Quantifying Levels of Automation

THESIS FOR THE DEGREE OF DOCTOR OF TECHNOLOGY

Quantifying Levels of Automation ‐

to enable competitive assembly systems ÅSA FASTH

Department of Product and production Development CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2012

Quantifying Levels of Automation ‐

to enable competitive assembly systems

ÅSA FASTH ISBN 978-91-xxxx-xxx-x

©ÅSA FASTH. 2012

Doktorsavhandlingar vid Chalmers tekniska högskola. Ny serie Nr 3320 ISSN 0346-718X

Department Product and production Development Chalmers University of Technology SE-412 96 Gothenburg Sweden Telephone + 46 (0)31-772 3686 (0730346288) E-mail: [email protected]

Cover: [en förklarande bildtext till eventuell omslagsbild, med sidhänvisning till utförligare information i uppsatsen.]

Chalmers Reproservice Gothenburg, Sweden 2012

Quantifying Levels of Automation - to enable competitive assembly systems ÅSA AB FASTH Department of Product and production development Chalmers University of Technology

ABSTRACT Production companies frequently have to meet demands and requirements, both internal and external, which trigger a plan for change in different production areas. Assembly systems of today are heading towards more customised production, i.e. smaller batches, shorter cycle times and increased number of variants. As a result, companies have to find more flexible methods for assembling their products and become more proactive in the assembly system itself. Identifying new strategies becomes vital and can be achieved by designing the assembly system in a structured way with the most advantageous cognitive and mechanical Level of Automation. The aim of this thesis is to show that by quantifying, measuring and analysing physical and cognitive Levels of Automation it is possible to enable competitive assembly systems. ‘If it is not Quantifiable it is not true’ – this is a common statement among engineers who rather use numbers than words when describing a phenomenon. This thesis will discuss the phenomenon of Levels of Automation from both a quantitative and qualitative point of view. Furthermore it will discuss whether it is necessary to have more than one level or dimension of automation and if so, what benefits this creates for industry when measuring and analysing Levels of Automation in their assembly systems. ‘The future assembly systems will consist of highly skilled operators...’ In order to choose and use the right level of automation, the choice itself has to be structured and well based. This thesis will discuss the importance of a structured method and an easy-to-use tool to visualise and quantify the levels of automation in the current state of a system. Furthermore, it will show how this could be used in a future analysis. Lastly, this thesis will discuss the effects of changing mind-set from primary look at cost and productivity to also consider other parameters that could influence the system in order to be competitive.

Keywords: Levels of Automation, LoA, Assembly System, Quantify, Competitive, Cognitive

ACKNOWLEDGMENT Five years of education to become a PhD will soon come to an end and new exciting doors will open. During this time many people have joined my path for a short or longer period of time. Therefore, there are a lot of people that deserve to be acknowledged for inspiration, interesting and inspiring discussions and kindness about being disturbed and answering a lot of questions. I cannot mention you all in person, but thanks for all support! Some persons have contributed in a special way and those I’d like to thank below: First of all I would like to give special thanks to my supervisor and examiner Prof. Johan Stahre; thank you for all the inspiring discussions and debates about my work. It has helped me tremendously in my growth as a researcher. You have been with me from the start of my journey and I hope that it will continue for many more years to come! Thank you! I want to thank Anna Davidsson for tutoring me for the last months; it helped me a lot in my writing! I want to acknowledge the organizations that have supported my research on a financial level, VINNOVA (Swedish agency for innovation systems), the Sustainable Production Initiative and the Production Area of Advance at Chalmers. I also thank the colleges in all research projects that I have been involved in: ProAct, Simter, MyCar, COMPLEX and The operator of the future. Thank you for fruitful and inspiring discussions. A special thanks to all the companies that have supported the research, especially the personnel at Stoneridge Electronics AB for an inspiring production and inspiring mind-set! I want to give special thanks to all colleagues at my department for a warm and inspiring climate. Particular thanks are due to Dr. Anders Skoogh for the ‘every-day’ talk and discussions. This journey would not have been the same without all the PhD-student friends within ProViking and PADOK – thanks to all of you! Special thanks to a group of extraordinary ladies: Lic. Kerstin Dencker, PhD. Jessica Bruch, Marie Jonsson, PhD. Victoria Rogstrand, Senior lecture Kerstin Johansen, Docent Carin Andersson and Prof. Lena Mårtensson thank you all for inspiration, discussions and good times! Thanks to all my friends who have believed in me. Finally I would like to thank my fiancé Johan, my daughter Tova and the rest of my family for helping and supporting me – love you all!

Åsa Fasth Guldheden, Gothenburg, February, 2012

LIST OF APPENDED PAPERS The results presented in this thesis are primarily based on the work in the following appended papers.

Paper 1

Fasth, Å. and Stahre, J. (submitted 29 June, 2011), Task allocation in assembly systems –Measuring and analyzing Levels of Automation, special issue (Journal of Theoretical Issues in Ergonomics Science)

Contribution Fasth, Å. initiated the paper and wrote the paper with Stahre, J. as a reviewer.

Paper 2

Fasth, Å. Stahre, J. and Dencker, K. (2008), Measuring and analyzing Levels of Automation in an assembly system. Proceedings of the 41st CIRP International Conference on Manufacturing Systems (ICMS), Tokyo, Japan.

Contribution Fasth, Å. initiated the paper and wrote the paper with Stahre, J. and Dencker, K. as reviewers. Fasth, Å. was the corresponding author and presented the paper at the conference. Paper 3

Fasth, Å., and Stahre, J., Does Levels of Automation need to be changed in an assembly system? - A case study. Proceedings of the 2nd Swedish Production Symposium (SPS), Stockholm, Sweden, 2008.

Contribution Fasth, Å. initiated the paper and wrote the paper with Stahre, J. as a reviewer. Fasth, Å. was the corresponding author and presented the paper at the conference.

Paper 4

Fasth, Å, Reviewing methods for analyzing task allocation in a production system (Accepted for publication), Journal of Logistic management

Paper 5

Fasth, Å., Bruch, J., Dencker, K., Stahre, J., Mårtensson, L. and Lundholm, T. (2010) Designing proactive assembly systems (ProAct) - Criteria and interaction between automation, information and competence, Asian International Journal of Science and Technology in production and manufacturing engineering (AIJSTPME), vol 2 issue 4, pp.1-13

Contribution Fasth, Å. initiated the paper and wrote the paper with the other authors as reviewers. Fasth, Å., Bruch, J., Dencker, K. collected the data and Fasth, Å. analysed it for the paper. Fasth, Å. was the corresponding author and presented the early version of the paper at the 42nd CIRP conference on manufacturing systems Grenoble, France, 2009. Paper 6

Fässberg, T., Fasth, Å., Hellman, F., Davidsson, A., and Stahre, J (2012), Interactions between complexity, quality and cognitive automation, 4th CIRP Conference On Assembly Technology Systems, Ann Arbor, USA

Contribution Fässberg, T., Fasth, Å., Hellman, F., initiated the paper and wrote the paper with Davidsson, A., and Stahre, J. as reviewers.

LIST OF ADDITIONAL PAPERS 1. FASTH, Å., PROVOST, J., STAHRE, J. & LENNARTSON, B. (2012) From task allocation towards resource allocation when optimizing assembly systems. IN PATRAS, U. O. (Ed.) 45th CIRP Conference of Manufacturing Systems (CMS). Athen, Greece. 2. MATTSSON, S., GULLANDER, P., HARLIN, U., BÄCKSTRAND, G., FASTH, Å. & DAVIDSSON, A. (2012) Perceived Production Complexity at Assembly Stations - A Case Study. IN PATRAS, U. O. (Ed.) 45th CIRP Conference of Manufacturing Systems (CMS). Athen, Greece. 3. FÄSSBERG, T., FASTH, Å., HELLMAN, F., DAVIDSSON, A. & STAHRE, J. (2012) Interaction between Complexity, Quality and Cognitive Automation. 4th CIRP Conference on Assembly Technologies and Systems (CATS). Ann Arbor, USA. 4. FÄSSBERG, T., FASTH, Å. & STAHRE, J. (2012) Classification of carrier and content of information. 4th CIRP Conference on Assembly Technologies and Systems (CATS). Ann Arbor, USA. 5. PROVOST, J., FASTH, Å., STAHRE, J., LENNARTSON, B. & FABIAN, M. (2012) Human operator and robot resource modelling for planning purposes in assembly systems 4th CIRP Conference on Assembly Technologies and Systems (CATS). Ann Arbor, USA. 6. FASTH, Å. (accepted for publication) Reviewing methods for analysing task allocation in a production system. International journal of logistic management 7. MATTSON, S., FÄSSBERG, T., STAHRE, J. & FASTH, Å. (2011) MEASURING INTERACTION USING LEVELS OF AUTOMATION OVER TIME. 21st International Conference on Production Research. Stuttgart, Germany 8. FÄSSBERG, T., HARLIN, U., GARMER, K., GULLANDER, P., FASTH, Å., MATTSSON, S., DENCKER, K., DAVIDSSON, A. & STAHRE, J. (2011b) AN EMPIRICAL STUDY TOWARDS A DEFINITION OF PRODUCTION COMPLEXITY. 21st International Conference on Production Research, Stuttgart, Germany 9. FASTH, Å., MATTSSON, S., FÄSSBERG, T., STAHRE, J., HÖÖG, S., STERNER, M. & ANDERSSON, T. (2011) Development of production cells with regard to physical and cognitive automation - A decade of evolution International Symposium on Assembly and Manufacturing (ISAM 11). Tampere, Finland. 10. FASTH, Å. & STAHRE, J. (Submitted) Task allocation in assembly systems - Measuring and analysing Levels of Automation. Special issue 11. FASTH, Å. (2011) Comparing methods for design and measurement of production systems, Proceedings of the 4th Swedish Production Symposium (SPS). Lund, Sweden 12. FÄSSBERG, T., FASTH, Å., MATTSSON, S. & STAHRE, J. (2011) Cognitive automation in mass customised assembly systems. Proceedings of the 4th Swedish Production Symposium (SPS) 13. FASTH, Å., STAHRE, J. & DENCKER, K. (2010b) Level of automation analysis in manufacturing systems. Proceedings of the 3rd international conference on applied human factors and ergonomics. Miami, Florida, USA. 14. DENCKER, K., MÅRTENSSON, L. & FASTH, Å. (2010) ”The operator saves our day?” - Why do we need the operator? Proceedings of the 3rd international conference on applied human factors and ergonomics. Miami, Florida, USA. 15. FASTH, Å. & STAHRE, J. (2010) Concept model towards optimising Levels of Automation (LoA) in assembly systems. Proceedings of the 3rd CIRP Conference on Assembly Technologies and Systems, Trondheim, Norway 16. NORDIN, G., FÄSSBERG, T., FASTH, Å. & STAHRE, J. (2010) iPod Touch - an ICT tool for assembly operators in factories of the future? - Technical solutions and requirements. 3rd CIRP Conference on Assembly Technologies and Systems (CATS), Trondheim, Norway

17. FÄSSBERG, T., NORDIN, G., FASTH, Å. & STAHRE, J. (2010) iPod Touch - an ICT tool for assembly operators in factories of the future?. Proceedings of the 43rd CIRP International Conference On Manufacturing Systems (ICMS). Vienna, Austria. 18. FASTH, Å., BRUCH, J., DENCKER, K., STAHRE, J., MÅRTENSSON, L. & LUNDHOLM, T. (2010a) Designing proactive assembly systems (ProAct) - Criteria and interaction between automation, information, and competence Asian International Journal of Science and Technology in production and manufacturing engineering (AIJSTPME), 2 (4), 1-13. 19. DENCKER, K. & FASTH, Å. (2009) A MODEL FOR ASSESSMENT OF PROACTIVITY POTENTIAL IN TECHNICAL RESOURCES, Proceedings of DET2009 6th International Conference on Digital Enterprise Technology, Hong Kong 20. JOHANSSON, B., FASTH, Å., STAHRE, J., HEILALA, J., LEONG, S., LEE, Y.-T. T. & RIDDICK, F. (2009) Enabling Flexible Manufacturing Systems by Using Level of Automation as Design Parameter. Winter Simulation Conference 21. DENCKER, K., FASTH, Å., STAHRE, J., MÅRTENSSON, L., LUNDHOLM, T. & AKILLIOGLU, H. (2009) Proactive assembly systems-realising the potential of human collaboration with automation. Annual Reviews in Control, In Press, Corrected Proof. 22. FASTH, Å., LUNDHOLM, T., MÅRTENSSON, L., DENCKER, K., STAHRE, J. & BRUCH, J. (2009) Designing proactive assembly systems – Criteria and interaction between Automation, Information, and Competence. Proceedings of the 42nd CIRP conference on manufacturing systems Grenoble, France 23. FASTH, Å. (2009) Measuring and Analysing Levels of Automation in assembly systems - For future proactive systems. Licentiate thesis, Product and production Development, Production systems, Gothenburg, Chalmers University of Technology 24. JOHANSSON, B., FASTH, Å., STAHRE, J., HEILALA, J., LEONG, S., LEE, Y.-T. T. & RIDDICK, F. (2008) Simulation Tool for Design of Flexible Manufacturing Systems Using Level of Automation as Design Parameter. Winter Simulation Conference. 25. FASTH, Å. & STAHRE, J. (2008) Does Levels of Automation need to be changed in an assembly system? - A case study. Proceedings of the 2nd Swedish Production Symposium (SPS), Stockholm, Sweden. 26. FASTH, Å., STAHRE, J. & DENCKER, K. (2008 -a) Analysing changeability and time parameters due to levels of Automation in an assembly system. Proceedings of the 18th conference on Flexible Automation and Intelligent Manufacturing – FAIM, Skövde, Sweden 27. LIND, S., KRASSI, B., JOHANSSON, B., VIITANIEMI, J., HEILALA, J., STAHRE, J., VATANEN, S., FASTH, Å. & BERLIN, C. (2008) A Production Simulation Tool for Joint Assessment of Ergonomics, Level of Automation and Environmental Impacts. Proceedings of the 18th conference on Flexible Automation and Intelligent Manufacturing - FAIM. Skövde, Sweden. 28. FASTH, Å., STAHRE, J. & DENCKER, K. (2008 -b) Measuring and analysing Levels of Automation in an assembly system. Proceedings of the 41st CIRP conference on manufacturing systems Tokyo, Japan 29. DENCKER, K., STAHRE, J., FASTH, Å., GRÖNDAHL, P., MÅRTENSSON, L. & LUNDHOLM, T. (2008) Characteristic of a Proactive Assembly System. Proceedings of the 41st CIRP conference on manufacturing systems., Tokyo, Japan 30. FASTH, Å., FROHM, J. & STAHRE, J. (2007) Relations between Performers/parameters and Level of Automation. IFAC workshop on manufacturing modelling, management and control, Budapest, Hungary

TABLE OF CONTENTS 1 

Introduction ....................................................................................................................... 1  1.1 

Background ......................................................................................................................... 1 

1.2 

Definitions of Automation .................................................................................................. 2  o

1.2.1 When to automate? ..................................................................................................... 2 

o

1.2.2 The importance of quantifying ................................................................................... 3 

1.3 

Research aim and objective................................................................................................. 4 

1.4 

Research questions .............................................................................................................. 4 

1.5 

Delimitations ....................................................................................................................... 4 

1.6 

Outline of thesis .................................................................................................................. 5 

2 

Research approach ............................................................................................................ 7  2.1 

Research methodology ........................................................................................................ 7 

2.2 

Research process ................................................................................................................. 7  o

2.2.1 Theoretical collection ................................................................................................. 8 

o

2.2.2 Empirical collection and analysis ............................................................................... 9 

o

2.2.3 Theoretical and practical contribution...................................................................... 11 

3 

Theoretical framework ................................................................................................... 13  3.1 

Assembly systems ............................................................................................................. 13 

3.2 

Allocation of function, task or resource? .......................................................................... 14 

3.3 

Levels of automation ......................................................................................................... 15 

3.4 

Assessment methods ......................................................................................................... 17  o

3.4.1 Qualitative versus quantitative ................................................................................. 17 

o

3.4.2 DYNAMO ................................................................................................................ 21 

3.5 

Effects ............................................................................................................................... 21  o

3.5.1 Time ......................................................................................................................... 21 

o

3.5.2 Flexibility ................................................................................................................. 22 

o

3.5.3 Proactivity ................................................................................................................ 22 

o

3.5.4 Complexity ............................................................................................................... 23 

4 

Results .............................................................................................................................. 25  4.1 

Results related to RQ1 ...................................................................................................... 26  o

4.1.1 Paper 1...................................................................................................................... 26 

o

4.1.2 Paper 4...................................................................................................................... 28 

4.2 

Results related to RQ 2 ..................................................................................................... 30  o

4.2.1 Paper 2...................................................................................................................... 30 

o

4.2.2 Paper 3...................................................................................................................... 32 

o

4.2.3 Paper 5...................................................................................................................... 34 

o

4.2.4 Paper 1...................................................................................................................... 35 

4.3 

Results related to RQ 3 ..................................................................................................... 36  o

4.3.1 Paper 1...................................................................................................................... 36 

o

4.3.2 Paper 5...................................................................................................................... 38 

o

4.3.3 Paper 6...................................................................................................................... 39 

5 

Discussion......................................................................................................................... 41  5.1 

Summary of the appended papers related to the RQs ....................................................... 42  o

5.1.1 .. Why is it important to quantify Levels of Automation (LoA) in an assembly system context? ............................................................................................................................. 42 

o

5.1.2 How should LoA be measured and analysed in assembly systems? ........................ 42 

o

5.1.3 What are the expected effects of analysing and changing LoA?.............................. 42 

5.2 

Contribution matrix ........................................................................................................... 43  o

5.2.1 Practical contribution from a scientific perspective ................................................. 43 

o

5.2.2 Theoretical contribution from a scientific perspective ............................................. 45 

o

5.2.3 Theoretical contribution from an industrial perspective .......................................... 47 

o

5.2.4 Practical contribution from an industrial perspective............................................... 48 

5.3 

Empirical and theoretical collection and analysis ............................................................. 50  o

5.4  6 

5.3.1 Triangulation perspective ......................................................................................... 50  Future work ....................................................................................................................... 51  Conclusion........................................................................................................................ 53 

References ........................................................................................................................................... 55  Appended papers ................................................................................................................................. 63 

LIST OF FIGURES  Figure 1 Research process, inspired by (Eisenhardt, 1989) .................................................................... 8  Figure 2 Case studies performed between 2007 and 2011 ...................................................................... 9  Figure 3 Relations between the theoretical and practical contribution from a scientific and industrial perspective ............................................................................................................................................. 11  Figure 4 Relations between production and manufacturing systems, and assembly operations (Groover, 2001) ..................................................................................................................................... 13  Figure 5 structuring levels and views of a factory (Wiendahl, et al., 2007), edited. ............................. 14  Figure 6 illustrates a division of complexity into PERCEIVED, objective and subjective complexity24  Figure 7 Matrix illustrating physical and cognitive levels of automation (LoA) .................................. 27  Figure 8 Result of the method review (Fasth, Accepted for publication) ............................................. 29  Figure 9 An overview of the 4 phases in the DYNAMO ++ methodology (Fasth and Stahre, 2008) 32  Figure 10 Task and possible operation optimisation ............................................................................. 33  Figure 11 Methodologies used in the ProAct Loop and the different levels of measurement .............. 34  Figure 12 Concept model, further developed from DYNAMO++ (Fasth and Stahre, 2010)................ 35  Figure 13 Observed LoA and cycle time (C/T) value for station 3 (S3) ............................................... 36  Figure 14 Creating volume flexibility through task allocation.............................................................. 37  Figure 15 Effort needed to change loa within a system ........................................................................ 38  Figure 16 The investigated relations ..................................................................................................... 39  Figure 17 The selected area with total number of stations and selected stations .................................. 39  Figure 18 LoA measurement for all the seven stations ......................................................................... 40  Figure 19 Relations between the theoretical and practical contribution from a scientific and industrial perspective ............................................................................................................................................. 43  Figure 20 evolution from DYNAMO to the concept model ................................................................. 44  Figure 21 Observed tasks from ten case studies, documented in the LoA matrix................................. 48 

  LIST OF TABLES  Table 1 LoA-scales for computerized and mechanized tasks within manufacturing ................................ 10  Table 2 Summary of definitions of Levels of Automation (Frohm, et al., 2008), edited ...................... 17  Table 3 Features of qualitative and quantitative methodology (Miles and Huberman, 1994) .............. 18  Table 4 Definitions of Flexibility .......................................................................................................... 22  Table 5 How the papers contribute to the RQs and theoretical areas .................................................... 25  Table 6 summary of the assessment methods and focus areas .............................................................. 28  Table 7 Current state in six case studies ................................................................................................ 32  Table 8 Triggers for change and lean awareness................................................................................... 33

 

 

 

1 INTRODUCTION The aims of this chapter are to:    

Introduce the reader to the research area and issues of Automation Describe research aim Formulate the research questions Describe the delimitations of the research scope in this thesis

1.1 BACKGROUND  In order to maintain a sustainable production in an increasingly globalised industry, current traditions for design and usage of automation may not be adaptable to the needs and future challenges that the industry is facing. Manufacturing research has the potential to develop technologies for highly competitive manufacturing, adding value and sustainability by changing the orientation and the criteria of optimisation to support the structural change of manufacturing (Westkämper, 2008). Rapid changes of demands and requirements, both internal and external, frequently trigger plans for change in different manufacturing areas, i.e. lower product and production costs, higher quality and shorter throughput time. This demands a high degree of flexibility (Chryssolouris, 2006, Koren, et al., 1999) and more dynamic decision making later in the production chain. Flexibility and changeability are key enablers for meeting the challenges of a global market (Wiendahl, et al., 2007). Smaller batches and shorter time limits for set-up between products are normal demands on the assembly systems caused by increasing numbers of product variants, i.e. mass customisation. This results in increased amounts of information to and from the assembly personnel since information regarding every product variant needs to be available (Fässberg, et al., 2010). Furthermore the amount of information needed by the operators is individual and dependent on their level of expertise (Fjällström, 2007). As a result, there is a need for increasingly flexible methods for assembling products and means to make assembly systems more proactive. One solution could be to consider different levels of automation, both cognitive and physical. In order to make the right decisions regarding the level of automation and obtain a competitive assembly system, companies have to know their present system, the future changes and how and what to improve in order to make the assembly system as effective as possible. Evolution of an assembly system is argued to be one of several steps towards a competitive position for a manufacturing company (Säfsten, 2002). A statement by Percy Barnevik in 1991 (Barnevik, 1991) is still applicable today: “Modern technology has made it easy to copy products, but a process is not that easy to copy. A superior process can therefore give a more durable competitive advantage.” An addition to this is the importance of highly skilled resources: “... we expect that team assembly plants will be populated almost entirely by highly skilled problem solvers whose task will be to think continually of ways and means to make the system run more smoothly and productively.” (Womack, et al., 1990) Throughout history, different strategies have been used in order to meet internal and external demands. The future is always hard to predict. If aiming for a more sustainable production, then a more effective and adaptive assembly system needs to be developed. One solution could be to consider different automation solutions, not necessarily to increase the automation, but to investigate pros and cons of the use of humans and technique. 1

1.2 DEFINITIONS OF AUTOMATION  A broad spectrum of assembly systems with varying degrees of automation exists in industry today, such as manual assembly, semi-automatic assembly or automatic assembly (Rampersad, 1994). The term ‘automated assembly’ refers to the use of mechanised and automated devices to perform the various assembly tasks in an assembly line or cell (Groover, 2001). Depending on the context, the term automation has different meanings:“...in their language, or in their region of the world or their professional domain, automation has a unique meaning and we are not sure it is the same meaning for other experts”(Nof, 2009). Furthermore, because automaton is context-dependent and because it describes technology that facilitates human performance, cognitive or physical, what is considered automation will therefore change over time. The term automation evolves from ‘automatos’ in Greek which means acting by itself (Williams, 2009). The scope of this thesis will be limited to the final assembly system context within production systems and therefore the definitions and discussions about automation will also be limited to that context. Automation can be defined as ‘the execution by a machine agent (usually a computer) of a function that was previously carried out by a human’ (Parasuraman and Riley, 1997). This could be related to the more cognitive tasks of an assembly system. Another definition more related to the physical automation is: ‘Automation is a way for humans to extend the capability of their tools and machines’ (Williams, 2009). A definition that could contain both the cognitive and physical automation is: a technology by which a process or procedure is accomplished without human assistance (Groover, 2001).

1.2.1 WHEN TO AUTOMATE? Mass customisation puts great strains on the product developers and the system designers. There are many different solutions available on the market today in terms of technologies and methodologies. Companies have to be at the front of the evolution in their field by adapting to future changes and trends, on both macro and micro level. But it is costly to invest in resources (both human and machines) and technology that will not be used to their full potential. Appropriate Levels of Automation (LoA), both cognitive and mechanical, must therefore be selected to meet the increased information flow and to avoid over or under automated systems. This means that suitable allocation of tasks between resources (human operators and machines) and technique has to be made and must be able to be dynamically changeable over time. One of the primary design dilemmas engineers and designers face is determining what level of automation should be introduced into a system that requires human intervention (Parasuraman and Riley, 1997). Do companies think automation is an important factor to consider in order to be competitive? Until recently, the primary criteria for applying automation were technological feasibility and cost. To the extent that automation could perform a function more efficiently, reliably, or accurately than the human operator, or merely replace the operator at a lower cost, automation has been applied at the highest level possible (Parasuraman and Riley, 1997). In line with Parasuraman and Riley, results from an industrial Delphi survey conducted in 2006 (Frohm, 2008) show that the top three answers regarding benefits with automation or when to automate were: cost savings, to get higher efficiency and to increase competitiveness. On the other hand, the top three answers regarding disadvantages with automation or when not to automate were: too many products or variants, investment cost, and adapting the product for manufacturing (Frohm, 2008).

2

In order for industry both to stay competitive and to handle the increasing demand of mass customisation, the question when to automate becomes a little more complex. The assembly area is where most manual work takes place in the process. Humans and technology have to cooperate in order to simplify the job and make the overall system more efficient and productive. Companies must obtain deeper knowledge about new production solutions and be willing to evaluate them with reference to their own production in order to create a long-term sustainable system. One of the considerations preventing the total removal of human operators from systems has been the perception that humans are more flexible, adaptable, and creative than automation and thus are better able to respond to changing or unforeseen conditions (Parasuraman and Riley, 1997). Most system design tools focused solely on the physical system (Parasuraman, et al., 2000) towards a more flexible assembly (Feldmann and Slama, 2001) but all resources contributed to flexibility. Completely automated systems almost always have a human operator somewhere, at some level (Dekker and Woods, 2002), so Chapanis’ dream in 1970 (Chapanis, 1996), ‘to automate everything you possibly can towards autonomous systems’, remains a dream, forty years later. Further, current research (Parasuraman and Wickens, 2008, Sheridan and Parasuraman, 2006, Stoessel, et al., 2008) argues that operators still have not been surpassed by conventional automation in terms of flexibility and high product variation. Therefore, operators should be used for more than supervision of machines and be integrated and seen as complementary to machines rather then divide the resources in manmachine thinking when performing task allocation in system production design (Hancock and Chignell, 1992, Hou, et al., 1993, Jordan, 1963, Kantowitz and Sorkin, 1987, Sheridan and Parasuraman, 2006). In comparison with technical capabilities, human capabilities, human performance and cognition in automated systems are much less frequently described in literature or discussed in public forums (Parasuraman and Riley, 1997). Automation of physical functions has freed humans from many timeconsuming and labour-intensive activities. However, full automation of cognitive functions such as decision making, planning, and creative thinking remains rare (Parasuraman and Riley, 1997). Flexible technology cannot be effective without flexible operators and vice versa (Slack, 2005) so humans and machines/technique should be seen as complementary, rather than conflicting, resources when designing a man-machine system (Jordan, 1963). The automation should aim for extending the physical and cognitive capacity of people to achieve what might otherwise be impossible (Lee, 2008).

1.2.2 THE IMPORTANCE OF QUANTIFYING Existing methods used for designing assembly systems, do not consider LoA and different aspects regarding LoA as a prime parameter. To be able to compare current and future state and to turn the term and methods around level of automation from a “human factor” and design tool into an easy-touse engineer tool, a more quantified method is needed. The key issue is not who make the decision or on what level the decisions are made, but why the decisions are made and upon what facts (Winroth, 2006). Case studies (Fasth and Stahre, 2008, Säfsten, 2002) show that these kind of decisions are often informal and unstructured. In order to make the decisions more objective and more structured, it is important to have a quantitative method to measure and analyse LoA.

3

1.3 RESEARCH AIM AND OBJECTIVE  As discussed, there are a lot of challenges when designing or improving an assembly system with regard to automation. There is a need for a structured method to determine when and how much to automate. The aim of this thesis is to show that: By quantifying, measuring and analysing physical and cognitive Levels of Automation, competitive assembly systems are enabled.

1.4 RESEARCH QUESTIONS  Based on the aim, the following three research questions (RQs) have been formulated; RQ 1: Why is it important to quantify Levels of Automation (LoA) in an assembly system context? The way companies choose and use their resources has been an issue ever since the craftsman became an operator, and the manual work became more automated. This research question will discuss why automation needs to be put in a quantitative context. Furthermore, the RQ will discuss why both the physical and cognitive automation need to be addressed when measuring and analysing an assembly system. RQ 2: How should LoA be measured and analysed in assembly systems? When the Levels of Automation have been quantified, this quantification could be used for further task allocation in an assembly system. The tasks in the assembly system are measured and analysed in a structured way in order to avoid under- or over-automated systems. This research question will explain the evolution from a method towards a more logical model aiming for presenting the measures and analyses of LoA, i.e. a concept model. RQ 3: What are the expected effects of analysing and changing LoA? Cost is a common parameter used when improving production systems. The final research question addresses the possible presence of other parameters linked to LoA that would provide equally good or better results than just focusing on cost when deciding if the LoA should be changed.

1.5 DELIMITATIONS       

The scope of this thesis is limited to the final assembly area of a production system. o Further, the shop-floor level is the primary scope. Product design will not be covered in this thesis Cost as a primary parameter will not be discussed in this thesis Even though production logistics is important to consider when redesigning a system it will not be discussed in detail in this thesis The thesis does not base its cognitive automation discussion on psychological theory The thesis does not discuss physical automation based on control theories, robotics engineering or computer science

4

1.6 OUTLINE OF THESIS  Chapter 1 provides the reader with a short introduction of the research area. Further, the research aim and research questions are presented. Chapter 2 describes the research approach, i.e. the overall methodology used in order to reach the aim. Further, the methods used in the different papers in order to answer the research questions are presented. Chapter 3 reviews earlier research and definitions that are used in the later part of the thesis and in the appended papers. Chapter 4 presents the results from the appended papers correlated to each research question. Chapter 5 discusses the overall aim, the research question in relation to the theoretical framework and the results from the appended papers. Chapter 6 provides conclusions regarding the research aim.

5

6

2 RESEARCH APPROACH This chapter presents the research design, thus providing a description of the diverse methods that were used to achieve the aim of the thesis.

2.1 RESEARCH METHODOLOGY  Research methodology means the strategy, path or design of the research, i.e. the choice of methods used to reach the aim (Crotty, 1998). The methods are the tools, i.e. techniques or procedures to analyse the gathered data and information related to the research questions (Crotty, 1998). The main methodology used in this thesis is applied research, which means that empirical data from industrial case studies are a major part of the research results. Independent related studies (Wilkinson, 1991) are also used as part of the research methodology, i.e. the theoretical framework and methods used are the same in all case studies but the research questions and discussion differ. This approach could also be referred to as a triangulation approach (Olsen, 2004), which aims to increase the quality of the research by mixing data or methods so that diverse viewpoints and standpoints cast light upon a topic. According to Deniz (Dezin, 1970), there are four different types of triangulation collections; 1. 2. 3. 4.

Data triangulation – Use of variety of data sources in a study Investigator triangulation – Use of several different researchers or evaluators Theory triangulation – Use of multiple perspectives to interpret a single set of data Methodology triangulation – Use of multiple methods to study a single problem or phenomenon

The use of the different triangulation collections will be further explained in the sections below and within the discussion chapter, relation to the papers. But, as a summary, the different data sources used were production data gathered from production engineer at the companies and assessment of Levels of Automation within the stations in chosen production cells. In the case studies multiple researchers and in some cases master students collected the data and the information, explained more in detail in the empirical collection section. The theory collected is both from a quantitative perspective and a qualitative perspective (results discussed in paper 3). A multiple perspective has also been used when it comes to explain and discuss the view and use of Levels of Automation in assembly systems and in other areas such as control rooms etc (results discussed in paper 1). Further, paper 4 discusses the phenomenon of operators’ action space, viewed from three different aspects (information flow to and from the operators group, level of competence and use of different levels of automation.). The gathering of this information was done by three different researchers which also resulted in an investigator triangulation.

2.2 RESEARCH PROCESS  The research in this thesis includes both a deductive research process (Patel and Davidsson, 2003) and an inductive research process (Starrin and Svensson, 1994). The process is illustrated in Figure 1. The first step is to formulate an overall aim with the research (purpose of research) in order to focus the theoretical and empirical collection; without such a research focus, it is easy to become overwhelmed by the volume of data (Eisenhardt, 1989). From both a theoretical collection and from empirical collections, a possible academic (theoretical) and industrial (empirical) gap is determined. These gaps are then formulated into a hypothesis which becomes RQs later in the process. The hypothesis is then synthesised (i.e. methodologies and methods are developed and validated within industrial case studies). Finally a more general methodology is developed and effects of its use are determined (theoretical and practical contribution). 7

The research process and the different types of triangulation used are illustrated in Figure 1.

FIGURE 1 RESEARCH PROCESS, INSPIRED BY (EISENHARDT, 1989)

The following sections will describe the different areas and methods used in the research process. The following sections will describe the different areas and methods used in the research process.

2.2.1 THEORETICAL COLLECTION According to (Wacker, 1998) a theory (or theoretical framework) contains four components: (1) a domain where the theory applies, (2) definitions of terms or variables, (3) a set of relationships of variables and (4) specific predictions and factual claims. Four areas have been chosen as a base for the theoretical framework and are connected to the components as follows: The domain in this thesis is defined as Assembly systems. Definitions of terms and variables are related to the theoretical definitions of Levels of Automation (LoA) and the definition of set of relationships of variables are related to different effects usually measured and improved in an assembly system (excluding cost), i.e. time, productivity, flexibility, complexity. Further, the assessment variables could be seen as a relation between the different effects and LoA based on results from a method performing qualitative and/or quantitative assessments. The fourth component, specific predictions and factual claims (4), will be brought up in each theoretical area and in the discussion chapter. The theoretical framework was gathered through secondary data (Merriam, 1994), i.e. books, papers (conference and journals), theses and reports.

8

2.2.2 EMPIRICAL COLLECTION AND ANALYSIS A case study (or empirical collection) is done in order to investigate a current phenomenon within its real-life context. Case studies typically combine data collection methods such as archives, interviews, questionnaires, and observations. The evidence may be qualitative (e.g., words), quantitative (e.g., 534 numbers), or both (Eisenhardt, 1989). When the boundaries between the phenomenon and context are not clearly evident multiple sources of evidence are used (Yin, 2003). In multiple case studies, detailed information is gathered at several sites, within the same company or at different companies (Flynn, et al., 1990). The empirical collection used in this thesis is thirteen industrial case studies that have been carried out from 2007 to 2011. Figure 2 illustrates when the case studies have been conducted, the relation to the appended papers and in what phase of the method development (i.e. DYNAMO++) they belong. The author has visited all companies and been part of all case studies, either as first part participant or as tutor for student thesis*

FIGURE 2 CASE STUDIES PERFORMED BETWEEN 2007 AND 2011

If there are enough cases, some forms of inferential statistical analysis are possible. The choice and number of cases should be decided with reference to the studied phenomena that are being examined, the context and the research questions posed (Easton and Harrison, 2004). In analysing the data, similarities and differences between companies are noted and documented, to the extent possible. For example, in (Fasth and Stahre, 2008), six of the thirteen case studies are compared in terms of triggers for change, lean awareness and production layout. The empirical collection in the case studies has contained four different types of qualitative and quantitative methods, i.e. Observations, Structured interviews, Measurements of cognitive and physical LoA and Workshops. The methods are explained in the following sections. Observations In observational studies, the researcher can take either the role of non-participant observer (outside observer) or the role of participant observer (inside observer) (Sekaran, 2000) (Flynn, et al., 1990). The approach in all case studies has been outside observation. Outside observation uses a neutral observer to collect data, often employing some methods for ensuring that data are collected systematically (Flynn, et al., 1990). The methods used for collecting data systematically have been interviews, measures and workshops. In some case studies, Value-Stream-Mapping (VSM) has been used in order to map the flow within the company.

9

Structured interviews A structured interview involves the use of a form, which specifies questions to be used (Flynn, et al., 1990). The relevant questions and the order in which they are posed are decided in advance (Merriam, 1994). Other questions may be asked, as well, based on the direction of the conversation; however, certain questions are standard. Structured interviews permit comparisons between interviewees, without sacrificing the depth of the personal interview. Structured interviews were performed in most of the case studies in the present thesis. All interviews were recorded and then transcribed. According to (Flynn, et al., 1990), the quality of the interview is raised significantly if the researcher does not have to take meticulous notes. Other questions were covered by means of an open face-to-face interview with men and women from different professional categories, i.e. operators, production technicians, production logistics and production managers. Each interview took approximately 15-45 minutes. In-depth results and analysis from the interviews can be read about in other papers (Fasth, 2009, Fasth and Stahre, 2008). Assessment of cognitive and physical Levels of Automation (LoA) In order to measure cognitive (information and control) and physical (mechanical and equipment) levels, LoA, a taxonomy developed by Frohm (Frohm, 2008), was used. The taxonomy has seven steps for each type of LoA and different examples are used as guidelines, illustrated in Table 1. TABLE 1 LOA-SCALES FOR COMPUTERIZED AND MECHANIZED TASKS WITHIN MANUFACTURING

Workshops Workshops were performed at the end of each case study in order to present the result from the observations and measures, but also to get more information and ideas about the methodology and the system. The participants at the workshops were almost always the same that had been participating in the interviews (at some companies there could be different operators participating in the workshops and at interviews depending on availability from production). The number of participants was from 5 to 15 persons.

10

Validity Validity is an assurance that the right thing is measured or to what degree the results are connected to reality (Merriam, 1994). Reliability could be explained as the accuracy of the measures (Sekaran, 2000) or to what degree the measures are repeatable (Merriam, 1994). According to Yin (Yin, 2003), validity could be determined by discussing four areas; 1. 2. 3. 4.

Construct validity – constructing correct measures for the concept that is being studied Internal validity – establishing a causal relation External validity – establishing a domain in which the study could be generalised Reliability - ensuring that the operation of the study could be repeated with the same result

In the thesis validation was done foremost by the DYNAMO++ methodology and how to measure and analyse the cognitive and physical LoA. Because the research has been performed in multiple case studies within industry it is always hard to have 100% reliability because the industry and the environment are changing over time. A further discussion about validation of the methods used is done in Chapter 5.

2.2.3 THEORETICAL AND PRACTICAL CONTRIBUTION The aim of this thesis could be seen from both a theoretical and a practical contribution. Theories are important for the social and natural sciences because they make possible robust explanations of previously or currently observed phenomena, and because they are points of departure for forecasts about future phenomena (May, et al., 2009). The research loop ends by adding new theory (or new pieces of the puzzle (Kuhn, 1962)) to the scientific world, i.e. theoretical contribution (otherwise it is not science (Danermark, et al., 2003)). On the other hand, when performing applied science the practical contribution is as important as the theoretical. In order to get the theoretical and methodology triangulation perspective, discussed earlier in this chapter, a discussion about the theoretical and practical contributions seen from a scientific and industrial perspective will be presented. Figure 3 illustrates the different areas and an explanation of the author’s interpretation of each quadrant. The results or Research Answers (RAs) from the different quadrants will be shown in Chapter 4 and discussed in Chapter 5.

Theoretical

Practical

Definitions etc

Empirical validation

Operative and structure

Use and effect

Scientific

Industrial

FIGURE 3 RELATIONS BETWEEN THE THEORETICAL AND PRACTICAL CONTRIBUTION FROM A SCIENTIFIC AND INDUSTRIAL PERSPECTIVE

11

12

3 THEORETICAL FRAMEWORK This chapter presents four theoretical areas that have been chosen in order to reach the aim within this thesis.

3.1 ASSEMBLY SYSTEMS  There are numerous definitions of a manufacturing and production system. There are two main differences in the view of the definition of a manufacturing and production system. The first is that the manufacturing system is superior to the production system (http://www.cirp.net, 2008). The second definition, used in this thesis, is that the production system is superior to the manufacturing system: A production system is a collection of people, equipment and procedures organised to perform manufacturing operations at a company. A production system covers all steps in the chain from raw material to end costumer (Groover, 2001, Löfgren, 1983, Ståhl, 2006, Tangen, et al., 2008). Production systems can be divided into two categories or levels: facilities and manufacturing support systems; see Figure 4.

FIGURE 4 RELATIONS BETWEEN PRODUCTION AND MANUFACTURING SYSTEMS, AND ASSEMBLY OPERATIONS (GROOVER, 2001)

The processes that accomplish manufacturing involve a combination of machinery, tools, power and manual labour (Groover, 2001). The manufacturing processes can be divided into two different types: processing operations and assembly operations. Assembly operations generally have one material flow (Alting, 1978): Converging flow, corresponding to assembly or joining processes in which the shape is obtained by joining pre-shaped parts, without removing material to form a new entity. Components of the new entity are connected together either permanently or semi permanently (Groover, 2001). The assembly system operates as an integral part of the total production system, which in turn consists of all the elements that support the manufacturing system (Cochran, 1998). The assembly system can be characterised as a transformation system, for the purpose of transforming (Bellgran, 1998) or converting (input) (Andreasen, et al., 1983) geometrically corresponding parts into subassemblies (Rampersad, 1994) or finished products (output) through manual and/or automated work tasks (Andreasen, et al., 1983). This integration is achieved by a process where the necessary operations are integrated in respect of material, energy and information (Andreasen, et al., 1983) that is given 13

additional values, properties and qualities, so that the final state of the operations (Bellgran, 1998) is an organised unit working towards a goal (Andreasen, et al., 1983) that satisfies the previously declared need (Bellgran, 1998). According to numerous research (Seliger, et al., 1987) (Westkämper, 2006) (Nyhuis, et al., 2005) (Wiendahl, 2002), a manufacturing system can be described from a hierarchical perspective, where every system can be divided into elements or stations. These can be further divided into part-elements or tasks. Figure 5 illustrates the hierarchical perspectives used in this thesis.

FIGURE 5 STRUCTURING LEVELS AND VIEWS OF A FACTORY (WIENDAHL, ET AL., 2007), EDITED.

There are two structuring levels: working area and working place, and two views: the resource view proposal by Westkämper (Westkämper, 2006) and the space view proposal from Nyhuis (Nyhuis, et al., 2005) based on H-P Wiendahl (Wiendahl, 2002). Tasks within the working place has been added as a level in the model (Fasth, 2011), it is done to be able to in count task allocation.

3.2 ALLOCATION OF FUNCTION, TASK OR RESOURCE?  In order to optimize a system some kind of allocation has to be done. In most modern workplaces there is a close sharing of tasks between people and machines (Prince, 1985). Throughout history there have been numerous of definitions according how and when to allocate a task or a function and to whom, man or machine? One of the most common and debated attempt to allocate different tasks to different resources is fits list from 1951 (Fitts, 1951), which describes humans’ and machines’ differences. Fitts (Fitts, 1951) said that using the criteria in his list as the sole determinant of the allocation of functions was to lose sight of the basic nature of a system containing humans and machines. The Fitts list had little impact on engineering design practice because such criteria are overly general, non-quantitative, and incompatible with engineering concepts, and because they assume that functions will be performed by humans or machines alone (Prince, 1985). Jordan (Jordan, 1963) argued whether you could actually compare men and machines, and that the two should be seen as complementary, rather than conflicting, resources when designing a man-machine system. Sheridan (Sheridan, 1995) proposed to “allocate to the humans the tasks best suited to humans and allocate to the automation the task best suited to it. It is only when both humans and machine can do the same task, the question of task allocation becomes an issue (Hancock and Chignell, 1992). Contrary to the widely accepted urge towards autonomy, the real need is to provide an organic relationship for mutual benefit between the human and the machine (Tesar, 2002).

14

Today there are three different allocation approaches that are used in different stages and at different levels at companies, i.e. static and dynamic (adaptive) function allocation (Lagu and Landry, 2010), task allocation (Older, et al., 1997) and resource allocation (Isard, et al., 2009, Jain, et al., 1984). The first kind of allocation, function allocation (FA), is mostly used early in the design phase before prototypes or design specification has been determined (Wright, et al., 2000); this makes this approach a little hard to use in real industry. Wirstad (1979) has summed up the FA process succinctly: “Although the principle is clear, function allocation has never worked in practice'' (Fuld, 2000). Task allocation of some kind is usually made later, often during system implementation (Cherns 1986,1987) (Waterson, et al., 2002). Different tasks could have multiple recourses suitable for it; Fasth (Fasth and Stahre, 2010) shows a LoA matrix, where both the physical and cognitive automation, current and future needed, could be illustrated and analysed. This type of allocation is often a static allocation based on global optimization. Suitable allocation of tasks between resources (human operators and machines) and technique has to be made and must be able to be dynamically changeable over time. However, it is common that designers automate every subsystem that leads to an economic benefit for that subsystem and leave the operator to manage the rest (Parasuraman and Wickens, 2008). Generally, the manufacturing requirements of the product need to be matched to the capabilities of actual resources. This product/resource mapping means that one or more possible resources are identified for each product operation. The desired degree of flexibility will decide how many alternative resources are included in this resource allocation. Among the possible ones, a final choice has to be determined, e.g., by optimization (Lennartson, et al., 2010). The resource allocation can be based on a simplistic model such as available/unavailable resources. Such a model can be easily applied if we suppose that there is no resource breakdown, no maintenance task, etc. In that case, a resource could be allocated to an operation as soon as it is available. There is a need for a dynamic allocation that can take advantage of the access to instantaneous evaluation of the situations to choose the best allocation (Hoc, 2000). A case study that uses dynamically changeable Levels of automation (LoAs) (Fasth, et al., 2008 ) shows that it is possible to change from a human operator to a robot-cell and vice versa in order to achieve volume and route flexibility. The issue to be shown in this paper is how to model and simulate this dynamic allocation when alternative resources could be allocated to some operations. Difference in LoA implies that different resources need to be modelled as precisely as possible so that these models correspond to these LoA and not to a global resource. Furthermore, models of behaviour, knowledge and skills for robots and human must be considered in different ways in order to better fit the real resources.

3.3 LEVELS OF AUTOMATION  A common industrial predisposition is to consider automation investments as “binary” decision, even though a simple choice between humans or machines for a specific task may be suboptimal. Several development trends towards highly automated production and shop floor workplaces were seen during the 1980's and early 1990's. At that time the predominant task allocation strategy was "left-over allocation". Since the late 1990's trends are changing, much due to obvious shortcomings of automation to fulfil cost and flexibility expectations. Thus, to identify, implement, and maintain the correct level of automation in a controlled way could be a way to radically improve the effectiveness of a system. According to Frohm (2008), to make a manufacturing system as robust, flexible and adaptable as possible, the system must be resilient to process variations, such as the introduction of new products, tool changes, product disturbances etc. It 15

is thus important to understand how to obtain a balanced manufacturing system that has the proper mix of operators and machines in order to e.g. obtain the highest profit possible without suffering loss of product quality. One way to achieve this balanced manufacturing system is to separate the system description into two basic classes of activities, i.e. information handling and physical work. The next step would then be to describe the allocation of tasks within each class, i.e. the “level of automation”. Extensive amounts of research have been done in the area of levels of automation, emphasising different perspectives. Automation research can be divided into three main groups, i.e.   

Mechanical automation (Duncheon, 2002, Groover, 2001, Kern and Schumann, 1985, March and Mannari, 1981) Information and control automation (Bright, 1958, Endsley, 1997, Hollnagel, 2003, Parasuraman, et al., 2000, Parasuraman and Wickens, 2008, Sheridan, 1992). Combinations of physical/mechanical and information/cognitive automation (Frohm, 2008).

To determine what to automate, a classical task allocation strategy from 1951 (the MABA-MABA list) was proposed by Fitts (Fitts, 1951). It was an attempt to suggest allocation of tasks between humans and machines by treating them as system resources, each with different capabilities. Two examples, i.e. “Machines Are Better At” performing repetitive and routine tasks while “Men Are Better At” improvising and using flexible procedures. At the time, this was a revolutionary thought causing a lot of debate. Jordan (Jordan, 1963) argued whether you could actually compare man and machine, and that the two should be seen as complementary rather than conflicting resources when designing a man-machine system. Sheridan (Sheridan, 1995) proposed to “allocate to the human the tasks best suited to humans and allocate to the automation the task best suited to it.” But if tasks in which machines are better become automated and operators are still required to monitor the automation, maintaining full situation awareness (Endsley and Kiris, 1995), we might lose more than we gain. Fifty years after Fitts published his list, Hollnagel (Hollnagel, 2003) argues that the machine (or automation) has been used for three main purposes over the years (which is in line with Fitts), i.e. to ensure more precise performance of a given function; to improve stability of performance by relieving people of repetitive and monotonous tasks; and to enable processes to be carried out faster and more efficiently. So, do Fitts' thoughts still prevail, or has research turned towards Jordan’s argument? The decision matrix suggested by Prince (Prince, 1985) was partly in line with Fitts in that some tasks were better performed by machines and some better by humans. But interestingly Prince also defined a set of tasks where the same task could and should be performed both by humans and by machines. Further, when there is no single allocation, the different resources need support from each other, which is in line with Jordan’s argument. Hancock (Hancock and Chignell, 1992) argues that it is only when both human and machine can do the same task that the question of task allocation becomes an issue. In line with Jordan, previous research (Hancock and Chignell, 1992, Hou, et al., 1993, Kantowitz and Sorkin, 1987, Sheridan, 2000) agrees that the task allocation should be seen as complementary between man and machine rather than assigning tasks solely to one resource. Thus, suitable allocation of tasks between resources (human operators and machines) and technique has to be made and must be able to be dynamically changeable over time. However, it is common that designers automate every subsystem which leads to an economic benefit for that subsystem but leaves the operator to manage the rest (Parasuraman and Wickens, 2008). Parasuraman et al. (Parasuraman, et al., 2000) argue that automation design is not an exact science; however, neither does it belong in the realm of the creative arts, with successful design dependent upon the vision and brilliance of individual creative designers. 16

Table 2, show a summary of definitions of levels of automation from each decade, from 1950 to 2010.

TABLE 2 SUMMARY OF DEFINITIONS OF LEVELS OF AUTOMATION (FROHM, ET AL., 2008), EDITED

Author (Bright, 1958) (Amber and Amber, 1962) (Williams, 1977) (Sheridan, 1980) (March and Mannari, 1981) (Kern and Schumann, 1985) (Billings, 1997) (Endsley, 1997) (Satchell, 1998) (Parasuram an, et al., 2000) (Groover, 2001) (Duncheon, 2002) (Frohm, et al., 2008)

Definition of Levels of Automation Divides the levels depending on who initiates the control, the human (1-4),the human together with automation (5-8) or the automation (9-17) The extent to which human energy and control over the production process are replaced by machines Automation is the capability of causing a machine to carry out a specific operation on command from an external source The level of automation incorporates the issue of feedback, as well as relative sharing of functions in ten stages Automaticity is defined in six levels from conducting the tasks manual, without any physical support, to fully automated cognition with computer control Degree of mechanization is defined as the technical level in five different dimensions or work functions

Mechanical scale

Information and control scale

17

-

-

-

-

5

-

10

6

3 (9)

The level of automation goes from direct manual control to largely autonomous operation, where the human role is minimal The level of automation in the context of expert systems is most applicable to cognitive tasks such as ability to respond to, and make decisions based on, system information The level of automation is defined as the sharing between the human and machines, with different degrees of human involvement The interaction and task division between the human and the machine should instead be viewed as a changeable factor which can be called the level of automation. Level of mechanization can be defined as the manning level, with focus on the machines ‘Manual’ tasks are those in which humans are responsible for conducting the task. ‘Semi-automatic’ is a higher level of automation and involves automated alignment and application by a robot. ‘Automatic’, where material handling is also automated. The allocation of physical and cognitive tasks between humans and technology, described as a continuum ranging from totally manual to totally automatic

6 10

5

10+4

3 3(6) 7

7

3.4 ASSESSMENT METHODS  In Paper 3, ten assessment methods have been reviewed. A summary of the methods (except DYNAMO++) is presented in the following two sections, divided into quantitative and qualitative methods.

3.4.1 QUALITATIVE VERSUS QUANTITATIVE 17

The aim with a qualitative method is to interpret and understand a phenomenon (Patel and Davidsson, 2003). The research is done by asking the “journalist triangle” plus two questions, i.e. why, what, who + when and where. A quantitative method could be described as gathering data that will be transformed into figures and numbers to enable statistical analysis (Holme and Solvang, 1997). The two approaches have different pros and cons depending on in what context the method is going to be used. Of course there are extremists in this matter: "There's no such thing as qualitative data. Everything is either 1 or 0" (Kerlinger, 1960) and "All research ultimately has a qualitative grounding" (Campbell, December,1976) Table 3 shows the main differences between the qualitative and quantitative approach. TABLE 3 FEATURES OF QUALITATIVE AND QUANTITATIVE METHODOLOGY (MILES AND HUBERMAN, 1994)

Qualitative

Quantitative

The aim is a complete, detailed description.

The aim is to classify features, count them, and construct statistical models in an attempt to explain what is observed.

Researcher may only know roughly in advance what he/she is looking for.

Researcher knows clearly in advance what he/she is looking for.

Recommended during earlier phases of research projects.

Recommended during later phases of research projects.

The design emerges as the study unfolds.

All aspects of the study are carefully designed before data are collected.

Researcher is the data-gathering instrument.

Researcher uses tools, such as questionnaires or equipment, to collect numerical data.

Data are in the form of words, pictures or objects.

Data are in the form of numbers and statistics.

Subjective Individuals’ interpretation of events is important, e.g., using participant observation, in-depth interviews etc.

Objective Seeks precise measurement & analysis of target concepts, e.g., uses surveys, questionnaires etc.

Qualitative data are more 'rich', timeconsuming, and less able to be generalized.

Quantitative data are more efficient, able to test hypotheses, but may miss contextual detail.

Researcher tends to become subjectively immersed in the subject matter.

Researcher tends to remain objectively separated from the subject matter.

18

The following sections contain theory about nine different assessment methods used as a comparison to DYNAMO++ methodology in paper 4.

A model for types and levels of human interaction with automation (Parasuraman, et al., 2000, Parasuraman and Wickens, 2008) The model is primarily used to analyze ATC (Air Trafic Control) systems with the issue: given specific technical capabilities, which system functions should be automated and to what extent? The human performance consequences of specific types and levels of automation constitute the primary evaluative criteria for automation design using the model. Secondary evaluative criteria include automation reliability and the costs of action consequences. Such a combined approach— distinguishing types and levels of automation and applying evaluative criteria—can allow the designer to determine what should be automated in a particular system. The model does not prescribe what should and should not be automated in a particular system. Hence, the model provides a more complete and objective basis for automation design than approaches based purely on technological capability or economic considerations. Ten levels of automation of decision and action selection are used for task allocation. Complementary Analysis and Design of Production Tasks in Socio-technical Systems (KOMPASS) (Grote, 2004, Wäfler, et al., 1997) The main aim with the COMPASS method is to design production systems where a human has control over technology, i.e. automated systems. Expert analysis of existing systems is done based on three levels of analysis criteria; work system, human work tasks and human machine system. The method is built on the complementary principle (Jordan, 1963) when designing a system, i.e. humans and machines are fundamentally different and can therefore not be compared on a quantitative basis but complement each other, performing tasks in a joint cognitive system (Hollnagel and Woods, 2005). Cognitive Reliability and Error Analysis Method (CREAM) (Hollnagel, 1996, Hollnagel, 1998) CREAM is a Human Reliability Analysis (HRA) method, i.e. modeling cognitive errors and error mechanisms into the risk assessment processes. The basic notion is that of contextual control modelling, i.e., describing human cognition in terms of the competence for actions and the way in which the actions are controlled. CREAM can be used to identify the most likely cause of an observed event--either an accident or an erroneous action. The method can also be used in a predictive way to derive the likely consequences of specific erroneous actions. Task Evaluation and analysis Methodology (TEAM) (Johansson, 1994, Wäfler, et al., 1997) The method was developed between 1994 and 1996. The main aim is to evaluate existing advanced manufacturing systems (AMS) from a user perspective in order to pinpoint efficiency problem areas. Further to provide support for humans to better interact with complex technology (Wäfler, et al., 1997). Task analysis is presented in an evaluation matrix, developed by Stahre (Stahre, 1995), based on a combination between Sheridan’s supervisory control and Rasmussen’s human behaviour levels. Four factors are considered: work environment, work tasks, information flow and system performance (Johansson, 1994). The method should ideally be performed by multidimensional system design teams with at least one human factor specialist. Three levels are used for task evaluation: 1) generally difficult, 2) differentially difficult, 3) tasks known by few operators Taxonomy for Cognitive Work Analysis (Rasmussen, 1985, Rasmussen, et al., 1990) 19

This taxonomy was first published in 80s and should be used for effective support of decision processes to create a work practise that suits the individual users’ cognitive resources (Rasmussen, et al., 1990). A work domain should be represented at five levels of abstraction, representing goals and requirements, general functions, physical processes and activities, as well as material resources (Rasmussen, et al., 1990). Any of these levels has a work function (what should be used) which can be seen both as a goal (why it is relevant) for a function at a lower level, and as a means for a function at a higher level (how this is realized), (Rasmussen, et al., 1994). Moving from a lower level to a higher level of abstraction means a change in the representation of system properties. TUTKA production assessment tool (Koho, 2010) The TUTKA production assessment tool was developed during the end of 2000s. The main aim with the tool is to assess the current state of a production system and to identify potential and means for improvements. The tool is comparing the current state of the system with a desired state, i.e. a well performed production system, by using 33 key characteristics, 6 decision areas and 6 production objectives. Systematic Production Analysis (SPA) (Ståhl, 2007) The SPA was developed in 2007-2008 with focus on manufacturing processes such as machining. The main aim is to measure the existing production condition and to simulate (Jönsson, et al., 2008) different outcomes regarding three main parameters, i.e. Quality (Q), Down-time parameters (S) and Production speed/tact (P) in order to reduce cost. The methodology has also been used in assembly operations (Andersson, et al., 2009), focusing on capacity flexibility and part cost. Two levels of automation is used to describe the assembly stations (manual/ automatic). Productivity Potential Assessment (PPA) (Almström and Kinnander, 2007 , Almström and Kinnander, 2009 ) The PPA method was developed during 2005-2006 by the institute of innovation and management at Chalmers University of Technology, Sweden. The main aim is to show the improvement potential of productivity in Swedish manufacturing companies. The parameters forming the PPA-method are divided into different 4 levels:    

Level 1 is the core of the method, constituting two parameters for measuring efficiency in manual work and machine work respectively. Level 2 parameters affect productivity at corporate level, Level 3 parameters indicate the company’s ability to improve the production while maintaining a sound work environment. Level 4 treats the potential of improving productivity

Four levels of (mechanical) automation are used: 1) Man- Manual, 2) Semi – Semi-automatic 3) Auto – Automatic 4) Proc – Process industry Lean Customisation Rapid Assessment (LCRA) (Comstock, 2004) This method is a further development of the Rapid Plant Assessment (RPA) method, which was developed to help managers to fast determine if a factory was lean or not and discern the factory’s 20

strength and weaknesses where (Goodson, 2002). The main aim with the further develop method, LCRA, is to provide support in the analysis and/or design of a production system or even en entire company for mass customisation (Comstock and Bröte, 2003). This is done through three evaluation sheets divided into costumer elicitation, engineering and manufacturing.

3.4.2 DYNAMO The main aim with this method was to investigate how decisions of automation was made and to work out a possible framework for design, measuring and visulise automation to be used for strategy decisions in order to achieve dynamic automation in manufacturing (Säfsten, et al., 2007, Winroth, et al., 2005). The DYNAMO method was developed during 2004 to 2007 with help of six case studies and validated with a seventh (except the last step) (Granell, et al., 2007). The DYNAMO method contains of eight steps, resulting in minimum and maximum levels of the company’s automation strategy. This method could be seen mostly as a qualitative method because the tools used are interviews and observations. The taxonomy (se chapter 2) containing seven grades are used as a reference scale for the observations.

3.5 EFFECTS   The following sections will discuss different types of effects in an assembly system, both direct measurable and indirect measurable parameters. As described in the delimitations in chapter 1, cost is off course the primary KPI and the desired parameter to effect. Because it is primary it is will not be discussed nor brought up in this chapter, but seen as an effect of changing the secondary effects and LoA.

3.5.1 TIME Measuring different time parameters has always been important in industry. The basis for measurements and methods started with F.W Taylors scientific management (Taylor, 1911), motion studies by Frank and Lillian Gilbreth (Gilbreth, 1911) and later the MTM (Method-TimeMeasurement) by H.B Maynard in 1917 (Bicheno, 2006, Smith, 2004). While the Gilbreths' motion study work is commonly linked with Frederick Taylor's time studies and grouped within the various "laws and principles" of scientific management, in reality there is a great difference between the two. The components which was originally known as the "Taylor system" and later became scientific management changed how workers where paid, introduced a new division of labour, as well as expanded and strengthened the role of management. The use of stop watches to measure and set the proper time for tasks was important, but only as part of the overall system. The Gilbreths' motion studies where more focused on how a task was done, and how best to eliminate unnecessary, tiring steps in a process. The main difference between time studies and MTM is that in the former, the time to perform a task is measured while in MTM an average time is calculated based on different motions, which means that you do not have to measure on the shop floor to gain a first hint of how long the task takes. Today, measurement is mainly used for planning, but also to balance tact lines. This thesis focuses on three time parameters, which are described briefly below. o o

Cycle time -The time it takes to manufacture one individual product, (Mattson, 2004) and for the operator to finish all of his/hers work tasks (Rother and Shook, 2002). Operation time - Operation time is referred to the lead-time for carrying out one manufacturing step. It includes waiting time, transport and handling time to the production group, queue time in the production group, set-up time and production time. It represents one part of the throughput time. (Mattson, 2004) 21

o

Throughput time - The throughput time is the time it takes to manufacture an article from material and start of the first operation to delivery of a finished quality approved product. The throughput time is a part of the lead time and includes transport times, queuing time, set-up time and producing time (Mattson, 2004).

3.5.2 FLEXIBILITY An early definition of flexibility was provided by Stigler (1939) who defined it as: “Those attributes of a manufacturing technology which can accommodate grater output variations”. Since then, numerous definitions of flexibility have been suggested. Sethi and Sethi (1990) demonstrated the use of over 50 separate terms describing flexibility. Slack stated that flexibility is above all other measures of manufacturing performance, cited as a solution (Slack, 2005). More flexibility in manufacturing operations means more ability to adapt to customer needs, respond to competitive pressures, and to be closer to the market (Slack, 2005). The types of flexibility that will be discussed in this thesis are Volume flexibility, Routing flexibility and Production flexibility. Table 4 shows definitions of these three types of flexibilities from different perspectives TABLE 4 DEFINITIONS OF FLEXIBILITY

Authors

Volume Flexibility

Product Flexibility

The ability to handle a change in volume for a specific unit.

The ability to add and remove details from the mix over time

The ability to manufacture profitably in spite of a shifting manufacture volume.

The ability to manufacture a product in an economical way.

A company’s ability to adjust to variations in demand volumes.

A company’s ability be able to produce customised products within a given product concept and lead-time that fulfil customer demands for variation

(Slack, 2005)

The ability to change the level of aggregated output

The ability to introduce novel products, or to modify existing ones.

(Ståhl, 2006)

The ability to change production volume while retaining the effectiveness and the moving costs tied to the production tact.

The ability to develop, buy and produce new products and to modify the product and the assembly system for normal and nominal production.

(Gerwin, 1983) (Browne, et al., 1984)

(Mattson, 2004)

Routing Flexibility The ability to reroute a product’s path if a unit doesn’t work. The ability to continue manufactures a product in spite of a tool breakdown. An operation that could be used as an alternative manufacturing step in another production group if the usual operation and production group are unavailable or unusable due to under capacity or machine breakdown -

The ability to produce a multitude of products and handle changes in production planning.

3.5.3 PROACTIVITY According to (Frese and Fay, 2001), research focuses on reactive performance concepts, where people have to fit given tasks. Occurring needs and solutions become responses to existing problems, i.e. highly reactive actions. The introduction and ramp-up of a new product is often a discrete and unique event rather than part of the long-term development of the assembly system. It is questionable whether the reactive approach is sufficiently progressive and competitive. Instead, assembly systems need to be dynamic and evolvable to really constitute long-term assets for the manufacturing company (Onori, et al., 2006). Consequently, the preferred assembly system should have the ability to proactively meet emergent and long-term fluctuations. In dynamic environments, the activities of the operators’ job are no longer fixed and the work situations he/she faces are unlikely to be identified by work instruction 22

sheets. It is assumed that assembly work settings enabling proactive behaviour on the part of the human operators, which is important for managing the increasing uncertainty of work contexts (Griffin, et al., 2007). The approach is based on the concept of proactivity: the ability of operators to control a situation by taking action and effectuating changes in advance ensuring a favourable outcome (Dencker, et al., 2007). This is in line with (Griffin, et al., 2007), who defined proactivity as: “the extent to which the individual takes self-directed action to anticipate or initiate change in the work system or work roles” In work situations characterised by uncertainty, where aspects of work roles cannot be formalised, a proactive assembly system with a focus on active operator participation may be favourable and is supported by several studies [see e.g. (Crant, 2000, Dencker, et al., 2007, Frese and Fay, 2001, Parker, et al., 2006)] (Bruch, et al., 2008). Proactive behaviour by operators supports both the short and long term development of proactive assembly systems (Crant, 2000) stated that “Proactive behaviour can be a high-leverage concept rather than just another management fad, and can result in increased organizational effectiveness” Proactive behaviour can be characterised by 1) anticipation of problems related to change, 2) initiation of activities that lead to a solution of the change-related problems and improvements in the work, and 3) resolution of change-related problems (Sherehiy, et al., 2007). Proactive operators’ decisions are influenced by clues, early warnings, uncertain information, lack of information and the overall objectives. The latter is important, as proactive operators are expected to have a long-term view of and anticipatory perspective on the development of their work place (Bruch, et al., 2007). Such a system consists of technical components efficiently integrated with human operators to constitute reliable resources in the manufacturing system. In this way, present and future requirements for sustainability, flexibility and robustness can be met (Dencker, et al., conditionally accepted for publication). The effect of predicted and unpredicted disturbances can be minimized, thus enhancing the availability of the entire assembly system (Flegel, et al., 2005). Unfortunately, the use of proactivity as a competitive factor in assembly system design is not widespread (Dencker, et al., 2009)

3.5.4 COMPLEXITY Complexity can be defined as: “the complexity of a system is the degree of difficulty in predicting the system properties, given the properties of the system’s parts” (Weaver, 1948). Complexity can be divided into three sub parts, illustrated in Figure 6.

23

FIGURE 6 ILLUSTRATES A DIVISION OF COMPLEXITY INTO PERCEIVED, OBJECTIVE AND SUBJECTIVE COMPLEXITY

Perceived complexity is in research closely related to managing and handling critical events, production disturbances, frequent changes, unknown situations, unpredicted situations, and difficult work tasks etc. [6-8]. Hence, as production systems become more complex there is more that can go wrong, in several ways, and it is increasingly difficult to predict faults [9]. Human cognitive skills at different levels in the organization are increasingly crucial when manufacturing systems are becoming increasingly complex and subjected to changes and uncertainties [10]. Objective complexity can be further divided into:  

structural complexity Blecker et al. (Blecker, et al., 2005)or static complexity Frizelle (Frizelle and Suhov, 2001). Characteristics are related to fixed nature of products, hierarchical structures, processes, variety, and strength of interactions. Behavioral complexity (Asan, 2009) is characterized by dynamism, nonlinearity, deviation from equilibrium, history, adaptive, emergent structures, and self-organisation evolution. Dynamic complexity, which is caused by external and internal sources within the operation, like variations in dates and amounts due to material shortness, breakdowns, insufficient supplier reliability.

Regarding subjective production complexity, the same production system or situation may be perceived differently depending on a number of different factors such as individuals´ skills, competence and experience (Fässberg, et al., 2011).

24

4 RESULTS This chapter presents the results from each paper, how these are related to the research questions and the four theoretical areas. Table 5 shows a summary of contribution of each paper in relation to the three RQs and the theoretical areas, brought up in previous chapter. TABLE 5 HOW THE PAPERS CONTRIBUTE TO THE RQ:S AND THEORETICAL AREAS

Theoretical and practical contribution

x

Paper 1

x

x

Paper 4

x

x

RQ 2

x

x

Paper 2

x

x

x

x x

x

x

Paper 3

Assembly systems

x

Effects

Assessment methods

Levels of Automation RQ 1

Why is it important to quantify Levels of Automation (LoA) in an assembly system context? P1 present result from a literature study done to describe the lack of quantitative methods when assessing Levels of automation in assembly systems. P4 reviews ten assessment methods thru a literature study to determine if these models are assessing LoA and so, if this is done quantitative or qualitative. How should LoA be quantified, measured and analysed in assembly systems? P2 present a methodology, DYNAMO++, for measuring and analysing LoA based on five case studies and a literature review. P3 presents results from seven case studies using DYNAMO++ P5 is introducing the ProAct meta model which describe the relation between automation, information and competence. This is a further development of the methodology described in P2. P1 describe a concept model containing a main loop for measuring and analysing LoA and other important areas to consider when doing a task allocation between resources. A further development of the meta methodology presented in P4.

Paper 5

Paper 1

RQ 3

x

Paper 3

Comments and contribution

x

x

What are the expected effects of analysing and changing LoA?

x

x

P3 presents results from seven case studies to determine if to change LoA in order to achieve the triggers for change Considering LoA related to Time and Flexibility

Paper 1

x

x

Paper 5

x

x

x

P5 presents how LoA could affect Proactivity

x

x

P6 is describing thru an industrial case study, the relations between assembly errors, complexity and LoA and how to quantify these in an assembly context.

Paper 6

The sections below summarises the appended papers in relation to the RQs 25

4.1 RESULTS RELATED TO RQ1  The sections below will give a short description of the most important results in the papers related to RQ1: Why is it important to quantify Levels of Automation (LoA) in an assembly system context? The way companies choose and use their resources has been an issue ever since the craftsman became an operator, and the manual work became more automated. This research question will discuss why automation needs to be put in a quantitative context. Furthermore, why both the physical and cognitive automation needs to be addressed when measuring, analyzing and improving an assembly system.

4.1.1 PAPER 11 The aim of this paper was to describe the need for quantitative methods when wanting to measure and analyse different automation solutions in assembly systems. The part of the paper related to RQ1 is based on literature studies. According to Fasth et al (Fasth and Stahre, 2008) and Säfsten et al. (Bellgran and Säfsten, 2005, Säfsten and Aresu, 2000), a majority of companies studied, have a clear picture of why to change their system. However, the evaluations are often informal and unstructured, i.e. interpretation rather than facts. To choose solutions based solely on experience and interpretation rather than facts and numbers might not be the optimal solution when designing a system. A more reliable and objective quantitative method is therefore needed. A problem related to MABA-MABA-oriented methods is the simplicity e.g. “put your allocation problem into the method and the solution will emerge from the other end” (Dekker and Woods, 2002). The methods do not really explain the cognitive actions for how and when to intervene, nor do they describe how to switch from level to level. The relevance of a task allocation process is obvious, yet there is still lack of systematic methods and, more importantly, methods that can be applied to advanced technological systems (Older, et al., 1997). Another problem with new methods and tools in the human factors area concerns their lack of uptake and use by system developers. New methods must therefore be developed jointly with its users, i.e. adaptable to be put in practice (Older, et al., 1997, Waterson, et al., 2002), furthermore the method must be validated within its planned area of use. The paper compares requirements from (Older, et al., 1997, Waterson, et al., 2002) with the concept model developed by (Fasth and Stahre, 2010) in order to see if the concept model is fulfilling them. The concept model and the validation of the model are described under RQ2. The taxonomy is a seven-step reference scale, for cognitive and physical LoA aiming at quantifying tasks with help of different Levels of Automation. Frohm (Frohm, et al., 2008) defined physical tasks as the level of automation for mechanical activities, mechanical LoA, while the level of cognitive tasks is called information LoA. Mechanical LoA is WITH WHAT to assemble, while Cognitive LoA is HOW to assemble on the lower levels (1-3) and situation control on the higher level (4-7) (Fasth and Stahre, 2008).

1 Fasth, Å. and Stahre, J. (submitted 29 June, 2011), Task allocation in assembly systems –Measuring and analyzing Levels of Automation, special issue (Theoretical Issues in Ergonomics Science)

26

A matrix integrating the two reference scales, as seen in figure 7, forms a 7x7 matrix, resulting in 49 possible types of solutions for task allocation, each including a physical LoA and a cognitive LoA. The figure displays the division between human and machine assembling and monitoring the tasks. Machine assembling – A machine is performing the task and a human has a monitoring role or no role at all (in totally autonomous systems) Human assembling and monitoring – A human is performing the task and also monitoring her own work (or has no technique helping her monitoring the work) Machine/technique monitoring – A machine or technique is monitoring the task performed either by human or machine. When the machine/technique is both performing and monitoring the task, humans could still have a superior monitoring on station or factory level.

FIGURE 7 MATRIX ILLUSTRATING PHYSICAL AND COGNITIVE LEVELS OF AUTOMATION (LOA)

The matrix is used as a quantitative way of measuring the current LoA in the chosen areas´ tasks. The result is used for further analysis to meet triggers for change and also to make the company understand their mind set in a clearer and more objective way when it comes to automation.

27

4.1.2 PAPER 42 This paper reviews ten methods or models developed during the last twenty years used for redesign, measuring or analysing a production system (further information about the different methods and models is to be found in Frame of reference). Furthermore, a comparison is done between the methods and models based on four focus areas, seen in table 6. The comparison is done with the aim of putting the developed DYNAMO++ and concept model into perspective due to the other methods and models. A literature study is used in order to review the methods and the focus areas. TABLE 6 SUMMARY OF THE ASSESSMENT METHODS AND FOCUS AREAS

Assessment methods DYNAMO++ (Fasth, et al., 2010) and Concept model (Fasth and Stahre, 2010) TUTKA production assessment tool (Koho, 2010) Systematic Production Analysis (SPA) (Ståhl, 2007) Productivity Potential Assessment (PPA) (Almström and Kinnander, 2007 )

Focus areas 1. What assessment scale and level of change within the production system is the main focus? 2. Assessment objectives i.e. what is the methods’ main measurement parameters? 3. Assessment methods i.e. qualitative or quantitative methods? 4. Where within the dimensions of Socio-Technical and Physical Cognitive are the methodology’s main focus?

Lean Customisation Rapid Assessment (LCRA) (Comstock, 2004) A model for types and levels of human interaction with automation (Parasuraman, et al., 2000) Complementary Analysis and Design of Production Tasks in Socio-technical Systems (KOMPASS) (Grote, 2004, Wäfler, et al., 1997) Cognitive Reliability and Error Analysis Method (CREAM) (Hollnagel, 1996, Hollnagel, 1998) Task Evaluation and analysis Methodology (TEAM) (Johansson, 1994, Wäfler, et al., 1997) Taxonomy for Cognitive Work Analysis (Rasmussen, et al., 1990)

The selection of methods is always hard to do but in this case the author chose methods related to national and international well-known and well-cited developers. Focus area 1 has been chosen to be able to determine if the models or methods should or could be used in a design phase or in a running phase of the system, i.e. large or minor changes. Furthermore, if the methods had a strategic and organizational approach or a “shop-floor” approach. Results show that all of the methods had a shop floor approach. This resulted in an empty quadrant in the evaluation matrix (Socio-Physical), in focus area 4. Focus area 2 has been conducted in order to determine if the method is primarily investigating cost and/or productivity issues or if the methods are investigating other effects as well.

2

FASTH, Å. (Accepted for publication) REVIEWING METHODS FOR ANALYSING TASK ALLOCATION IN A PRODUCTION SYSTEM International journal of logistic management.

28

To illustrate the results from focus area 3 and 4 a matrix were developed with two axes. The first axis’ dimensions are Socio and Technical and the dimensions of the other axis are Physical and Cognitive. This forms four areas in which the methods and models could be placed. Focus area 3 are aiming to investigate if the more quantitative methods is placed in the technical-part quadrant and the more qualitative methods are place in the socio-part of the system. Focus area 2 and 3 are closely connected to each other. If the method is focused on cost and productivity it is defined as quantitative and is grouped in quadrant Technical-physical. If the methods are more focused on cognitive behaviour and sociological aspects they are usually qualitative and grouped in the Socio-cognitive quadrant. The Socio-Physical quadrant has a more organisational approach and the Technical-Cognitive quadrant is treating artificial intelligence and autonomous systems. None of the methods were place in these quadrants. Focus area 4 is a result of the first three focus areas and resulted in a matrix illustrated in Figure 8 where the ten assessment methods are placed due to results from the earlier focus areas and the definitions of the quadrants. The aim is to determine how the methods are handling the issue of assess task allocation and Levels of automation in their models.

FIGURE 8 RESULT OF THE METHOD REVIEW (FASTH, ACCEPTED FOR PUBLICATION)

Results show that the methods conducted before the year of 2000 had a clear Socio-cognitive and qualitative approach while the method conducted after the year of 2000 had a more Technical-physical and quantitative approach towards the issue of automation and resource and task allocation in the system. Only DYNAMO++ had the main focus on measuring and assesses Levels of automation. The result shows that the DYNAMO++ and the Concept model is in-between the socio-cognitive models and the technical-physical models when measuring and analysing a production system. The model takes into consideration both physical and cognitive Levels of Automation in a more delicate scale than the other methods and models which makes the task allocation measurements and analysis more precise. Furthermore, the model also considers the social aspects in terms of competence within the operator group and the information flow to and from the cell or station.

29

4.2 RESULTS RELATED TO RQ 2  The sections below will give a short description of the most important results in the papers related to RQ2: How should LoA be measured and analysed in assembly systems? This research question will explain the evolution of finding an easy-to-use methodology for industry and a more logic concept model to enable data analysis, developed by the author for measuring and analysing LoA in assembly systems.

4.2.1

PAPER 23

The aim of the second paper (Fasth, et al., 2008 ) was to present the further development of the DYNAMO methodology (Frohm, 2008) with focus on the analysis phase and its constituent steps. The methodologies used to write this paper were observations and semi-structured interviews and measures in the industrial case studies and literature study. A total of thirteen cases were analysed:  

Eight case studies [Two of the case studies within the ProAct project and six of the case from the DYNAMO project (Granell, et al., 2007)] were performed with the DYNAMO method. Five case studies were conducted to validate the developed methodology, DYNAMO++.

The DYNAMO method consists of eight steps. The developed method consists of twelve steps, divided into four phases. The skeleton of the DYNAMO method, i.e. the first six steps (with minor changes), is used as a base for the further development. The first two phases, i.e. pre-study and measurement, investigate the current state of the system e.g. product variants, number of components, number of operations and tasks within the chosen area, number of shifts and number of operators and levels of automation (LoA). Methods used are Value Stream Mapping (VSM), Hierarchic Task Analysis (HTA), structured interviews and the LoA taxonomy. The major contribution of the further development of the method is within the analysis and the development of the implementation phase. One of the main developments in the analysis phase is the attempt to transform the reference scale, developed by Frohm (2008) into a more logical description of a matrix. Described in equations 1.1-1.3: 1 ≤ LoAtotal ≤ 49

(Eq. 1.1)

LoAtotal → (LoAmech) ∧ (LoAinfo)

(Eq. 1.2)

WHERE LoAmech (y) = 1≤ y ≤ 7 and LoAinfo (x) =1≤ x ≤ 7

(Eq. 1.3)

Eq 1.1 describes the number of possible solutions within the matrix (LoAtotal) Eq 1.2 describes the relation between LoAmech (later LoAphysical) and LoAinfo (later LoAcognitive) Eq. 1.3 defines the discrete steps at each axis

The equations were formulated to get a logical ground and to be able to add dimensions or parameters to the methodology. This matrix is used to visualise the different levels of automation. It is also used in the analysis phase to show the results of the measurements and the suggestions of possible improvements. 3

Fasth, Å., Stahre, J. and Dencker, K. (2008) Measuring and analyzing Levels of Automation in an assembly system. Proceedings of the 41st CIRP International Conference on Manufacturing Systems (ICMS), Tokyo, Japan.

30

Furthermore a square within the LoA matrix was developed in order to simplify and visualise the result of the analysis step. The Square of Possible Improvements (SoPI) illustrates the span within the matrix by which the companies believe their systems could be improved, in terms of different parameters, resources and demands. The logic behind SoPI is illustrated in equations 2.1-2.6: SoPI → (LoAmech (min; max)) ∧ (LoAinfo (min; max))

(Eq. 2.1)

SoPI = LoAmech (min; max) * LoAinfo (min; max)

(Eq. 2.2)

WHERE LoAmech (y) = 1 ≤ min ≤ max ≤ 7 ∧ LoAinfo (x) = 1 ≤ min < max ≤ 7

(Eq. 2.3)

SoPItask ≤ LoAtotal

(Eq. 2.4)

SoPItask ⊆ LoAtotal

(Eq. 2.5)

Eq 2.1 defines the relation between the SoPI and the axis Eq 2.2 defines the square itself Eq. 2.3 defines the discrete steps within the square Eq 2.4 defines that the square itself should have less or equal number of solutions than the matrix Eq. 2.5 defines that the solutions of the SoPI should be e part of the solutions within the matrix.

In order to prevent sub-optimisation from task to operation optimisation, one condition regarding this was developed. In order to perform an operation optimisation, all the SoPItask has to be represented in the SoPIoperation in order to make an optimisation (Eq. 2.6); if not, one solution is to do an optimisation with some of the tasks and do a task optimisation on the others. It could be described as:

IFF SoPIoperation ⊆

, THEN

operation optimisation is possible

(Eq. 2.6)

The Square of Possible Improvements (SoPI) indicates the span within the matrix where company personnel believe their systems could be improved. The improvement potential is seen from different perspectives described by parameters, resources and demands. It is important to state that the information from the current state analysis is used as input for the future state solutions. The development of extended method logic and the addition of the time dimension to the existing LoA reference scales will make it easier to simulate different assembly system solutions. Moreover, it will provide a measurable value that can be used for comparing the present and future assembly system. This would provide companies with a deep foundation for decision making in the planning and implementation phases of their future assembly system. Focus on time parameters and follow-up would facilitate measurement of the change from the old to the new current stage. Moreover, it would provide a measurable value with which comparisons between the present and future assembly systems is made possible.

31

4.2.2 PAPER 34 The aim of this paper was to investigate if there were relations between the parameters time, flexibility and LoA. Three questions were discussed: 

Time and/or Flexibility has been important in the latest paradigm shifts; is it still important today?



Does Levels of Automation need to be changed to achieve time savings and/or increase flexibility in today's assembly systems?



Could Lean tools like JIT simplify the understanding about Levels of Automation, flexibility and time savings?

In order to answer these questions six case studies were conducted using the DYNAMO++ methodology illustrated in Figure 9.

FIGURE 9 AN OVERVIEW OF THE 4 PHASES IN THE DYNAMO ++ METHODOLOGY (FASTH AND STAHRE, 2008)

In the current state, data were gathered about flow, type of assembling and number of products in the measured area. Flow and time parameters were also documented. A measurement of the current state’s Level of Automation was carried out; the value is based on the automation level that the operator used to perform the task. The result of the current state phase is illustrated in Table 7. TABLE 7 CURRENT STATE IN SIX CASE STUDIES

4

Fasth, Å., and Stahre, J., Does Levels of Automation need to be changed in an assembly system? - A case study, Proceedings of the 2nd Swedish Production Symposium (SPS), Stockholm, Sweden, 2008.

32

Results from interviews regarding the companies’ trigger for change (illustrated in Table 8) show that Flexibility, e.g. Volume and product, and Time, e.g. through-put time, still are important factors for the companies to consider when redesigning their systems. TABLE 8 TRIGGERS FOR CHANGE AND LEAN AWARENESS

As a part of the analysis work, LoAs were measured and a SoPI was illustrated based on the logic from paper 2. Figure 10 illustrates an example from case study A, where the left matrix is an example of a task allocation for task 1.1 and the right matrix is an example of a possible operation optimisation between task 1.1-1.5. The matrix shows how the possible solutions decreases when going from a single task optimisation (18 possible solution) to a operation optimisation (6 possible solutions)

FIGURE 10 TASK AND POSSIBLE OPERATION OPTIMISATION

Results show that either the mechanical or information LoA needed to be changed in almost all companies in order to reach their triggers for change; LoAinformation (50 % of the companies) in terms of digital assembly instructions in different levels due to the operators’ competence and experience, or visualisation of the production in terms of state lamps. LoAmechanical (33 % of the companies) in terms of conveyers (transport automation), and variable automation in terms of redundancy or plug-and-play flows. The degree of Lean awareness also seemed to have an impact on how these three parameters were treated. The companies with middle or high lean awareness found it easier to understand the DYNAMO++ methodology and the term Levels of Automation. Further, it was easier for them to understand time savings in terms of non-valuable time.

33

4.2.3 PAPER 55 The aim of this paper was to describe a Meta methodology (the ProAct loop), for designing a proactive system in a structured way considering Automation (LoA), Competence (LoC) and Information (LoI), aiming to increase flexibility, achieve time minimization and increase the operators’ action space. The effect of investigating the three areas is discussed in RQ3. It is important to consider both the qualitative and quantitative aspects when redesigning an assembly system. The socio-technical school [42, 43] could be seen as an alternative in order to expand the operator action space and to find the interaction between the three areas, i.e. automation, competence and information. The “social system “could be related to the operators’ roles, Level of Competence (LoC), Level of Information (LoI) and the cognitive Level of Automation (LoA). The technical system could be connected to mechanical LoA and in some cases LoI in terms of technical solutions (information carriers). The ProAct-loop, shown in figure 11 (illustration to the left), is a Meta methodology connecting the three areas by combining theory and methods. The loop is used in two iterations, the current stage mapping and the future stage analysis. The methodology was tested and validated in five Swedish production companies in 2007 and 2008.

FIGURE 11 METHODOLOGIES USED IN THE PROACT LOOP AND THE DIFFERENT LEVELS OF MEASUREMENT

In order to make a first statement about the system and its proactive potential, an attempt to quantify the three areas were done, illustrated in Figure 11 (illustration to the right). Through a deep understanding of the interaction and relation between the three areas, companies can avoid over- or under-automated systems. Also, excessive information and incorrect competence levels can be avoided. Further, companies will be able to balance the three areas and do changes where it is most needed. For example, low competence level can to some extent be compensated by higher levels in the information and/or automation area and vice versa. The operators' situation awareness will improve, thus increasing their ability to act proactively.

5

FASTH, Å., BRUCH, J., DENCKER, K., STAHRE, J., MÅRTENSSON, L. & LUNDHOLM, T. (2010) Designing proactive assembly systems (ProAct) - Criteria and interaction between automation, information, and competence Asian International Journal of Science and Technology in production and manufacturing engineering (AIJSTPME), 2 (4), 1-13

34

4.2.4 PAPER 16 The aim of this paper was to describe a concept model containing a main loop for measuring and analysing LoA, illustrated in Figure 12. Furthermore, the concept model aims at mapping other areas that are important to consider when doing a task allocation between resources. The concept model is a further development and a leaner version of the ProAct methodology presented in P5. The concept model is aiming at determining appropriate task allocation with a span of various levels of automation in assembly operations. Methods used in this paper are empirical collections, i.e. interviews and measures and a theoretical framework. The concept model was first presented in a proceeding at the 3rd CIRP Conference on Assembly Technologies and Systems7.

FIGURE 12 CONCEPT MODEL, FURTHER DEVELOPED FROM DYNAMO++ (FASTH AND STAHRE, 2010)

The model contains two parts. The first part is the main loop which has its base in the 12-step methodology of DYNAMO++, described in paper 2. The second part is describing the other parameters that are important to consider when re-designing an assembly system. Both parts are related to the ProAct loop presented in paper 4. The results from this paper are an attempt to describe the underlying methodologies in a more visual and leaner way. Furthermore, the concept model is an attempt towards a more logical explanation in order to develop a future ontology. 

  6

FASTH, Å. & STAHRE, J. (2nd Review) Task allocation in assembly systems -Measuring and analysing Levels of Automation, Special issue (Theoretical Issues in Ergonomic Science) 7 FASTH, Å. & STAHRE, J. (2010) Concept model towards optimising Levels of Automation (LoA) in assembly systems. Proceedings of the 3rd CIRP Conference on Assembly Technologies and Systems, Trondheim, Norway

35

4.3 RESULTS RELATED TO RQ 3  The sections below will give a short description of the most important results in the papers related to: RQ3: What are the expected effects when analysing and changing LoA? Cost is a common parameter that is often used when improving production systems. This research question will discuss whether there are other parameters that are linked to LoA and could get as good or better results than just focusing on cost when deciding if the level of automation should be changed.

4.3.1 PAPER 18 Results from Paper 1 show that it is possible to compare two different levels of automation and to use them as alternatives in order to create flexibility in the system (volume and route). The results are illustrated through a case study. Part of the results has also been published (Fasth, et al., 2008 )and (Fasth and Stahre, 2008). The first two phases in the DYNAMO++ were used for a current state analysis. The company’s trigger for change was to increase volume flexibility for a specific product family. In order to analyse if this was achievable an illustration of the product flow and an investigation of the bottleneck was done. The company had integrated redundancy in the bottleneck station, as illustrated in Figure 13. The matrix shows two different alternatives that were used for assembling. Alternative 1, the main path, is a robot cell, LoA= (6; 5). The second alternative is a station with an operator and a fixture, LoA= (5; 3). This solution was used as an alternative when the robot cell was unusable. In order to fulfil the enhancement of volume flexibility, a solution is to use both stations and to perform task allocation, i.e. dynamically changeable LoA depending on the order status on a daily basis.

FIGURE 13 OBSERVED LOA AND CYCLE TIME (C/T) VALUE FOR STATION 3 (S3)

8

FASTH, Å. & STAHRE, J. (2nd Review) Task allocation in assembly systems -Measuring and analysing Levels of Automation, Special issue (Theoretical Issues in Ergonomic Science)

36

The result of this task allocation gave four alternatives to elaborate regarding volume and route flexibility. The robot cell is used as the main resource in the system. The productivity for one normal day is 24 products per hour, i.e. 192 products per shift. Assume that a breakdown for two hours happens on the robot cell. Without the routing flexibility the loss will be 24 parts per hour = 48 products. With the routing flexibility the company is able to use the static work station under the reparation producing 18 products per hour, i.e. a loss of 5 products per hour = 10 products. The four alternatives give a variance of 144 products (min 152, max 296). The amount of products produced with different alternatives is shown in Figure 14.

FIGURE 14 CREATING VOLUME FLEXIBILITY THROUGH TASK ALLOCATION

This case shows how task allocation and changing recourses could be illustrated in an easy way for the company so that it can take well-founded decisions due to the volume it wants to produce.

37

4.3.2 PAPER 59 The results in this paper related to RQ3 are to investigate the criteria and interaction between three areas: automation, information, and competence in order to design a proactive assembly system. The methods used to achieve this were the DYNAMO++ methodology, definition of operators’ roles and work tasks(Sheridan, 1995, Stahre, 1995) and the Skill-Role-Knowledge(SRK)-model (Mårtensson and Stahre, 2003, Rasmussen, 1983) . Through a deep understanding of this interaction, an optimising of the separate areas in a structured way using the proposed Meta-method (the ProAct-loop, described under RQ2) could be done. As a result, companies will be able to balance the three areas and optimise where it is most needed. One example is the short-time planning and the ability to adapt to for example increased need for more products, or the need for another machine when the main one is done, i.e. volume and route flexibility shown in the previous example from Paper 1. Further, a low competence level can to some extent be compensated by higher levels of information and/or through higher cognitive LoA and vice versa. The square within the LoA matrix (Fig. 15) represents the tasks or operation’s action space in which automation solutions for the future assembly system may vary. Depending on how movement in this square is done, the effort to change to other technical solutions, is reflected on both the level of information and the level of competence.

FIGURE 15 EFFORT NEEDED TO CHANGE LOA WITHIN A SYSTEM

In proactive assembly systems, operators are encouraged to take more responsibility, far beyond managing operation and disturbances. Requirements on frequent reconfiguration, either initiated by explicit demands, or by changes due to proactively foreseeable problems or requirements, are normal. This means that in the future there is a need to take operator abilities and limitations into consideration, as well as the operators' various ways of using information and making decisions in different working situations (Dencker, et al., 2007).

9

Fasth, Å., Bruch, J., Dencker, K., Stahre, J., Mårtensson, L. and Lundholm, T. (2010) Designing proactive assembly systems (ProAct) - Criteria and interaction between automation, information and competence, Asian International Journal of Science and Technology in production and manufacturing engineering (AIJSTPME), vol 2 issue 4, pp.1-13

38

4.3.3 PAPER 610 The aim of this paper is to investigate if cognitive automation can be used to increase quality in a complex final assembly context. An industrial case study has been executed to test if there is a relation between cognitive automation, quality and quantitative (objective) station complexity, illustrated in Figure 16.

FIGURE 16 THE INVESTIGATED RELATIONS

The increased task complexity in assembly needs to be handled; otherwise the quality of the product and productivity in the system could be affected. In order to maintain high quality and reduce the complexity, one solution could be to consider cognitive automation for the operator, e.g. technical support to know how and what to assemble and to be in situation control. An industrial case study has been executed in order to investigate the effects cognitive automation have on quality, in terms of assembly errors, in a complex final assembly context. In order to test the aim of the paper, an area of interest was selected. The area is one of the most complicated in the final assembly with a very high product variety and a large number of parts. The chosen area consists of a total number of sixteen stations where seven have been studied within this project (the grey operators in Figure 17 represent the chosen stations). The chosen stations are a part of the pre-assembly area for the preparation line of engines. In the line the engines are customised with correct driveshaft, cables etc. The engines assembled are used in all models and variants on the main assembly line. There are two areas for the pre-assembly of the engines and this is the second area, Power Pack 2 (PP2).

FIGURE 17 THE SELECTED AREA WITH TOTAL NUMBER OF STATIONS AND SELECTED STATIONS

Measuring Levels of Automation (LoA) was made from direct observations and from standardised assembly instructions. An advantage of the use of two sources of information is that the standardised assembly instruction does not always correspond with the reality which we want to capture.

10

Fässberg,T., Fasth, Å., Hellman, F., Davidsson, A., and Stahre, J (accepted for publication), Interactions between complexity, quality and cognitive automation, Proceedings of 4th CIRP Conference On Assembly Technology Systems

39

Two models were assessed for each station, the most common model regarding demand and the heaviest model to produce regarding time. The distributions of the tasks for the two models are presented in the matrix illustrated in Figure 18.

FIGURE 18 LOA MEASUREMENT FOR ALL THE SEVEN STATIONS

By looking at the results, it is clear that the majority of the cognitive support is either none, LoAc=1, or very high, LoAc=5. Results show that 62 percent (H) and 64 percent (C) were made with LoA level= (1; 1), i.e. by hand and with own experience. The fact that so many tasks are done without cognitive support could have an impact on quality. Further, 25 percent (H) and 24 percent (C) is done with LoAcog =5 (often pick-by-light or indicators of what bit to use for the pneumatic screwdrivers). This paper shows that it is possible to use quantitative measures in order to show relations between station complexity, quality and cognitive automation. These methods could be further used in order to improve both the resource efficiency and resource allocation in order to get an effective assembly system. Then, the operators’ competence and experience should also be taken into consideration, which is not fully covered by using the three methods. The main conclusion is that there is evidence that cognitive support is needed in final assembly to minimize the negative effects of complexity.

40

5 DISCUSSION This chapter will discuss the most important results in the appended papers connected to the research questions; this will be done in relation to four areas of theoretical and practical contribution. Further, a discussion regarding choice of methods will be done. Lastly, a discussion of future work will be addressed. According to (Wiendahl, et al., 2007), on assembly level, an enabler to achieve flexible and changeability systems is the ability to upgrade or downgrade the degree of automation. For assembly operations, in contrast to machining operations, there is often the possibility to perform them either manually or automatically. Earlier research (Dekker and Woods, 2002, Parasuraman and Riley, 1997, Sheridan, 1995) show that in order to make well founded decisions, automation has to be seen as different levels of automation. Furthermore the automation decisions must have a well-founded ground and be a part of the company’s strategy plan (Säfsten, et al., 2007). Inagaki (Inagaki, 2003) argues that in order to analyse the choice of automation with regard to cost and benefit, quantitative models are needed. Further, he lists the main arguments regarding the choice of automation and the need for quantitative methods in order to make a well-structured decision: 1. Without a quantitative model, it is not possible to evaluate tradeoffs between cost and benefit. 2. Intuition is not powerful enough to analyze time-dependent characteristics of cost and benefit in a dynamically changeable situation. 3. Robustness or sensitivity of solution (whether to automate or not) can be analysed only when quantitative models are available. 4. Suppose a current plan is to ‘‘automate’’ a specific function. With a quantitative model, we can investigate to what extent the plan is superior to an alternative that leaves the function under manual control. By quantifying means it is possible to compare some kind of data with each other in a statistical way (Miles and Huberman, 1994). In line with Inagaki, Older et al. (Older, et al., 1997) stated the need for a more easy-to-use quantitative method for analysing Levels of Automation with their seventeen requirements. Results from the appended papers show that automation can be used as a primary parameter in order to analyse assembly systems in relation to competence, information, quality, time and flexibility. Furthermore, the results show the importance of having more than two levels of automation and to also in count the cognitive automation when searching for possible solutions in order to get a more levelled and precise automation solution. Therefore the aim of this thesis, quantifying, measuring and analysing the physical and cognitive Levels of Automation to enable competitive assembly systems, is of relevance and will be further discussed in this chapter. Therefore the aim of this thesis; quantifying, measuring and analysing the physical and cognitive Levels of Automation to enable competitive assembly systems is of relevance and will be further discussed in this chapter.

 

41

5.1 SUMMARY OF THE APPENDED PAPERS RELATED TO THE RQS  Below follows a short summary of the appended paper with regard to each RQ.

5.1.1 WHY IS IT IMPORTANT TO QUANTIFY LEVELS OF AUTOMATION (LOA) IN AN ASSEMBLY SYSTEM CONTEXT? Summary of the results from appended paper 1 and 3 related to RQ1: Results from the appended papers (1 and 3) show that it is important to consider LoA as a quantitative measure in order to be able to compare with other parameters important for the company (connected to RQ3): “talk to farmers in farmers’ way”. Further to be able to go from a more socio-soft way of thinking of primarily cognitive automation, towards a more technical explanation. The conducted case studies show that the cognitive automation is often forgotten or not prioritised. In order to increase competitiveness and to survive in a more complex environment, cognitive automation needs to be considered and developed further. Moreover, the case studies showed that the ways the automations were chosen were based on experience and “common sense” which is not always optimal; this shows a need for a structured way of measuring and analysing LoA in order to chose solutions in a more structured way (RQ2).

5.1.2 HOW SHOULD LOA BE MEASURED AND ANALYSED IN ASSEMBLY SYSTEMS? Summary of the appended papers 1,2 and 4 related to RQ2: Results from papers 1, 2 and 4 show that it is important to create a method that involves both the socio-cognitive (Qualitative) and the technical-physical (Quantitative) part when redesigning an assembly system: current state analyses to collect data about LoA, maturity in LEAN and system parameters (number of products, variants, type of layout etc.), measurement of LoA (both cognitive and physical), analysis due to triggers for change (relate to RQ3), implementation and follow-up.

5.1.3 WHAT ARE THE EXPECTED EFFECTS OF ANALYSING AND CHANGING LOA? Summary of the results, i.e. RA3: According to papers 1, 4, 5 it is important to investigate the triggers for change within a company before redesigning the system. The next step will be to determine how to achieve this change. This has to be done in a structured way as discussed in RQ2. The most expected and obvious change might not be the most optimal. According to several case studies (Fasth, et al., 2011, Fasth and Stahre, 2008) the most common triggers beside decreasing cost are: increase (volume and route) flexibility, increase proactivity, decrease or manage complexity and decrease cycle time. The prime focus with this RQ is to determine if a change of mindset, from looking primarily at cost to also consider pros and cons with different LoAs and other parameters related to both LoA and cost, could be a part of being more competitive. 

 

42

5.2 CONTRIBUTION MATRIX  The further discussion will be based on the contribution matrix, brought up in Chapter 2, p.11. The quadrants have been filled in with the contributions from the appended papers (RA) (illustrated in Figure 19) and will be discussed based on the four areas in the theoretical framework in chapter 3 and the results from the appended papers listed in Chapter 4.

Theoretical

Practical

RA1: Development of LoA matrix, illustration of a quantitative tool. Definition of LoA

RA2: Validation of DYNAMO++ and the concept model

RA2: Development of DYNAMO++ and the concept model

RA3: Use of DYNAMO++ and the concept model. Shown effects of measuring and analyzing LoA

Scientific

Industrial

FIGURE 19 RELATIONS BETWEEN THE THEORETICAL AND PRACTICAL CONTRIBUTION FROM A SCIENTIFIC AND INDUSTRIAL PERSPECTIVE

5.2.1

PRACTICAL CONTRIBUTION FROM A SCIENTIFIC PERSPECTIVE

This section will discuss the practical contribution with the results from the appended papers related to RQ2 and in relation to the theoretical framework of assessment methods (RA2: Development of DYNAMO++ and the concept model)

Without a quantitative model, it is not possible to evaluate tradeoffs between cost and benefit of different levels of automation (Inagaki, 2003). According to (Säfsten, et al., 2007) it has proved to be more successful to formulate an automation decision in congruence within the company rather than having automation as an only concern to reduce cost, i.e. the automation decision is pushed on the organisation (Boyer et al., 1996; Winroth et al., 2007). Results from papers 1, 4 and 5 show that considering LoA based solely on cost savings might not be the optimum solution. The quantitative approach and illustration of the LoA matrix could be used as a tool in order to analyse the current stage and to achieve the triggers for change and design a new current stage – even though results from paper 4 and 5 show that it is important to consider the more qualitative aspects such as Level of Competence and Level of Information when doing an in-depth study after the first analysis. According to Older et al. requirements on a quantitative method are: “new methods must be developed jointly with its users, i.e. adaptable to be put in practice (Older, et al., 1997, Waterson, et al., 2002) and the method must be validated within its planned area of use”. Existing models about the content and process of manufacturing strategy,(Skinner, 1969) deal with automation very briefly as a question that is included in the process technology decision (Säfsten, et al., 2007). Results from paper 3 show that the models or methods on “shop-floor” level are either focusing on the socio-cognitive of the system, i.e. competence, skill, operator group control/planning etc. or the more physical-technical part of the system, i.e. pre-defined criteria for a “perfect system”, measurements and cost savings. In line with Olsen (Olsen, 2004), the developed methods and model presented in papers 1, 2 and 4 are an attempt to combine qualitative (socio-cognitive) and quantitative (physical-technical) approaches rather than see contradictions between the two modes of analysis, and in line with Older et al, develop an easy-to-use method for the planned area of use, i.e. final assembly that could be used from shop-

43

floor level up to a strategic level. The evolution of how to measure and analyse LoA presented in paper 1, 2, 3 and 5 are illustrated in Figure 20.

FIGURE 20 EVOLUTION FROM DYNAMO TO THE CONCEPT MODEL

Results from paper 2 and paper 3 describe the DYNAMO++ methodology which was developed in association with five case companies and further validated in six cases. The focus was to develop the LoA matrix and the analysis steps with regard to the logic within the matrix. According to Secondly, the ProAct loop was presented in paper 4. The loop or Meta model consisted of three methodologies, where DYNAMO++ was one of them. In the ProAct loop, the DYNAMO++ was more focused on quantifying LoA and related the measured and analysed industrial result to the other two areas (competence analysis and information mapping). The third and final evolution is the concept model, presented in paper 1, which is a lean version of the ProAct loop aiming towards a more logical mathematical model. Is it possible to only use either one when it comes to quantitative and qualitative methods? Two of the most obvious alternatives when developing a methodology or method are either to have a quantitative or qualitative approach. The two approaches have different pros and cons depending on in what context the method is going to be used. Of course there are extremists in this matter: "There's no such thing as qualitative data. Everything is either 1 or 0" (Kerlinger, 1960) and "All research ultimately has a qualitative grounding" (Campbell, December,1976). According to (Miles and Huberman, 1994) these two research approaches need each other more often than not. Results from paper 3 show a clear line between methods developed before and after the year 2000. The methods developed between 1990 and 2000 had a clear qualitative approach while the methods conducted between 2000 and 2010 had a clear quantitative approach. The aim of the concept model is to try to combine these two approaches into one model. Using quantitative methods is a way to fast-scan and get results on where the problem area is. Furthermore it is an easy way to compare different stations or cells with each other, statically. The qualitative methods could then be used for in-depth study of the issues. The thought with the concept model is to give the companies an easy overview of aspects or parameters to consider when redesigning a system. The main loop is aiming for a more quantitative approach. Results from paper 5 show that the quantitative approach within the concept model could be used as a first step to find relations between LoA, Complexity and quality. Then, the operators’ competence (LoC, in Figure 24) and experience should also be taken into consideration, which is not fully covered by using only the quantitative methods. The areas are a mapping over areas to consider when automating an assembly system. These areas are a mix between qualitative and quantitative measures. By combining both quantitative and qualitative measurement approaches and combining different methods, the concept model could be used both as a fast scanning method to catch the problem areas and handle the triggers for change but also as an in-depth method.

44

5.2.2

THEORETICAL CONTRIBUTION FROM A SCIENTIFIC PERSPECTIVE

This section will discuss the theoretical contribution with the results in RQ1 in relation to the theoretical framework regarding Levels of Automation (LoA). (RA1: Development of LoA matrix, illustration of a quantitative tool. Definition of LoA)

In order to be able to measure and analyse LoA, a quantifiable way of illustrating LoA has to be developed. To determine what to automate, a classical task allocation strategy from 1951 (the MABAMABA list) was proposed by Fitts (Fitts, 1951). It was an attempt to suggest allocation of tasks between humans and machines by treating them as system resources, each with different capabilities. Two examples, i.e. “Machines Are Better At” performing repetitive and routine tasks while “Men Are Better At” improvising and using flexible procedures. At the time, comparing man and machine, was a revolutionary thought that caused a lot of debate. The questions are: is the list still is applicable in an assembly context and is the list quantitative enough? A problem related to MABA-MABA-oriented methods is the simplicity, e.g. “put your allocation problem into the method and the solution will emerge from the other end” (Dekker and Woods, 2002). The methods do not readily explain the cognitive actions for how and when to intervene, nor do they describe how to switch from level to level. Many researchers after Fitts have argued whether it is impropriate or not to automate or to compare man and machines, and how to allocate the different tasks to man or machine: Jordan (Jordan, 1963) argued whether you could actually compare man and machine, and that the two should be seen as complementary rather than conflicting resources when designing a man-machine system. Sheridan (Sheridan, 1995) proposed to “allocate to the human the tasks best suited to humans and allocate to the automation the task best suited to it”. Hancock (Hancock and Chignell, 1992) argues that it is only when both human and machine can do the same task that the question of task allocation becomes an issue. These arguments are still not quantitative enough, in order to measure and analyse LoA. An attempt after Fitts to assess LoA was Prince in 1985 (Prince, 1985) who designed a decision matrix, partly in line with Fitts in that some tasks were better performed by machines and some better by humans. But interestingly Prince also defined a set of tasks where the same task could and should be performed both by humans and by machines. Further, when there is no single allocation, the different resources need support from each other, which is in line with Jordan’s argument. Until today a lot of different “lists” and methods have been published. The relevance of a task allocation process is obvious, yet there is still lack of systematic methods and, more importantly, methods that can be applied to advanced technological systems (Older, et al., 1997). Older et al. (Older, et al., 1997) in 1997 (and 2002) compares eighteen different methods (developed 1965 and 1992).Seven of them contains quantitative evaluations but none of the methods are considering both cognitive and physical automation regarding task allocation. Results from publication 3 (Fasth, 2011) show that five of ten methods compared (developed between 1990 and 2010) have a quantitative approach, but only the DYNAMO++ method considers both cognitive and physical automation on a task level. Automation design is not an exact science, however; neither does it belong in the realm of the creative arts, with successful design dependent upon the vision and brilliance of individual creative designers (Parasuraman, et al., 2000). Without a quantitative model, it is not possible to evaluate tradeoffs between cost and benefit of different levels of automation (Inagaki, 2003). Attempts to describe different levels of automation between extremes, from totally manual to totally automatic have been done. 45

According to Parasuraman and Riley, automation can be characterized by a continuum of levels rather than as an all-or-none concept (Parasuraman and Riley, 1997). In line with Parasuraman and Riley, Frohm (Frohm, 2008) defines Levels of Automation as: “The allocation of physical and cognitive tasks between humans and technology, described as a continuum ranging from totally manual to totally automatic” The question today is not either continuum ranging or rather than. Results from paper 1 and 3 show that both industry and assessment methods developed by academia still sometimes see automation (partially the physical) as a binary decision. The author believes the question that should be asked is how to quantify the levels so that a more logic approach could be made when measuring and analyzing LoA, i.e. discrete steps from totally manual to totally automatic. To be able to make it simple and in order to design a LoA matrix, Levels of Automation is defined as: The allocation of physical and cognitive tasks between resources (humans or technology), described as discrete steps from 1 (totally manual) to 7 (totally automatic), forming a 7x7 LoA matrix containing 49 possible types of solutions. [Fasth, 2012] Results from paper 1 and paper 2 show a combination of the taxonomy and the matrix with discrete steps. The physical level 1-4 is defined as the task is performed by humans, i.e. the human is assembling (or performing) the task. Level 5-7 is defined as the machine assembling (or performing) the task. At the cognitive level, level 1-3 is the human who is monitoring the task while level 4-7 is the technique or machine that is monitoring the task.

46

5.2.3

THEORETICAL CONTRIBUTION FROM AN INDUSTRIAL PERSPECTIVE

This section will discuss the theoretical contribution with the results from the appended papers related to RQ3 in relation to the theoretical framework regarding assessment methods and assembly systems and based on quality of research i.e. from a validity perspective (RA2: Validation of DYNAMO++ and the concept model)

The validity perspective can be divided into four different types: External validity, Construct validity, Internal validity, and Reliability (Yin, 2003). The DYNAMO++ methodology and the concept model could be quality-checked by discussing these four areas. External validity means establishing a domain in which the study could be generalised. The externally validity has mostly been done in the assembly area context. External validity has been achieved through the thirteen case studies that have been conducted; according to Glaser and Strauss (Glaser and Strauss, 1967), it is the intimate connection with empirical reality that permits the development of a testable, relevant, and valid theory. Many researchers (Miles and Huberman, 1994, Strauss, 1987, Yin, 2003) agree that using multiple case studies could be one way to ensure quality and external validity. To reach further external validity the approach of triangulation has been used, described in Chapter 5.3.1. It is important to consider both qualitative and quantitative methods when performing a case study to broaden the perspective in terms of cause and effect and relation between different sets of parameters (Wacker, 1998). Construct validity is accomplished by constructing correct measures for the concept that are being studied. Since the concept of Levels of Automation was the key point to measure and analyse, this is what has been measured. Frohm’s taxonomy was used as an assessment method because it has been tested and validated in previous research projects and case studies (Granell, et al., 2007). Further it contains both cognitive and physical levels of automation which was the concept that should be tested. By using data and investigator triangulation, this increases the validity of the research (Yin, 2003). Internal validity is achieved by establishing a causal relation. This was done by performing the thirteen case studies using triangulation of multiple investigators, within-case and cross-case analyses, and existing literature, which according to (Eisenhardt, 1989) are good tools in order to build theory based on case study research. Structured interviews and well documented measures were done in most of the cases, and this information was then used in order to do cross-case analyses. One tactic is to select categories or dimensions, and then to look for within-group similarities coupled with intergroup differences. Categories chosen in these cases were Levels of Automation, Triggers for Change and lean awareness. Results from six of the studies are presented in paper 3 (Fasth and Stahre, 2008). Reliability is to ensure that the operation of the study could be repeated with the same results. Two issues are important in reaching closure: when to stop adding cases, and when to stop iterating between theory and data. Ideally, researchers should stop adding cases when theoretical saturation is reached (Eisenhardt, 1989) (Theoretical saturation is simply the point at which incremental learning is minimal because the researchers are observing phenomena seen before,(Glaser and Strauss, 1967)). The final products of building theory from case studies may be concepts (e.g., the Mintzberg and Waters, 1982, deliberate and emergent strategies) or a conceptual framework (e.g., Harris & Sutton's, 1986, framework of bankruptcy). This is in line with results from paper 1, e.g. the development of the concept model developed from case studies and the DYNAMO++ methodology. Further, both novices and experienced users have been able to follow the methodology in a simple way, i.e. investigator triangulation, which increases the reliability.

47

5.2.4

PRACTICAL CONTRIBUTION FROM AN INDUSTRIAL PERSPECTIVE

This section will discuss the practical contribution with the results in RQ1 in relation to the theoretical framework of Levels of Automation (LoA). (RA3: Use of DYNAMO++ and the concept model.)

Results from the case studies show that the issue of Levels of Automation is often seen from a physical perspective, focusing on reducing cost as a primary parameter. Furthermore, results from the literature study of assessment methods, presented in paper 4, reveal that there is a lack of methods which consider both physical and cognitive automation as a prime parameter. By introducing the DYNAMO++ and the concept of LoA, i.e. to consider both the physical and cognitive LoA and to investigate other triggers for change, the companies’ view broadens when it comes to solutions for redesigning the system. The most illustrative and quantitative step in the method is the LoA-matrix. The matrix, illustrated in Figure 21, shows collected measures from ten case studies. Results from paper 1 [Fasth, 2008; Fasth, 2010; etc.] show that there is a need for a more detailed scale. If these tasks had been assessed with only three levels of physical automation (manual, semi-automatic, automatic), all tasks would have been classified as manual tasks. The more detailed scale with four (sometimes five) levels could be used in order to make finer improvements. In line with Porras and Robertsson (Porras and Robertsson, 1992), there is still a need for minor changes in a system. For example the physical LoA, level 1-4 could be used in order to determine and improve the (hand) tool that is used. This information could also be used in ergonomic studies. For the cognitive LoA, Levels 1-4 could tell the engineer what kind of support is used, for example whether it is working orders, sequencing (or kitting), pick-by-light etc.

FIGURE 21 OBSERVED TASKS FROM TEN CASE STUDIES, DOCUMENTED IN THE LOA MATRIX

The LoA-matrix is an easy-to-use tool for quantifying LoA in an illustrative way that could be used and understood from shop-floor level up to site manager level.

48

This section will discuss the practical contribution with the results from the appended papers related to RQ3 in relation to the theoretical framework related to Effects and Assembly systems. (RA3: Shown effects of measuring and analyzing LoA)

Results from the appended papers have shown possible effects and consequences when considering LoA and three other relations (1, Level of Competence (LoC), Level of Information (Lei) and Proactivity; 2, Time and Flexibility; and 3, Quality and Complexity). These will be discussed in the following section. Competence (LoC), Information (LoI) and Proactivity In order to create a possible proactive work setting, the ability in terms of competence and tools in terms of information system and automation solutions has to be available to help the individual to take self-directed action, to anticipate or initiate change in the work system or work roles (Griffin, et al., 2007) and to handle frequent changes within the system (Dencker, 2011, Fasth, et al., 2010). Results from paper 5 show that it is a strong relation between the areas of automation, competence and information flow. Further, they show that all areas have to be considered in order to increase proactive behaviour among the operators. It is shown that if changing LoA the other areas will be affected; the challenge is to understand how much effort it takes for the operators and the system if changing LoA in terms of education, investments of soft- and hardware etc. Time and Flexibility More flexibility in manufacturing operations means more ability to adapt to customer needs, respond to competitive pressures, and be closer to the market (Slack, 2005). Results from papers 1 and 3 show that it is important to consider the production layout when examine possible solution. By combining different levels of automation, routing flexibility was created. Further, volume flexibility and cycle time were also affected in a positive way by the solutions. Quality and Complexity Earlier empirical results (Guimaraes, et al., 1999) show that in general, system complexity does affect performance negatively, and that training and the man/machine interface play important roles in minimizing the negative effect of system complexity on performance. Results from previous sections show that relations could be made between quality, complexity and cognitive automation. Beliefs are that cognitive automation can be used as a means to reduce the negative effects of choice complexity in terms of quality. Results from paper 6 show that it is possible to use quantitative measures in order to show relations between station complexity, quality and cognitive automation. These methods could be further used in order to improve both the resource efficiency and resource allocation in order to get an effective assembly system. Then, the operators’ competence and experience should also be taken into consideration, which is not fully covered by using only these quantitative methods. The main conclusion is that there is evidence that cognitive support is needed in final assembly to minimize the   negative effects of complexity.

49

5.3 EMPIRICAL AND THEORETICAL COLLECTION AND ANALYSIS  The following sections will discuss the results from the empirical and theoretical collection base on two perspectives i.e. the triangulation perspective and the validation perspective.

5.3.1 TRIANGULATION PERSPECTIVE Within the triangulation perspective there are four areas of empirical collection that could be discussed according to (Dezin, 1970) and (Olsen, 2004). The use of these four perspectives will be discussed in the following section. 1. Data triangulation Data triangulation means using a variety of data sources in a study. Both quantitative and qualitative data were collected within the case studies in order to achieve data triangulation, but also to get a more nuanced picture of the examined area, i.e. both “numbers and words” (Miles and Huberman, 1994). Furthermore, multiple case studies give a more generalised data collection. The results from the empirical collections were mostly used in papers 1, 2, 4 and 5. A drawback is that lab experiments have not been done to the extent that was wished for in order to collect even more data to strengthen some relations between LoA and other parameters in a more controlled way. This will be done in future work. 2. Investigator triangulation The use of several different researchers or evaluators creates investigator triangulation but also reliability within the validation step of a methodology. Results in papers 4 and 5 are based on investigator triangulation though data gathering and the analysis were made in cooperation between three or more researchers. This method has not been used solely by industrial partners; the aim is that the method will be used by production engineers as a statistic and analytical tool for every-day followups and for strategic investments. 3. Theory triangulation - To use a multiple perspective to interpret a single set of data. Theory has been used from different domains in order to achieve theory triangulation. There are still other areas of interest that could be combined in order to make the concept model more logical, automated and generalised. 4. Methodology triangulation - Using multiple methods to study a single problem or phenomenon. This is mainly done in papers 4 and 5 where three different methods in each paper are used in order to study the relation between Levels of automation and Proactivity in paper 4 and the relation between complexity, quality and Cognitive automation in paper 5. Furthermore, the use of both quantitative and qualitative methods also strengthens the methodology triangulation. .

50

5.4 FUTURE WORK  Results from the case studies conducted for this thesis show that more than 80 % of the tasks are performed based on the operators’ own experience (see Figure 24, p.48). Some assumptions on why it looks like this could be: experienced operators that work on final assembly line at some case companies. The fact that some companies have u-cells with one product family and thereby decrease the variant complexity could be another explanation. This manual level could affect other parameters in the system, such as time, quality etc. Results in paper 5 show a possible correlation between cognitive LoA, Quality and complexity. And as a part of a future study, the relation between different Levels of Automation and other parameters will be investigated in a more controlled environment, i.e. lab tests. This will be done in order to accomplish more statistically controlled experiments. A second part of the future work will be to further develop the automated planning system regarding local resource allocation (Fasth, et al., 2012, Provost, et al., 2012) and automated generated work instruction due to different LoAs. Also it will investigate further the work done on algorithm developments (de Winter and Dodou, 2011) according to dynamically changeable function allocation with a human centred approach (not left-over function allocation) (Hoc, 2001). A related part will be to perform lab experiments regarding cognitive automation and different types of carrier and content to further develop and test these concepts (Fasth and Stahre, 2010, Fässberg, et al., 2012). In order to make the DYNAMO++ and the concept model more generalised it would be interesting to apply the concept within other contexts such as manufacturing, health, mining etc.

51

52

6 CONCLUSION This chapter contains conclusions related to the aim. The aim of this thesis is: By quantifying, measuring and analysing the physical and cognitive Levels of Automation, enable competitive assembly systems. Results presented in Paper 1 show that there is RQ 1: Why is it important to quantify Levels of an ongoing debate, both in industry and Automation (LoA) in an assembly system context? academia on how to define physical automation, RQ 2: How should LoA be quantified, measured and cognitive automation; and Levels of analysed in assembly systems? Automation. The debaters might not even be explicit that it is automation they are arguing RQ 3: What are the expected effects of analysing and about, especially when it comes to cognitive changing LoA? automation. Therefore, in order to reach the aim, definitions of physical automation, cognitive automation and Levels of Automation need to be clear and agreed upon as a first step. The author defines the terms as follows:   

Physical automation is defined as: “technical solutions helping the operator to assembly the products e.g. WITH WHAT to assemble”. Cognitive automation is defined as: “technical solutions helping the operator e.g. HOW to assemble (Levels 1-4) and situation control (Levels 5-7)”. Levels of Automation is defined as:” the allocation of physical and cognitive tasks between resources (humans and technology), described as discrete steps from 1 (totally manual) to 7 (totally automatic), forming a 7 by 7 LoA matrix containing 49 possible types of solutions”.

With that definition of Levels of Automation, results from paper 1 show that it is possible to quantitatively measure automation based on seven ranking steps and also to visualise the result in the matrix. Further, companies could develop a company-specific “LoA-language” based on own examples within the matrix. This common language could be a help when discussing new LoA– investments. The LoA matrix with quantitative discrete steps could also be one way to illustrate the current state and be a base for discussion from operator level up to management level. Furthermore it is an easy way to compare different cells with each other, and could be used to find correlation with other parameters, i.e. quality, time, flexibility etc. Furthermore, results have shown possible effects and consequences when considering LoA and:   

Competence (LoC), Information (LoI) and Proactivity Time and Flexibility and Quality and Complexity

[Paper 5] [Paper 1 and 3] [Paper 6]

The developed and validated methodology, DYNAMO++, and concept model [Paper 1,2, 3 and 4], gives the possibility to, in a structured way, analyse the current state and to find possible solution for the future that could be further analysed and specified in terms of simulation models etc. The concept model gives an understanding of the different areas that are important to consider in relation to LoA. I therefore conclude that dynamically changeable automation and collaboration between highlyskilled humans and technology is the answer for competitive assembly systems.

53

54

REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27.

Almström, P., and Kinnander, A., PRODUCTIVITY POTENTIAL ASSESSMENT OF THE SWEDISH MANUFACTURING INDUSTRY. Proceedings from the 1st Swedish Production Symposium, Gothenburg, Sweden, 2007 Almström, P., and Kinnander, A., PRODUCTIVITY POTENTIAL ASSESSMENT OF 30 SUPPLIERS TO THE AUTOMOTIVE INDUSTRY. Proceedings from the 3rd Swedish Production Symposium, Stockholm, Sweden, 2009 Alting, L., 1978. “A Systematic Theory of Manufacturing”. Environment and Planning B vol. 5, 131-156. Amber, G. H., and Amber, P. S., 1962. Anatomy of Automation. Prentice Hall. Andersson, C., Jönsson, M., and Ståhl, J.-E., Capacity flexibility and part cost. Proceedings of the 3rd Swedish Production Symposium (SPS), Stockholm, Sweden, 2009. Andreasen, M. M., Kähler, S., and Lund, T., 1983. Design for Assembly. U.K. Asan, S. S., 2009. A Methodology Based on Theory of Constraints’ Thinking Processes for Managing Complexity in the Supply Chain, Univ. Berlin, Germany. Barnevik, P., 1991. Article in work life in transition, Sympohosium om arbetsliv och produktivitet. Arbetsmiljöfonden och LO, Stockholm. Bellgran, M., 1998. Systematic design of assembly systems, Linköping University, Linköping. Bellgran, M., and Säfsten, K., 2005. Produktionsutveckling - Utveckling och drift av produktionssystem. Studentlitteratur, Lund, Sweden. Bicheno, J., 2006. Ny verktygslåda för Lean - för snabbt och flexibelt flöde. Revere AB. Billings, C., 1997. Aviation Automation: The search for a Human-centered approach. Lawrence Erlbaum Associates, Mahwah, New Jersey, USA. Blecker, T., Kersten, W., and Meyer, C. M., 2005. Development of an Approach for Analyzing Supply Chain Complexity, Hamburg Univ. of Techn.,MPRA Paper #5284, http://mpra.ub.unimuenchen.de/5284/. Bright, J., 1958. Automation and Management. Boston, USA. Browne, J., Dubios, D., Rathmill, K., Sethy, S. P., and Stecke, K. E., 1984. Classification of Flexible manufacturing systems. The FMS Magazine, April. Bruch, J., Johansson, C., Karltun, J., and Winroth, M., Considering design demands of a proactive assembly system - A position paper. 1st Swedish Production Symposium (SPS), Gothenburg, 2007. Bruch, J., Karltun, J., and Dencker, K., A proactive assembly work setting – Information requirements. The 41st CIRP Conference on Manufacturing Systems, Tokyo, Japan, 2008. Campbell, D. T., December,1976. Assessing the Impact of Planned Social Change. In: The Public Affairs Center, D. C., (Ed., Hanover New Hampshire. Chapanis, A., 1996. Human Factors in Systems Engineering. John Wiley & Sons, Inc., New York. Chryssolouris, G., 2006. Manufacturing systems - theory and practice. Springer, New York. Cochran, D. S., 1998. Introduction to Production System Design: The System Design Differences between Mass and Lean Production, MIT Visteon Lean Manufacturing Workshop, Monroe MI, USA. Comstock, M., 2004. Production systems for mass customization - bridging theory and practice, Linköping university, Linköping. Comstock, M., and Bröte, S., Beyond 'Read a plant - fast' (for lean): read an enterprise for mass customization? , Proceedings of the 2nd international conference on mass customization and personalization Munich, Germany, 2003. Crant, J. M., 2000. Proactive behavior in organizations. Journal of Management 26, 435-462. Crotty, M., 1998. The foundations of social research: meaning and perspective in the research process. St Leonards, Australia. Danermark, B., Ekström, M., Jakobsen, L., and Karlsson, J. C., 2003. Att förklara samhället. Studentlitteratur, Lund, Lund. de Winter, J., and Dodou, D., 2011. Why the Fitts list has persisted throughout the history of function allocation. Cognition, Technology & Work, 1-11. 55

28. 29. 30.

31.

32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44.

45.

46. 47.

Dekker, S., and Woods, D., 2002. MABA_MABA or ABRAKADABRA? Progress on Human-Automation Co-ordination. Cognition, Technology and Work 4, 240-244. Dencker, K., 2011. An analysis of a proactive approach as a potential tool for adaptability in production systems, KTH, Stockholm. Dencker, K., Stahre, J., Gröndahl, P., Mårtensson, L., Lundholm, T., Bruch, J., and Johansson, C., 2007. PROACTIVE ASSEMBLY SYSTEMS- Realizing the potential of human collaboration with automation, Proceedings of the IFAC-CEA, cost effective automation in networked product development and manufacturing, Monterey, Mexico. Dencker, K., Stahre, J., Gröndahl, P., Mårtensson, L., Lundholm, T., and Johansson, C., 2007. An Approach to Proactive Assembly Systems - Towards competitive assembly systems, IEEE international symposium on assembly and manufacturing (ISAM), University of Michigan, Ann Arbor, USA.ISAM. Dencker, K., Stahre, J., Mårtensson, L., Fasth, Å., and Akillioglu, H., 2009. PROACTIVE ASSEMBLY SYSTEMS- Realizing the potential of human collaboration with automation. Annual Reviews in Control, 33:3 IFAC/Pergamon, ISSN: 1367-5788. Dencker, K., Stahre, J., Mårtensson, L., Fasth, Å., and Akillioglu, H., conditionally accepted for publication. PROACTIVE ASSEMBLY SYSTEMS- Realizing the potential of human collaboration with automation. Elsevier, Annual Reviews in Control. Dezin, N. K., 1970. The research act: A theoretical introduction to sociological methods. Aldine, Chicago. Duncheon, C., 2002. Product miniaturization requires automation - but with a strategy. Assembly Automation 22, 16-20. Easton, G., and Harrison, D., 2004. Temporally embedded Case comparison in industrial marketing research. In: Critical realism in management studies. Eisenhardt, K. M., 1989. Building Theories from Case Study Research The Academy of Management Review 14, 532-550. Endsley, M., and Kiris, E., 1995. Out-of-the-loop performance problem and level of control in automation. HUMAN FACTORS - Human Factors and Ergonomics Society. 37, 381-394. Endsley, M. R., Level of Automation: Integrating humans and automated systems. Proceedings of the 1997 41st Annual Meeting of the Human Factors and Ergonomics Society. Part 1 (of 2), Albuquerque, NM, USA, Santa Monica, CA, USA, 1997. Fasth, Å., 2009. Measuring and Analysing Levels of Automation in assembly systems - For future proactive systems, Chalmers University of Technology, Gothenburg. Fasth, Å., Comparing methods for design and measurement of production systems. Proceedings of the 4th Swedish Production Symposium (SPS), Lund, Sweden, 2011. Fasth, Å., 2011. Comparing methods for redesigning, measuring and analysing Production systems, Proceedings of the 4th Swedish Production Symposium (SPS), Lund, Sweden. Fasth, Å., Accepted for publication. REVIEWING METHODS FOR ANALYSING TASK ALLOCATION IN A PRODUCTION SYSTEM International journal of logistic management. Fasth, Å., Bruch, J., Dencker, K., Stahre, J., Mårtensson, L., and Lundholm, T., 2010. Designing proactive assembly systems (ProAct) - Criteria and interaction between automation, information, and competence Asian International Journal of Science and Technology in production and manufacturing engineering (AIJSTPME) 2 (4), 1-13. Fasth, Å., Mattsson, S., Fässberg, T., Stahre, J., Höög, S., Sterner, M., and Andersson, T., Development of production cells with regard to physical and cognitive automation - A decade of evolution International Symposium on Assembly and Manufacturing (ISAM 11), Tampere, Finland, 2011. Fasth, Å., Provost, J., Stahre, J., and Lennartson, B., From task allocation towards resource allocation when optimising assembly systems. 45th CIRP Conference of Manufacturing Systems (CMS), Athen, Greece, 2012. Fasth, Å., and Stahre, J., Does Levels of Automation need to be changed in an assembly system? - A case study. Proceedings of the 2nd Swedish Production Symposium (SPS), Stockholm, Sweden, 2008.

56

48. 49. 50. 51. 52. 53. 54.

55. 56. 57. 58. 59. 60. 61. 62.

63. 64. 65. 66. 67. 68. 69.

Fasth, Å., and Stahre, j., Concept model towards optimising Levels of Automation (LoA) in assembly systems. Proceedings of the 3rd CIRP Conference on Assembly Technologies and Systems, Trondheim, Norway, 2010. Fasth, Å., Stahre, J., and Dencker, K., Analysing changeability and time parameters due to levels of Automation in an assembly system. Proceedings of the 18th conference on Flexible Automation and Intelligent Manufacturing - FAIM, Skövde, Sweden, 2008 Fasth, Å., Stahre, J., and Dencker, K., Measuring and analysing Levels of Automation in an assembly system. Proceedings of the 41st CIRP conference on manufacturing systems Tokyo, Japan, 2008 Feldmann, K., and Slama, S., 2001. Highly flexible Assembly - Scope and Justification. CIRP Annals - Manufacturing Technology 50, 489-498. Fitts, P., 1951. Human engineering for an effective air navigation and traffic control system, Ohio state university, Columbus,OH. Fjällström, S., 2007. The role of information in production ramp-up situations, Chalmers university of technology, Gothenburg. Flegel, H., Jovane, F., Williams, D., Westkämper, E., Bueno, R., Richet, R., Hanisch, C., Groothedde, R., Chlebus, E., and Mendonça, J. M., 2005. ManuFuture Platform-STRATEGIC RESEARCH AGENDA, assuring the future of manufacturing in Europe, Manufuture Platform-Executive Summary, Manufuture Conference 2005 “Making it in Europe”, Rolls Royce, Derby, UK, EC. Flynn, B. B., Sakakibara, S., Schroeder, R. G., Bates, K. A., and Flynn, E. J., 1990. Empirical research methods in operations management. Journal of Operations Management 9, 250-284. Frese, M., and Fay, D., 2001. Personal Initiative: An Active Performance Concept for Work in the 21st Century, in Research in Organizational Behavior. Edited by Staw,B.M., Sutton, E.L., Elsevier Science, 133-187. Frizelle, G., and Suhov, Y. M., 2001. An entropic measurement of queuing behaviour in a class of manufacturing operations. In proc. Royal Society London 457, 1579-1601. Frohm, J., 2008. Levels of Automation in production systems, (Doctorial), Chalmers University of technology, Gothenburg. Frohm, J., Lindström, V., Winroth, M., and Stahre, J., 2008. Levels of Automation in Manufacturing. Ergonomia IJE&HF, 30:3. Fuld, R. B., 2000. The fiction of function allocation, revisited. International Jurnal of HumanComputer Studies 52, 217-233. Fässberg, T., Fasth, Å., and Stahre, J., 2012. Classification of carrier and content of information, 4th CIRP Conference on Assembly Technologies and Systems (CATS), Ann Arbor, USA. Fässberg, T., Harlin, U., Garmer, K., Gullander, P., Fasth, Å., Mattsson, S., Dencker, K., Davidsson, A., and Stahre, J., AN EMPIRICAL STUDY TOWARDS A DEFINITION OF PRODUCTION COMPLEXITY. 21st International Conference on Production Research, Stuttgart, Germany, 2011. Fässberg, T., Nordin, G., Fasth, Å., and Stahre, J., iPod Touch - an ICT tool for assembly operators in factories of the future? , Proceedings of the 43rd CIRP International Conference On Manufacturing Systems (ICMS), Vienna, Austria, 2010. Gerwin, D., 1983. A framework for Analyzing the Flexibility of Manufacturing Process, University of Wisconsin, Wisconsin. Gilbreth, F., 1911. Motion Study. D. Van Nostrand Co. Glaser, B. G., and Strauss, A., 1967. The discovery of grounded theory: Strategies of qualitative research. Goodson, R., 2002. Read a plant - fast. Harvard business review. Granell, V., Frohm, J., Bruch, J., and Dencker, K., 2007. Validation of the DYNAMO methodology for measuring and assessing Levels of Automation, Proceedings of the 1st Swedish Production Symposium (SPS), Gothenburg. Griffin, M. A., Neal, A., and Parker, S. K., 2007. A New Model of Role Performance: Positive Behaviour in Uncertain and Interdependent Contexts. Academy of Management Journal 50, 327-347. 57

70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96.

Groover, M. P., 2001. Automation, production systems, and computer-integrated manufacturing. Prentice Hall, Upper Saddle River, N.J. Grote, G., 2004. Uncertainty management at the core of system design. Annual Reviews in Control 28, 267-274. Guimaraes, T., Mårtensson, N., Stahre, J., and Igbaria, M., 1999. Empirically testing the impact of manufacturing system complexity on performance. International Journal of Operations & Production Management 19, 1254-1269. Hancock, H. A., and Chignell, M. H., 1992. Adaptive allocation by intelligent interfaces. Hoc, J.-M., 2000. From human-machine interaction to human-machine cooperation. Ergonomics 43, 833. Hoc, J.-M., 2001. Towards a cognitive approach to human-machine cooperation in dynamic situations. International Journal of Human-Computer Studies 54, 509-540. Hollnagel, E., 1996. Reliability analysis and operator modelling. Reliability Engineering and System Safety 327-337. Hollnagel, E., 1998. HRA -- The First Generation, Cognitive Reliability and Error Analysis Method (CREAM). Elsevier Science Ltd, Oxford, pp. 120-150. Hollnagel, E., 2003. Handbook of Cognitive Task Design. Routledge, Mahwah, New Jersey. Hollnagel, E., 2003. The role of Automation in joint cognitive systems. IFAC. Hollnagel, E., and Woods, D. D., 2005. Joint cognitive systems - foundation of cognitive system engineering. CRC press, Taylor and Francis group. Holme, I. M., and Solvang, B. K., 1997. Forskningsmetodik. Om kvalitativa och kvantitaiva metoder. Studentlitteratur, Lund. Hou, T., Lin, L., and Drury, C. G., 1993. An empirical study of hybrid inspection system and allocation of inspection functions. International journal of human factors in manufacturing systems, 351-367. http://www.cirp.net, 2008. Inagaki, T., 2003. Automation and the cost of authority. International Journal of Industrial Ergonomics 31, 169-174. Isard, M., Prabhakaran, V., Currey, J., Wieder, U., Talwar, K., and Goldberg, A., Quincy: Fair Scheduling for Distributed Computing Clusters. SOSP 2009. Jain, R., Chiu, D.-M., and Hawe, W., 1984. A quantitative measure of fairness and discrimination for resource allocation in shared computer systems. DEC Research Report TR301 9. Johansson, A., Decision support tools in manufacturing systems: a task evaluation method. European conference on integration in manufacturing, Amsterdam, 1994. Jordan, N., 1963. Allocation of functions between human and machine in automated systems. Journal of applied psychology 47, 161-165. Jönsson, M., Andersson, C., and Ståhl, J.-E., 2008. Implementation of an Economic Model to Simulate Manufacturing Costs, The 41st CIRP conference on manufacturing systems Tokyo, Japan. Kantowitz, B. H., and Sorkin, R. D., 1987. Handbook of human factors. Ch 3.3 Allocation of functions. Wiley, New York. Kerlinger, F. N., 1960. The mythology of educational research: The methods approach. School and Society 88, 149-151. Kern, and Schumann, 1985. Das Ende das Arbeitsteilung Verlag Beck. Koho, M., 2010. Production System Assessment and Improvement - A tool for MTO and ATO companies, Tampere university of technology, Tampere. Koren, Y., Heisel, U., Jovane, F., Moriwaki, T., Pritschow, G., Ulsoy, G., and Van Brussel, H., 1999. Reconfigurable Manufacturing Systems. CIRP Annals - Manufacturing Technology 48, 527-540. Kuhn, R., 1962. Structure of scientific revolution Lagu, A. V., and Landry, S. J., 2010. Roadmap for the Next Generation of Dynamic Function Allocation Theories and Strategies. Human Factors and Ergonomics in Manufacturing & Service Industries 00, 1-15.

58

97. 98. 99. 100. 101. 102.

103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. 116. 117. 118. 119. 120. 121.

Lee, J. D., 2008. Review of a pivotal Human Factors Article: "Human and Automation: Use, Misuse, Disuse, Abuse". Golden anniversary special issue of Human Factors 50, 404-410. Lennartson, B., Bengtsson, K., Yuan, C., Andersson, K., Fabian, M., Falkman, P., and Åkesson, K., 2010. Sequence Planning for Integrated Product,Process and Automation Design. IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING 7. Löfgren, K. G., 1983. Produktion - Teknik och Ekonomi. LiberHermods, Malmö. March, R., and Mannari, H., 1981. Technology and size as determinants of the organizational structure of Japanese factories. Administrative science quarterly 26, 33-57. Mattson, S.-A., 2004. Logistikens termer och begrepp. PLAN Föreningen för Produktionslogistik, Stockholm. May, C., Mair, F., Finch, T., MacFarlane, A., Dowrick, C., Treweek, S., Rapley, T., Ballini, L., Ong, B., Rogers, A., Murray, E., Elwyn, G., Legare, F., Gunn, J., and Montori, V., 2009. Development of a theory of implementation and integration: Normalization Process Theory. Implementation Science 4, 29. Merriam, S. B., 1994. Fallstudien som forskningsmetod. Studentlitteratur, Lund. Miles, and Huberman, 1994. Qualitative Data Analysis. Mårtensson, L., and Stahre, J., 2003. Skill and role development in swedish industry, Proceedings of the 8th IFAC Symposium on Automated Systems Based on Human Skill and Knowledge. Nof, S. Y., 2009. Handbook of Automation, Chapter 3. Nyhuis, P., Kolakowski, and M., H., C. L, 2005. Evaluation of Factory Transformability, 3rd International CIRP Conference on Reconfigurable Manufacturing, Ann Arbor, USA. Older, M. T., Waterson, P. E., and Clegg, C. W., 1997. A critical assessment of task allocation methods and their applicability. Ergonomics 40, 151-171. Olsen, W., 2004. Developments in sociology (Ch. 4 Triangulation in social research: Qualitative and quantitative methods could really be mixed). Onori, M., Barata, J., Hanisch, C., Maraldo, T., and D, S., 2006. EUPASS Technology Roadmap v.1.0. NMP-2-CT-2004-507978. Parasuraman, R., and Riley, V., 1997. Human and Automation: Use, Misuse, Disuse, Abuse. Human factors 39(2). Parasuraman, R., Sheridan, T. B., and Wickens, C. D., 2000. A model for types and levels of human interaction with automation. IEEE transactions on system, man, and cybernetics - Part A: Systems and humans 30, 286-296. Parasuraman, R., and Wickens, C. D., 2008. Humans: Still Vital After All These Years of Automation. Golden anniversary special issue of Human Factors 50, 511-520. Parker, S. K., Williams, H. M., and Turner, N., 2006. Modeling the Antecedents of Proactive Behaviour at Work. Journal of Applied Psychology 91, 636-652. Patel, R., and Davidsson, B., 2003. Forskningsmetodikens grunder. Studentlitteratur, Lund. Porras, J. I., and Robertsson, P. J., 1992. Organizational Development: Theory, Practice and Research in Handbook of industrial and organisational psycology. Consulting Psychologist Press inc, California. Prince, H., 1985. The allocation of function in systems. Human Factors 27, 33-45. Provost, J., Fasth, Å., Stahre, J., Lennartsson, B., and Fabian, M., Human operator and robot resource modelling for planning purposes in assembly systems 4th CIRP Conference on Assembly Technologies and Systems (CATS), Ann Arbor, USA, 2012. Rampersad, H. K., 1994. Integrated and simultaneous design for robotic assembly. Chichester, England. Rasmussen, J., 1983. Skills, Rules, Knowledge, Signals, Signs and Symbols and other Distinctions Human Performance Models. IEEE Transactions on Man, Systems and Cybernetics, 257-266. Rasmussen, J., 1985. The Role of Hierarchical Knowledge Representation in Decision-making and System Management. IEEE Transactions on Man, Systems and Cybernetics SMC-15/2, 234-243.

59

122. 123. 124. 125. 126. 127. 128. 129. 130. 131. 132. 133. 134. 135. 136. 137. 138. 139. 140. 141. 142. 143. 144. 145. 146. 147. 148. 149.

Rasmussen, J., Pejtersen, A. M., and Goodstein, L. P., 1994. Taxonomy for work analysis. Design of work and development of personnel in advanced manufacturing. John Wiley & Sons,inc p. 41-48, New York. Rasmussen, J., Pejtersen, A. M., and Schmidt, K., 1990. Taxonomy for Cognitive Work Analysis. Grafisk service, Roskilde, Denmark. Rother, M., and Shook, J., 2002. Learn how to see. In: Lean, (Ed. Lean enterprise institute, Sweden. Satchell, P., 1998. Innovation and Automation. London, England. Sekaran, U., 2000. Research methods for business - a skill-building approach. Seliger, G., Viehweger, B., and Wieneke, B., 1987. Decision Support in design and optimization of flexible automated manufacturing and assembly. Robotics and computerintegrated manufacturing 3, 221-227. Sherehiy, B., Karwowski, W., and Layer, J. K., 2007. A review of enterprise agility: Concepts, frameworks, and attributes. International Journal of Industrial Ergonomics, 445-460. Sheridan, T., and Parasuraman, R., 2006. Human-automation interaction. Human factors and ergonomics 1, 89-129. Sheridan, T. B., 1980. Computer control and Human Alienation. Technology Review 83, 6073. Sheridan, T. B., 1992. Telerobotics, automation and human supervisory control. MIT Press, Cambridge Massachussetts. Sheridan, T. B., 1995. Human centred automation: oxymoron or common sense?, Intelligent Systems for the 21st Century, IEEE international. Proceedings of: Systems, Man and Cybernetics, pp. 6. Sheridan, T. B., 2000. Function allocation: algorithm, alchemy or apostasy? International Journal of Human-Computer Studies 52, 203-216. Skinner, W., 1969. Manufacturing—missing link in corporate strategy Harvard Business Review May-June, 136-145. Slack, N., 2005. The flexibility of manufacturing Systems. International Journal of Operations & Production Management Vol. 25 No. 12. Smith, K., 2004. Living the Vision Seventy Years Later. Maynard exchange 9. Stahre, J., 1995. Evaluating human/machine interaction problems in advanced manufacturing. Computer Integrated Manufacturing Systems 8, 143-150. Stahre, J., 1995. Towards human supervisory control in advanced manufacturing systems, Chalmers university of technology, Gothenburg. Starrin, B., and Svensson, P.-G., 1994. Kvalitativ metod och vetenskapsteori. Studentlitteratur, Lund. Stoessel, C., Wiesbeck, M., Stork, S., Zaeh, M. F., and Schuboe, A., 2008. Towards optimal worker assistance: Investing Cognitive processes in manual assembly, The 41st CIRP conference on manufacturing systems Tokyo, Japan. Strauss, A., 1987. Qualitative research for social scientists. Camebridge university press. Ståhl, J.-E., 2006. Industriella tillverkningssystem - länken mellan teknik och ekonomi. Lunds tekniska högskola, Lund. Ståhl, J.-E., 2007. Industriella tillverkningssystem - länken mellan teknik och ekonomi. Lunds tekniska högskola, Lund. Säfsten, K., 2002. Evaluation of assembly systems - an exploratory study of evaluation situations, Linköping university, Linköping. Säfsten, K., and Aresu, E., 2000. Vad är bra monteringssystem?: En studie av utvärdering och utformning på 15 industriföretag i Sverige. Linköpings universitet, Linköping. Säfsten, K., Winroth, M., and Stahre, J., 2007. The content and process of automation strategies. International Journal of Production Economics 110, 25-38. Tangen, S., Axelsson, J. v., Dencker, K., and Gröndahl, P., 2008. Strategi och produktionsutveckling, chapter 1. WoxénCentrum, Stockholm. Taylor, F. W., 1911. Principles of Scientific Management. Tesar, D., 2002. Meeting Human Needs While Reducing Human Drudgery The University of Texas at Austin 60

150. 151. 152. 153. 154. 155. 156. 157. 158. 159. 160. 161. 162. 163. 164.

165.

Wacker, J. G., 1998. A definition of theory: research guidelines for different theory-building research methods in operations management. Journal of Operations Management 16, 361-385. Waterson, P. E., Gray, M. T. O., and Clegg, C. W., 2002. A socio-technical method for designing work systems. Human Factors 44, 376. Weaver, W., 1948. Science and Complexity. American Scientist 36. Westkämper, E., Digital Manufacturing in the global Era. Proceedings of the 3rd International CIRP Conference on Digital Enterprise Technology, Setúbal, Portugal, 2006. Westkämper, E., Manufuture and sustainable manufacturing. Proceedings of the 41st CIRP conference on manufacturing systems Tokyo, Japan, 2008. Wiendahl, H.-P., 2002. Wandlungsfähigkeit: Schlüs-selbegriff der zukunftsfähigen Fabrik (Transformabi-lity: key concept of a future robust factory), Werkstattstechnik online. Wiendahl, H. P., ElMaraghy, H. A., Nyhuis, P., Zäh, M. F., Wiendahl, H. H., Duffie, N., and Brieke, M., 2007. Changeable Manufacturing - Classification, Design and Operation. CIRP Annals - Manufacturing Technology 56, 783-809. Wilkinson, A. M., 1991. The scientists hanbook for writing papers and dissertations. Englewood cliffs, New jersey. Williams, T., 1977. 20 years of computer control. Can. Control 16 (12). Williams, T. J., 2009. Handbook of Automation, Chapter 2. Winroth, M., 2006. DYNAMO WP1, Automation Strategies, Jönköping. Winroth, M., Säfsten, K., and Stahre, J., 2005. Automation strategies - existing theroy or ad hoc decisions?, 16th Annual conference of POMS, Chicago, IL. Womack, J., Jones, D., and Roos, D., 1990. The Machine that Changed the World. Harper Collins, New York. Wright, P., Dearden, A., and Fields, B. O. B., 2000. Function allocation: a perspective from studies of work practice. International Journal of Human-Computer Studies 52, 335-355. Wäfler, T., Johansson, A., Grote, G., and Stahre, J., Complementary interaction of humans and machines in highly automated production systems - KOMPASS and TEAM. 13th Triennial Congress of the international Ergonomic Association, Finnish institute of occupational health, 1997, pp. 227-279. Yin, R. K., 2003. Case Study Research: Design and Methods. London, Sage.

61

62

APPENDED PAPERS

63

PAPER 1 Fasth, Å. and Stahre, J. (submitted 29 June, 2011), Task allocation in assembly systems – Measuring and analyzing Levels of Automation, special issue (Journal of Theoretical Issues in Ergonomics Science)

Task allocation in assembly systems - Measuring and analysing Levels of Automation Åsa Fasth and Johan Stahre Chalmers University of Technology, Department of Product and Production development

ABSTRACT The paper discusses the need for a quantitative and easy-to-use method, which simultaneously considers physical and cognitive automation. A concept model for task allocation is presented. It consists of a fivestep main loop supported by knowledge needed to enable the main loop. The model and method are compared to Olders’ et al. (1997) sixteen requirements for a task allocation method. The method was developed, validated, and verified together with end-users (industry and novice users) in fifteen companies to verify its practical ease-of-use. Keywords: Task allocation, levels of automation, assembly systems INTRODUCTION In order to maintain sustainable manufacturing in an increasingly globalised industry, current traditions for design and usage of automation in assembly systems may not be adaptable to the needs and future challenges that companies are facing. Rapid changes of demands and requirements, both internal and external, frequently trigger plans for change in different manufacturing areas. Smaller batches and shorter time limits for set-up between products are normal demands on the assembly systems caused by increasing numbers of product variants due to new industrial paradigms e.g. mass customisation. As a result, companies need increasingly flexible methods for assembling products and means to make assembly systems more proactive. One popular solution is automation, thus making it vital to determine the appropriate level of automation. A common industrial predisposition is to consider automation investments as “binary” decision, even though a simple choice between humans or machines for a specific task may be suboptimal. Several development trends towards highly automated production and shop floor workplaces were seen

 

during the 1980's and early 1990's. At that time the predominant task allocation strategy was "left-over allocation". Since the late 1990's trends are changing, much due to obvious shortcomings of automation to fulfil cost and flexibility expectations. Thus, to identify, implement, and maintain the correct level of automation in a controlled way could be a way to radically improve the effectiveness of a system. According to Frohm (2008), to make a manufacturing system as robust, flexible and adaptable as possible, the system must be resilient to process variations, such as the introduction of new products, tool changes, product disturbances etc. It is thus important to understand how to obtain a balanced manufacturing system that has the proper mix of operators and machines in order to e.g. obtain the highest profit possible without suffering loss of product quality. One way to achieve this balanced manufacturing system is to separate the system description into two basic classes of activities, i.e. information handling and physical work. The next step would then be to describe the allocation of tasks within each class, i.e. the “level of automation”. Extensive amounts of research have been done in the area of levels of automation, emphasising different perspectives. Automation research could be divided into three main groups, i.e. 

Mechanical automation (March and Mannari, 1981, Kern and Schumann, 1985, Groover, 2001, Duncheon, 2002)



Information and control automation (Bright, 1958, Sheridan, 1992, Parasuraman and Wickens, 2008, Endsley, 1997, Parasuraman et al., 2000, Hollnagel, 2003a).



Combinations of physical/mechanical and information/cognitive automation (Frohm, 2008).

To determine what to automate, a classical task allocation strategy from 1951 (the MABA-MABA list) was proposed by Fitts (Fitts, 1951). It was an attempt to suggest allocation of tasks between humans and machines by treating them as system resources, each with different capabilities. Two examples, i.e. “Machines Are Better At” performing repetitive and routine tasks while “Men Are Better At” improvising and using flexible procedures. At the time, this was a revolutionary thought causing a lot of debate. Jordan (Jordan, 1963) argued whether you could actually compare man and machine; and that the two should be seen as complementary rather than conflicting resources when designing a man-machine system. Sheridan (Sheridan, 1995) suggested to “allocate to the human the tasks best suited to humans and allocate to the automation the task best suited to it. But, if tasks in which machines are better become automated and operators are still required to monitor the automation, maintaining full situation awareness (Endsley and Kiris, 1995), we might lose more than we gain. Fifty years after Fitts published his list, Hollnagel (Hollnagel, 2003b) argues that the machine (or automation) has been used for three main purposes over the years (which is in line with Fitts) i.e. to ensure more precise performance of a given function; to improve stability of performance by relieving people of repetitive and monotonous tasks; and

to enable processes to be carried out faster and more efficiently. So, do Fitts' thoughts still prevail, or has research turned towards Jordan’s argument? The decision matrix suggested by Prince (Prince, 1985) was partly in line with Fitts in that some tasks were better performed by machines and some better by humans. But interestingly Prince also defined a set of tasks where the same task could and should be performed both by humans and by machines. Further, when there is no single allocation, the different resources need support from each other, which is in line with Jordan’s argument. Hancock (Hancock and Chignell, 1992) argues that it is only when both human and machine can do the same task, the question of task allocation becomes an issue. In line with Jordan, previous research (Hancock and Chignell, 1992, Kantowitz and Sorkin, 1987, Hou et al., 1993, Sheridan, 2000) agrees that the task allocation should been seen as complementary between man and machine rather than dividing tasks solely to one resource. Thus, suitable allocation of tasks between recourses (human operators and machines) and technique has to be made and must be able to be dynamically changeable over time. However, it is common that designers automate every subsystem which leads to an economic benefit for that subsystem but leaves the operator to manage the rest (Parasuraman and Wickens, 2008). Parasuraman et al. (Parasuraman et al., 2000) argues that automation design is not an exact science, however, neither does it belong in the realm of the creative arts, with successful design dependent upon the vision and brilliance of individual creative designers. According to Fasth et al (Fasth and Stahre, 2008) and Säfsten et al. (Säfsten and Aresu, 2000, Bellgran and Säfsten, 2005), a majority of companies studied, have a clear picture of why to change their system. However, the evaluations are often informal and unstructured, i.e. interpretation rather than facts. To choose solutions based solely on experience and interpretation rather than facts and numbers might not be the optimal solution when designing a system. A more reliable and objective quantitative method is therefore needed. A problem related to MABA-MABA-oriented methods is the simplicity e.g. “put your allocation problem into the method and the solution will emerge from the other end” (Dekker and Woods, 2002). The methods do not readily explain the cognitive actions for how and when to intervene, nor do they describe how to switch from level to level. The relevance of a task allocation process is obvious, yet there is still lack of systematic methods and, more importantly, methods that can be applied to advanced technological systems (Older et al., 1997). Another problem with new methods and tools in the human factors area concerns their lack of uptake and use by system developers. New methods must therefore be developed jointly with its users i.e. adaptable to be put in practice (Waterson et al., 2002, Older et al., 1997), furthermore the method must be validated within its planned area of use.

Attempts to create such methods have been made and two have been chosen in this paper; 

Older et al. (Older et al., 1997) in 1997 (and 2002); o

Presents sixteen requirements that a task allocation method should contain, furthermore seventeen predefined criteria for allocation of tasks.

o

Compares eighteen different methods for task allocation, which were developed between 1965 and 1992, to the sixteen requirements.

o 

Presents a socio-technical method, fulfilling the requirements (Waterson et al., 2002)

Fasth, in 2011 (Fasth, 2011); o

Presents four focus areas with focus on assessment objectives, scale and methods, divided into o

Qualitative methods i.e. based on interviews and observations (Hollnagel, 2003a).

o

Quantitative methods i.e. compare the system or tasks to pre-defined criteria of a “perfect system” (Almström and Kinnander, 2009 , Koho, 2010) or task allocation, for example the classic MABA-MABA – list.

o

Compares ten methods (developed between 1990 and 2010) due to two dimensions (sociotechnical and physical-cognitive). Results show that the methods consider either the cognitive/socio (Hollnagel, 2003a, Parasuraman et al., 2000, Trist and Bamforth, 1951, Endsley and Kaber, 1999) or the physical/technical (Taylor, 1911, Almström and Kinnander, 2007 ) part of the system, not both at the same time. The main part of the methods has been put in practice for evaluation of task allocation/Levels of Automation in manufacturing systems.

o

Presents a concept model dealing with both socio-technical and physical-cognitive issues (Fasth and Stahre, 2010)

This article will present the concept model developed by Fasth and Stahre in 2010. Furthermore, results from industrial case studies using the model will be presented. Lastly, discuss if the model is fulfilling Older’s et al. sixteen requirements (Older et al., 1997) (Waterson et al., 2002).

METHOD: A CONCEPT MODEL AND METHODOLOGY FOR TASK ALLOCATION The concept model was developed in steps, between the years 2007 and 2010, as shown in figure 1. In development of the concept model, the hypothesis was: “Could an assembly system be designed in a structured way with the most advantageous cognitive and physical level of automation to cope with future changes?” (Fasth, 2009). The methodology, DYNAMO++ (Fasth et al., 2008 -b, Fasth et al., 2010), was developed in collaboration with four companies. Furthermore, the methodology was validated in six companies (Fasth and Stahre, 2008). Validation was done to distinguish if the method could be generalised for use outside the first four companies. Further, if a novice user could understand and use the methodology. In these cases, the novice users where master students with supervision of an expert. This is in line with Waterson and Older (Older et al., 1997, Waterson et al., 2002) two approach requirements; 

Encourage participative use by various stakeholders, including the potential end-users of the system (method)



Be easy to learn, usable and require minimal training and support

Finally the methodology was verified in five companies before creating the concept model. DYNAMO++, contains of twelve steps divided into four phases; pre-study, measurement, analysis, and implementation. The methodology is built partly on other methods such as Value Stream Mapping (VSM) (Liker, 2004), Hierarchical Task Analysis (HTA) (Stanton et al., 2005, Annette and Duncan, 1967) and Levels of Automation (LoA)-taxonomy (Frohm et al., 2008) and partly new developed theory (explained in the following sections) and structure of the overall methodology. In line with Waterson et al. and Older et al. design requirements;  

Have a structured and systematic format (applies for the concept model as well) Be consistent with existing tools and techniques in use

  A further development of the methodology was made in 2010, resulting in a concept model, presented in figure 2. The aim of the concept model was to visualize relations between different areas and actions within a company when redesigning a system in terms of levels of automation through task allocation. Furthermore, to have a clear view why to change the system in order to avoid over- or under- automated systems. Additionally, the model also considers the cognitive LoA, the information system (here called the Level of Information - LoI) to and from the operators and different Levels of Competence (LoC) in the operator group.

Figure 2 Concept model, further developed from DYNAMO++ (Fasth and Stahre, 2010)

The following sections will explain the most important steps in the model, both theoretical and with industrial examples.

TRIGGERS FOR CHANGE - WHY CHANGE LEVELS OF AUTOMATION?

Porras and Robertsson (Porras and Robertsson, 1992) use a 2 by 2 matrix to describe changes in a structured way; Level of change (first and second degree) i.e. changes in the current system or redesigning the system and reason for change (external or internal demands). A majority of the fifteen case companies wanted to do minor changes (first degree) in the current system, although two of the companies wanted to redesign the assembly system (second degree). Three of the companies were conductors and had external quality demands (these demands were achieved but required a lot of extra internal work) Therefore, they wanted to improve in-house quality and First-Time-Thru (FTT) by do minor changes in the current system i.e. internal change, first degree. The other companies wanted to either decrease time (throughput and cycle time) or increase flexibility (product or volume). SOPI (LOAC;LOAP)- SQUARE OF POSSIBLE SOLUTIONS In 2008 Frohm (Frohm et al., 2008) proposed a taxonomy and definition for levels of automation used in manufacturing systems, i.e. “The allocation of physical and cognitive tasks between humans and technology, described as a continuum ranging from totally manual to totally automatic”.

The taxonomy is a seven-step reference scale, for cognitive and physical LoA aiming at quantifying tasks due to LoA. Frohm (ibid) defined physical tasks as the level of automation for mechanical activities, mechanical LoA, while the level of cognitive tasks is called information LoA. Mechanical LoA is WITH WHAT to assemble, while Cognitive LoA is HOW to assemble on the lower levels (1-3) and situation control on the higher level (4-7). A matrix integrating the two reference scales, as seen in figure 3, forms a 7x7 matrix, resulting in 49 possible types of solutions for task allocation, each including a physical LoA and a cognitive LoA. The figure also displays the division between human and machine assembling and monitoring the tasks.

Figure 3 Joint matrix of physical and cognitive LoA (Fasth et al., 2009 ) The matrix is used as a quantitative way of measuring the current LoA in the chosen areas´ tasks. The result is used for further analysis to meet triggers for change and also to make the company understand their mind set in a clearer and more objective way when it comes to automation. Step 7 in DYNAMO++, is a workshop where different levels of the involved employees in the chosen area participate, in the case studies it where operators, production engineer, logistics and production managers. Together they discuss the measured values and how this could be changed in order to fulfil their trigger for change. This reduces the number of possible solutions, resulting in a Square of Possible Improvements (SoPI) where companies can chose their appropriate solutions (perform Task allocation).

Figure 4 show a result from a case where the SoPI went from 49 possible solutions to 15, the company also thought that they wanted to increase the cognitive LoA i.e. change recourses in terms of technology (mobile assembly instructions).

SoPI= 3-5;(1-5)

 

 

 

 

Figure 4 Measured values of different tasks and suggested SoPI for future state analysis and discussion (Nordin et al., 2010)

In line with Waterson and Older´s types of allocation requirements; the matrix covers allocation to the humans and to the machine. To be able to cover the second requirement according types of allocation; Cover allocations between humans, and examine human role and the issue requirement; Examine the content and quality of human´s job, the concept model uses a mix between Sheridan’s five operator roles; Plan, teach (programming), perform, intervene, and learn (Sheridan, 1992) and the work tasks in the automatic assembly system. Together they form a 15-point list, used as a reference in the case studies in order to determine how many of the 15 roles the operators have fully, partly or non-control over in the present system. According to (Fasth et al., 2010), operators had the main responsibility for less than 20 % of these tasks. In most companies, operators have no participation in planning and maintenance (only in small disturbance handling). Partly involvement differs from 40 to 65 % between the five case companies. This result is used to understand the importance of the Level of Information (LoI) to and from the chosen area in the system and Level of Competence (LoC) in the operator group.

COMPARING CURRENT AND NEW CURRENT STATE - INDIRECT AND DIRECT MEASURABLE PARAMETERS

To determine if the goals have been achieved, there is a need to divide the triggers for change into indirect measurable parameters (PIDM) and direct measurable parameters (PDM). This is in line with Older´s issue requirement; specify decision criteria.

To determine a change in an assembly system, two types of parameters could be described; indirect and direct measurable parameters. The indirect measurable parameters (PIDM) could be described as qualitative parameters i.e. flexibility [7], complexity etc. The direct measurable parameters (PDM) could be described as quantitative parameters i.e. time parameters, number of products, number of tasks etc. These parameters will be related to the change of physical (LoAp) and cognitive (LoAc) LoA [7]. This could be described as: Company A wishes to increase its production volume flexibility (PIDM). This could be done by: 1. decreasing the throughput time (PDM) through the cell by increasing the physical LoA 2. increasing the competence (PIDM) among operators by learning more tasks (PDM) and by increasing the cognitive LoA THE CONCEPT MODEL PUT IN PRACTISE – INDUSTRIAL CASE STUDIES During 2007-2010, fifteen case studies using DYNAMO++ and the concept model have been carried out. Both quantitative and qualitative methods have been used to collect empirical data; measurement of LoA observations and semi structured interviews. The case question was whether the mechanical and cognitive level of automation should be changed in order to fulfil the industries' reasons for redesigning their assembly systems (their triggers for change). Furthermore, the case studies are used to show links between levels of automation and direct measurable parameters (PDM) and indirect measurable parameters (PIDM). The case studies have been executed in assembly operations. The chosen assembly system contained of 3-8 operations and approximately 3-15 operators depending on shifts etc. Four main reasons for redesigning the system were observed in the fifteen case studies; -

Two direct measurable parameters (PDM); 1) to increase quality, 2) to decrease timeparameters, and

-

Two indirect measurable parameters (PIDM); 1) to increase product and volume flexibility (Fasth et al., 2008 -a, Fasth and Stahre, 2008) and 2) to decrease subjective complexity (Fässberg et al., 2011).

MEASURING LEVELS OF AUTOMATION (LOA) IN THE CURRENT STATE

A current state analysis was made in the fifteen companies, using HTA and reference measurements of LoA. Measurement data from nine of the case studies is gathered in the LoA matrix, in figure 5. The tasks in each case in the table where measured in one or two stations in final assembly cells.

Figure 5 Observed tasks from nine case studies, documented in the LoA matrix Approximately 87% of the tasks (fig 5) in the cases are assembled and monitored by humans (LoA= 13;1-4), most of these tasks are done by hand, based on own experience (LoA= 1; 1). Examples of tasks on LoA level 1; 1 are complex or precise pick-and-place tasks (seen in picture 1, fig. 6).

  Figure 6 Example of different assembly tools used in the case studies An analysis based on measurement results was done in order to determine the case question of task allocation. Results showed that the cognitive automation should be increased (Fasth and Stahre, 2008) in almost all case studies, although not extensively. Company goals towards increasing customisation resulting in increasing information flow that requires new solutions for presenting information to the operators (Fässberg et al., 2010). The human should still be in control of the task, but not merely in a monitoring role. In the LoA /SoPI matrix this means moving from LoAcog= 1 towards LoAcog= 3-5. A majority of the companies did not prioritise increase in mechanical LoA after a DYNAMO++ analysis.

AN EXAMPLE OF TASK ALLOCATION – LINKING LOA WITH PDM AND PIDM Company D is an example of how companies can use LoA measurements to improve their triggers for change. According to Fasth et al (Fasth et al., 2008 -a), the company wanted to increase volume flexibility for a specific product family. A current state analysis was done, using HTA, VSM, and LoA measurement. To achieve volume flexibility the company integrated redundancy in the bottleneck station, as illustrated in figure 7. The matrix shows two different alternatives that could be used for assembling. The company used a manual station LoA= (3; 5) as a solution when the robot cell LoA= (6; 5) was down. In order to fulfil the incensement of volume flexibility, they could use the two stations to perform a task allocation depending on the order status on a daily basis.

Figure 7 Observed LoA and cycle time (C/T) value for the redundant station S3

The redundancy made it possible, not only to have a more robust system but to create routing flexibility and volume flexibility, thru four different alternatives to assemble the product i.e. task allocation. The amount of products produced with different alternatives is shown in figure 8.

Figure 8 Creating volume flexibility through task allocation

The robot cell is used as the main machine in the system. The productivity for one normal day is 24 products/hour, i.e. 192 products per shift. Assume that a breakdown for two hours is happened on the robot cell. Without the routing flexibility the loss will be 24 parts per hour = 48 products. With the routing flexibility the company is able to use the static work station under the reparation producing 18 products/hour i.e. a loss of 5 products per hour = 10 products. The four alternatives give a variance of 144 (min 152, max 296) number of products. This case shows how task allocation and changing recourses could be illustrated in an easy way for the company so that they could take well-founded decisions due to the volume they wanted to produce. The example covers four of Older´s requirement; 

Issues requirements; o

Trade-offs between decision criteria, in this case between the four alternatives (the more manual station and the robot cell).

o

Enable quantitative evaluations to be made of the alternative choices, in this case it is shown how the volume flexibility and number of products could be alternated by changing LoA.



Design requirement; Be cost effective and efficient to use, the matrix is an easy way to show and discuss the different alternatives



Types of allocation requirements; Incorporate the concept of dynamic allocations dependent on real time contingencies

DISCUSSION AND CONCLUSION The proposed method (DYNAMO++) for an analytical approach to Level of Automation analysis is the result of a broad review of reference scales addressing task allocation. The method emphasises tasks that can be performed by humans as well as by automation, thus embracing the fundamental ideas of Fitts (Fitts, 1951), Jordan (Jordan, 1963), and Sheridan (Sheridan, 1995). In this paper the authors propose a separation of cognitive and physical level of automation, while maintaining their relation in the LoA matrix. The paper therefore claims to demonstrate an advance in the state of the art in task allocation. The concept model and the DYNAMO++ method can be seen as a step closer towards quantitative measures of task allocation (i.e. changes in both physical and cognitive LoA) and dynamic changes over time. One step to examine if the concept model could be seen as a useful method is a comparison with Older’s and Waterman’s sixteen requirements. A summary of these requirements and how the concept model is fulfilling them is shown in appendix 1. Based on this summary and the examples shown above it is shown that the concept model could be used for task allocation in order to improve the system by letting the human man and the machine become equally possible solutions when it comes to task allocation, rather than a binary decision. Further, the concept model is partly fulfilling three of the sixteen requirements: 

Examine the whole system, as well as individual tasks and roles - The manufacturing complexity of increasing product customisation in combination with cost awareness has forced companies to abandon previous strategies for full assembly automation. Task allocation in industrial manufacturing is therefore becoming increasingly important, especially in assembly. The concept model is therefore used to improve the production system and cannot be seen as an organisational, top-down method but as a hands-on, bottom-up approach.



Be applicable to complex environments and different systems within the same environment – Because Older and Waterman developed their method for naval, i.e. control room, environments, this requirements is beyond the scope of the concept model but it would be interesting to investigate if the model could be useful also in these areas.



Be capable of use in new and existing systems – The model has only been used in order to improve existing systems and not to design a totally new system

Fifteen cases in real assembly system indicate that the concept model provides industrially relevant results and increases quality of advanced, semi-automated manufacturing system analysis. Further, the validation and verification show that the concept model is easy to use for the end-users and provides quantitative data for different solutions to be compared and analysed.

ACKNOWLEDGEMENT The authors want to express their deep gratitude to VINNOVA (The Swedish Governmental Agency for Innovation Systems) for funding this research. Moreover, special thanks to the fifteen case companies.

This work has been carried out within the Sustainable Production Initiative and the Production Area of Advance at Chalmers. The support is gratefully acknowledged. REFERENCES ALMSTRÖM, P. & KINNANDER, A. (2007) PRODUCTIVITY POTENTIAL ASSESSMENT OF THE SWEDISH MANUFACTURING INDUSTRY. Proceedings from the 1st Swedish Production Symposium. Gothenburg, Sweden. ALMSTRÖM, P. & KINNANDER, A. (2009) PRODUCTIVITY POTENTIAL ASSESSMENT OF 30 SUPPLIERS TO THE AUTOMOTIVE INDUSTRY. Proceedings from the 3rd Swedish Production Symposium. Stockholm, Sweden. ANNETTE, J. & DUNCAN, K. D. (1967) Task analysis and training design. Occupational Psychology, 41, 211-221. BELLGRAN, M. & SÄFSTEN, K. (2005) Produktionsutveckling - Utveckling och drift av produktionssystem, Lund, Sweden, Studentlitteratur. BRIGHT, J. (1958) Automation and Management, Boston, USA. BROOKS, R., ROBINSON, S. & LEWIS, C. (2001) Simulation-inventory control. Palgrave. DEKKER, S. & WOODS, D. (2002) MABA_MABA or ABRAKADABRA? Progress on HUmanAutomation Co-ordination. Cognition, Technology and Work, 4, 240-244. DUNCHEON, C. (2002) Product miniaturization requires automation - but with a strategy. Assembly Automation, 22, 16-20. ENDSLEY, M. & KABER, D. (1999) Level of automation effects on performance, situation awareness and workload in a dynamic control task. Ergonomics, 42, 462-492. ENDSLEY, M. & KIRIS, E. (1995) Out-of-the-loop performance problem and level of control in automation. HUMAN FACTORS - Human Factors and Ergonomics Society., 37, 381-394. ENDSLEY, M. R. (1997) Level of Automation: Integrating humans and automated systems. Proceedings of the 1997 41st Annual Meeting of the Human Factors and Ergonomics Society. Part 1 (of 2), Albuquerque, NM, USA. Santa Monica, CA, USA, Human Factors and Ergonomics Society, Inc. FASTH, Å. (2009) Measuring and Analysing Levels of Automation in assembly systems - For future proactive systems. Product and production Development, Production systems. Gothenburg, Chalmers University of Technology. FASTH, Å. (2011) Comparing methods for redesigning, measuring and analysing Production systems. Proceedings of the 4th Swedish Production Symposium (SPS). Lund, Sweden. FASTH, Å., BRUCH, J., DENCKER, K., STAHRE, J., MÅRTENSSON, L. & LUNDHOLM, T. (2010) Designing proactive assembly systems (ProAct) - Criteria and interaction between automation, information, and competence Asian International Journal of Science and Technology in production and manufacturing engineering (AIJSTPME), 2 (4), 1-13. FASTH, Å., LUNDHOLM, T., MÅRTENSSON, L., DENCKER, K., STAHRE, J. & BRUCH, J. (2009 ) Designing proactive assembly systems – Criteria and interaction between Automation, Information, and Competence. Proceedings of the 42nd CIRP conference on manufacturing systems Grenoble, France.

FASTH, Å. & STAHRE, J. (2008) Does Levels of Automation need to be changed in an assembly system? - A case study. Proceedings of the 2nd Swedish Production Symposium (SPS). Stockholm, Sweden. FASTH, Å. & STAHRE, J. (2010) Concept model towards optimising Levels of Automation (LoA) in assembly systems. Proceedings of the 3rd CIRP Conference on Assembly Technologies and Systems. Trondheim, Norway. FASTH, Å., STAHRE, J. & DENCKER, K. (2008 -a) Analysing changeability and time parameters due to levels of Automation in an assembly system. Proceedings of the 18th conference on Flexible Automation and Intelligent Manufacturing - FAIM. Skövde, Sweden. FASTH, Å., STAHRE, J. & DENCKER, K. (2008 -b) Measuring and analysing Levels of Automation in an assembly system. Proceedings of the 41st CIRP conference on manufacturing systems Tokyo, Japan. FITTS, P. (1951) Human engineering for an effective air navigation and traffic control system. Columbus,OH, Ohio state university. FROHM, J. (2008) Levels of Automation in production systems. Department of production system. Gothenburg, Chalmers University of technology. FROHM, J., LINDSTRÖM, V., WINROTH, M. & STAHRE, J. (2008) Levels of Automation in Manufacturing. Ergonomia IJE&HF, 30:3. FÄSSBERG, T., HARLIN, U., GARMER, K., GULLANDER, P., FASTH, Å., MATTSSON, S., DENCKER, K., DAVIDSSON, A. & STAHRE, J. (2011) AN EMPIRICAL STUDY TOWARDS A DEFINITION OF PRODUCTION COMPLEXITY. 21st International Conference on Production Research. Stuttgart, Germany. FÄSSBERG, T., NORDIN, G., FASTH, Å. & STAHRE, J. (2010) iPod Touch - an ICT tool for assembly operators in factories of the future? . Proceedings of the 43rd CIRP International Conference On Manufacturing Systems (ICMS). Vienna, Austria. GROOVER, M. P. (2001) Automation, production systems, and computer-integrated manufacturing, Upper Saddle River, N.J., Prentice Hall. HANCOCK, H. A. & CHIGNELL, M. H. (1992) Adaptive allocation by intellegent interfaces. HOLLNAGEL, E. (2003a) Handbook of Cognitive Task Design, Mahwah, New Jersey, Routledge. HOLLNAGEL, E. (2003b) The role of Automation in joint cognitive systems. IFAC. HOU, T., LIN, L. & DRURY, C. G. (1993) An emperical studyof hybrid inspection system and allocation of inspection functions. Internetional journal of human factors in manufacturing systems, 351367. JORDAN, N. (1963) Allocation of functions between human and machine in automted systems. Journal of applied psychology, 47, 161-165. KANTOWITZ, B. H. & SORKIN, R. D. (1987) Handbook of human factors. Ch 3.3 Allocation of functions, New York, Wiley. KERN & SCHUMANN (1985) Das Ende das Arbeitsteilung Verlag Beck. KOHO, M. (2010) Production System Assessment and Improvement - A tool for MTO and ATO companies. Production engineering. Tampere, Tampere university of technology. LIKER, J. K. (2004) The Toyota Way: 14 Management Principles from the World's Greatest Manufacturer p. 303, USA, McGraw-Hill. MARCH, R. & MANNARI, H. (1981) Technology and size as determinants of the organizational structure of japanese factories. Administrative science quarterly, 26, 33-57. NORDIN, G., FÄSSBERG, T., FASTH, Å. & STAHRE, J. (2010) iPod Touch - an ICT tool for assembly operators in factories of the future? - Technical solutions and requirements. 3rd CIRP Conference on Assembly Technologies and Systems (CATS). Trondheim, Norway. OLDER, M. T., WATERSON, P. E. & CLEGG, C. W. (1997) A critical assessment of task allocation methods and their applicability. Ergonomics, 40, 151-171.

PARASURAMAN, R., SHERIDAN, T. B. & WICKENS, C. D. (2000) A model for types and levels of human interaction with automation. IEEE transactions on system, man, and cybernetics - Part A: Systems and humans, 30, 286-296. PARASURAMAN, R. & WICKENS, C. D. (2008) Humans: Still Vital After All These Years of Automation. Golden anniversity special issue of Human Factors, 50, 511-520. PORRAS, J. I. & ROBERTSSON, P. J. (1992) Organizational Development: Theory, Practice and Research in Handbook of industrial and organisational psycology, California, Consulting Psychologist Press inc. PRINCE, H. (1985) The allocation of function in systems. Human Factors, 27, 33-45. SHERIDAN, T. B. (1992) Telerobotics, automation and human supervisory control, Cambridge Massachussetts, MIT Press. SHERIDAN, T. B. (1995) Human centred automation: oxymoron or common sense? Intelligent Systems for the 21st Century, IEEE international. Proceedings of: Systems,Man and Cybernetics. SHERIDAN, T. B. (2000) Function allocation: algorithm, alchemy or apostasy? International Journal of Human-Computer Studies, 52, 203-216. STANTON, N. A., SALMON, P. M., WALKER, G. H., BABER, C. & JENKINS, D. P. (2005) Human Factors Methods - A Practical Guide for Engineering and Design, Ashgate. SÄFSTEN, K. & ARESU, E. (2000) Vad är bra monteringssystem?: En studie av utvärdering och utformning på 15 industriföretag i Sverige. Linköping, Linköpings universitet. TAYLOR, F. W. (1911) Principles of Scientific Management. TRIST, E. L. & BAMFORTH, K. W. (1951) Some Social and Psychological Consequences of the Longwall Method of Coalgetting. Human Relations, 4, 35. WATERSON, P. E., GRAY, M. T. O. & CLEGG, C. W. (2002) A sociotechnical method for designing work systems. Human Factors 44, 376.

Appendix 1 A comparison of Older’s and Waterson’s requirements and the concept model Category Types of allocation

Issues

Requirements [Older et. Al, 1997; Waterson, 2002] (IV)Cover allocations to the humans and to the machines

Concept model [Fasth, 2010] H;M;H&M

LoA matrix

(V) Cover allocations between humans, and examine different human roles

LoC

(XI) Incorporate the concept of dynamic allocations dependent on real time contingencies

Shown in the industrial Ex.

(VIII) Examine the content and quality of the human´s job (VI) Specify decision criteria

LoC, LoI Direct measurable parameters Shown in the industrial Ex.

(VII) Consider the trade-offs between the decision criteria

Approach

Comment

(X) Enable quantitative evaluations to be made of the alternative choices

Shown in the industrial Ex.

(XII) Encourage participative use by various stakeholders, including the potential end users of the system (method) (IX) Enable the users of the method to make informed choices

Developed, validated and verified together with end-users Get a common language when discussing LoA, LoA matrix

(I) Be used early in the design process

Have been used when redesigning systems

(III) Allow iterative use, throughout the design process

Get a common language when discussing LoA, LoA matrix Developed, validated and verified together with end-users

(XII,XIII) Be easy to learn, usable, and require minimal training and support Examine the whole system, as well as individual tasks and roles

Partly

Mostly focuses on individual tasks and roles within the system

(XIV) Be applicable to complex environments and different systems within the same environment (XV) Be adaptable (to different situations and tailored for unique applications)

Partly

Mostly tested in assembly systems The companies Triggers For Change decides the rest of the design process

Be capable of use in new and existing systems

Partly

Mostly focus on existing systems

Coverage

Cover the rationale for its use (II) Have a structured and systematic format

Twelve steps and the main loop

(XVIII) Be cost effective and easy to use

Validated by novice users Using both new (LoAmatrix) and existing methods (HTA, VSM etc.)

Design

(XIX) Be consistent with existing tools and techniques in use

PAPER 2 Fasth, Å. Stahre, J. and Dencker, K. (2008) Measuring and analyzing Levels of Automation in an assembly system, Proceedings of the 41st CIRP International Conference on Manufacturing Systems (ICMS), Tokyo, Japan

Å. Fasth, J. Stahre, K. Dencker

Measuring and analysing Levels of Automation in an assembly system 1

1

Åsa Fasth Johan Stahre and Kerstin Dencker

2, 3

1

Department of Product and Production Development, Division of Production system, Chalmers University of Technology, Sweden 2

Royal Institute of Technology, Department of Production engineering 3

Swerea IVF, Industrial Research & Development Corp.

Abstract The level of automation employed in semi-automated assembly systems is crucial, both to system performance and cost. This paper presents a methodology to enable selection of the right Level of Automation. The method thoroughly maps existing product and information flows as well as the automation level in separate parts of the system. It then analyses and identifies future automation possibilities, i.e. the automation potential seen from an industrial perspective. Further development of the method is based on validations and industrial case studies.

Keywords: Levels of Automation, Assembly system, DYNAMO

1

2

INTRODUCTION

THE DYNAMO METHODOLOGY

A method was developed in the DYNAMO project (20042007) [6] in association with five companies. The aim with this project was to develop a methodology for measuring and get an accurate picture of today’s information flow and automation level in production systems. Furthermore to develop a reference scale for different Levels of Automation (LoA) that could be used in the manufacturing area [7], this is shown in table 1.

Products of today are becoming increasingly customized. Smaller batches and decreasing time limits for set-ups of new products are some of the resulting demands on the assembly systems, due to the increasing number of variants in the assembly flow. Consequently, assembly systems have to get the right things, to the right place, at the right time, in the right quantity to achieve perfect work flow while minimizing waste and being flexible and able to change [1]. To achieve this, the companies can adopt automated solutions, when doing this there is a need to determine the correct amount of automation. It is also necessary to identify the optimal parts of the valueflow to be automated. In automation decisions it is necessary to consider human resources, as well as mechanical technology and information flow. By definition, automation is a technology by which a process or procedure is accomplished without human assistance [2]. Unfortunately, automation does not always fulfil expectations; the need for human intervention in cases of disturbances and system failures is still high. Smart automation is defined by [3] as the human aspect of 'autonomation' whereby automation is achieved with a human touch. However, there is a tendency among industry to consider automation investments as a” black or white” decision. This may be suboptimal, since there is not always a need to distinctly choose between humans or machines. The interaction and task division between the human and the machine should instead be viewed as a changeable factor which can be called the level of automation [4]. Thus, identifying and implementing the right level of automation in a controlled way could be a way to maintain the effectiveness of a system[5]. In this paper we will discuss a methodology that could be used to measure today’s automation level and that analysis the level of automation that is possible in the future. This could help companies to choose the “right” level of automation due to their production, requirements and demands. This gives great benefits when it comes to time and cost savings in the planning and implementation phase.

Levels 1 2 3 4 5 6 7

Table I Levels of Automation Mechanical Information Totally manual Totally manual Static Hand tool Decision giving Flexible hand tool Teaching Automatic hand tool Questioning Static work station Supervising Flexible workstation Interventional Totally automatic Totally automatic

The concept Levels of Automation was defined as; “The allocation of physical and cognitive tasks between humans and technology, described as a continuum ranging from totally manual to totally automatic” By physical support, [7], mean the level of automation for mechanical activities, mechanical LoA while the level of cognitive activities is called information LoA. Due to [8] the conclusion is that most tasks in manufacturing often involves a mix of both mechanised and computerised tasks and the companies has to consider both areas when automating their system. 2.1 Validation of the methodology The DYNAMO methodology to measure the Levels of automation consists of eight steps [9], seen in figure 2. These steps were validated at an industrial company which has not been participating in the development of the

The 41st CIRP Conference on Manufacturing Systems, 2008

169

Å.Fasth, J.Stahre and K.Dencker

methodology [9]. The validation group contained of four people; two who developed the method and two that began looking at the methodology in 2006 as part of the ProAct project [10]. The group validated each step, except step 8. This step was not validated because the company where the validation took place did not have any purpose regarding LoA strategy, [9].

Modification of the DYNAMO method

1, Choose system (On-site)

3

Method for analysing the Levels of Automation

3, VSM (Value Stream Mapping) to identify the flow and time parameters

4, Identify the main operations and subtasks. Design a HTA for the chosen area 5, Measure LoA (Information, Mech; Task and Transport)

Measurement

The methodology for analysing levels of automation is done as one step in the ProAct [11] research project. This is done to be able to generate possible improvements for the present assembly system. The skeleton of the methodology [7] is intact. The further development is based on the validation that was done in the DYNAMO project and six case studies that have been done mainly within the ProAct project, throughout the period May-Nov, 2007 [12, 13]. The case studies were performed in SME companies. The modification of the method contains of four different subgroups or phases, as visualised in figure 1.

Pre-study

2, Walk the process

6, Document the results Pre study 7, Workshop (Ws) to decide the relevant Min- and Max levels for the different tasks in the system

Measurement Analysis Implementation

Analysis

8, Design of the Square of Possible Improvements (SoPI) based on the Ws 9, Analysis of the Square of Possible Improvements (SoPI); task and operation opt. due to the time and flow parameters

Figure 1 Phases in the measurement methodology These phases each contains of three steps, shown in figure 2. Results from the modification are; • Video or tape documentation has been used in all case studies to easier analyse the assembly system. • “Lean awareness” in the measured companies is seen as necessary (4 of 6 companies had this) to be able to perform and understand the usage of step 3-10 (modification steps). • A simplification of the equation has been developed in order to decrease the time in step 4-9 (modification steps), and to decrease the subjective assessment in the work shop. • No information were given out before the people who were doing the measurement were in place at the companies, i.e. off-site measurement could not been done in any of the case studies. • Doing a Value Stream Mapping (VSM) to gather information about the time parameters, the informationand material flow in the system to get a deeper analysis of today’s system. • Measure LoA based on value adding and not value adding tasks in the system. • Logic has been developed to simplify future simulation of levels of automation in an assembly system. • Consideration also has to be taken about the operators’ competence and education about future changes.

10, Write and/or visualise some suggestions of improvements based on the SoPI analysis and the company’s wishes and demands. Implementation

11, Implementation of the chosen suggestions 12, Follow-up when the suggestions have been implemented to see what effects the suggestions have had on time end flow parameters

Figure 2 Method for measuring and analysing Levels of Automation

170

Measuring and analysing Levels of Automation in an assembly system

To be able to perform an operation optimisation there is one condition, all the SoPItask has to be represented in the SoPIoperation in order to make an optimisation, if not, one solution is to do an optimisation with some of the tasks and do a task optimisation on the others. It could be described as;

The most important development is the implementation of measurable parameters, time parameters, within the methodology. The logic ha also been very important to develop considering future simulation and visualising of the result in the analysis phase, see subsequent sections. The analysis and implementation phase is completely new and the sections below will describe these phases.

IFF SoPIoperation ҧ task=1n SoPItask THEN operation optimisation is possible

3.1 THE MEASUREMENT PHASE The reference scale, seen in table 1 has been developed into a matrix. This is done to get a logical ground and to be able to add dimensions or parameters to the methodology. This matrix is used to visualise the different levels of automation. It is also used in the analysis phase to show the results of the measurements and the suggestions of possible improvements. The matrix is called LoAtotal and contains of the vectors LoAmech and LoAinfo . The logic for the matrix is seen in equation 1;

Equation 3 Operation optimisation The next step in the analysis phase is to evaluate what the effects are when choosing different solutions in the SoPItask or SoPIoperation depending on the goal with the measurement. 4

Discussion

From the six case studies, we can declaim that the SoPI is being limited by two obvious things; • The persons who are leading the work shop - If the leaders of the work shop do not have the skill or knowledge to ask open questions to widen the companies view about the future automation possibilities. It is also important that these persons have a deep knowledge about the methodology and some experience from the industry.

1  LoAtotal  49 LoAtotal  (LoAmech) ш (LoAinfo) WHERE LoAmech (y) = 1 y  7 and LoAinfo (x) =1 x  7 Equation 1 the matrix of LoAtotal This means that there are 49 possible solutions that could exist or be developed in the assembly system. It also means that a measured task has to contain both a mechanical and an information part otherwise the structure of the hierarchical task analysis (HTA) is too deep.

•

The companies thoughts and knowledge about future automation - If the companies refrain from “thinking outside the box”

This could result in a smaller SoPi which limit both task and operation optimisations. This has resulted in further development of the assessment in the work shop. This is done to decrease the subjective assessment. Furthermore to be able to simulate and visualise possible improvements based on the companies’ demands and needs rather than their thoughts. Future research efforts wiil be to investigate how the different solutions in the matrix are related to each other in terms of different time parameters [14]. Furthermore an investigation will be carried out to see if and how it is possible to change level of automation in real time and over time in an assembly system. Deeper knowledge will be used to improve modelling and simulation tools for different levels of automation; this is a part of a research project called SIMTER. Future research aims at simulating and visualising assembly systems with varying LoA in the system’s stations and tasks. Improvement of flexibility and the flow-and time parameters will be measured; this will be done to be able to develop a proactive assembly system in the project called ProAct.

3.2 THE ANALYSIS PHASE; Step 8 and 9 These steps are done after the work shop to analyse today’s assembly system and to map the possible improvements in the LoAtotal matrix. This is done with help of the relevant min and max value. These values form a span where the company could move within when it comes to a development of the companies’ assembly system; this span forms a square called Square of possible improvements (SoPI). The SoPI sets boundaries for the company’s future improvements in automation solutions seen from their demands. This is done to make it easier to analyse the effects when changing/ varying the LoA and also to see if it is possible to make task and operation optimisations within the measured system. Two different SoPI: s could be designed; task optimisation and operation optimisation. The first step is to design SoPIs´ for all the tasks in the operations. The logic for the SoPI and SoPItask is described in equation 2:

5

SoPI  (LoAmech (min; max)) ш (LoAinfo (min; max))

SUMMARY

This paper has presented a methodology for measurement of the level of automation in assembly systems. While based on the DYNAMO methodology, a development of the measurement- and analysis step has been shown to provide a good visualisation of the present assembly system. The resulting automation matrix provides 49 different solutions that can be compared and analysed. The Square of Possible Improvements (SoPI) shows the span within the matrix where company personal believe their systems could be improved. The improvement potential is seen from different perspectives described by parameters, resources and demands.

SoPI = LoAmech (min; max) * LoAinfo (min; max) WHERE

LoAmech (x) = 1  min  max  7 ш LoAinfo (y) = 1  min < max  7

SoPItask  LoAtotal SoPItask ҧ LoAtotal Equation 2 The Square of Possible Improvements (SoPI) and task optimisation

171

Å.Fasth, J.Stahre and K.Dencker

The development of extended method logic and the addition of the time dimension to the existing LoA reference scales will provide opportunities to easier simulate different solutions for assembly systems. Furthermore, it will provide measurable values that could be analysed in today’s assembly system in view future systems. This will give the companies a solid base for decision making in planning and implementation phases of developing their future assembly system. 6

[13]

[14]

ACKNOWLEDGEMENTS

The authors would like to express their deep gratitude to researchers and industries that participated in the three projects DYNAMO, SIMTER, and ProAct. This research was financially supported by the Swedish Foundation for Strategic Research (SSF) and also by the Swedish Governmental Agency for Innovation Systems (VINNOVA). 7 [1]

[2]

[3] [4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

REFERENCES J. K. Liker, The Toyota Way: 14 Management Principles from the World's Greatest Manufacturer. USA: McGraw-Hill, 2004. M. P. Groover, Automation, production systems, and computer-integrated manufacturing, 2. ed. ed. Upper Saddle River, N.J.: Prentice Hall, 2001. T. Ohno, Toyota Production System: Productivity Press, 1988. R. Parasuraman, T. B. Sheridan, and C. D. Wickens, "A model for types and levels of human interaction with automation," IEEE transactions on system, man, and cybernetics - Part A: Systems and humans, vol. 30, pp. 286-296, 2000. M. Bellgran and K. Säfsten, Produktionsutveckling Utveckling och drift av produktionssystem. Lund, Sweden: Studentlitteratur, 2005. J. Frohm, V. Lindström, and M. Bellgran, "A model for parallel levels of automation within manufacturing," in the 18th International Conference on Production Research, Fisciano, Italy, 2005. J. Frohm, "Levels of Automation in production systems," in Department of production system. vol. Doctorial Gothenburg: Chalmers University of technology, 2008. J. Frohm, V. Lindström, M. Winroth, and J. Stahre, "Levels of Automation in Manufacturing," Submitted for publication in Ergonomica. V. Granell, J. Frohm, J. Bruch, and K. Dencker, "Validation of the DYNAMO methodology for measuring and assessing Levels of Automation," in Swedish Production Symposium Gothenburg, 2007. K. Dencker, J. Stahre, P. Gröndahl, L. Mårtensson, T. Lundholm, J. Bruch, and C. Johansson, "PROACTIVE ASSEMBLY SYSTEMS- Relaizing the potential of human collaboration with automation," in IFAC-CEA, cost effective automation in networked product development and manufacturing Monterey, Mexico, 2007. K. Dencker, J. Stahre, P. Gröndahl, L. Mårtensson, T. Lundholm, and C. Johansson, "An Approach to Proactive Assembly Systems - Towards competitive assembly systems," in IEEE international symposium on assembly and manufacturing (ISAM), University of Michigan, Ann Arbor, USA.ISAM, 2007. K. Jonsson, "Flow analysis and simulation of new and old production plant at Eka Chemicals AB," in Dep. of product and production development. vol.

172

Master Gothenburg: Chalmers University of Technology, 2007, p. 43. P. Ottosson and G. Grundström, "Flow analysis and simulation of the cooling module production at NIBE AB," in Product and production development. vol. Master Gothenburg: Chalmers University of Technology 2008 to be submitted. Å. Fasth, J. Stahre, and K. Dencker, "Analysing changeability and time parameters due to levels of Automation in an assembly system," in FAIM Skövde, 2008 to be submitted.

PAPER 3 Fasth, Å. and Stahre, J., Does Levels of Automation need to be changed in an assembly system? - A case study, Proceedings of the 2nd Swedish Production Symposium (SPS), Stockholm, Sweden, 2008

Does Level of Automation need to be changed in an assembly system? - A case study Åsa Fasth* and Johan Stahre Chalmers University of Technology, Division of Production system, Gothenburg, Sweden [email protected] Abstract: Production of to day is getting more and more competitive and companies have to be on top in their area in order to survive. This paper discuss if Levels of Automation need to be changed in assembly systems in order to achieve companies goals when it comes to flexibility and time minimisation. The empirical data is gathered through case studies at six different companies.

Keywords: Levels of Automation, Flexibility, Assembly system

1 Introduction: Production companies are constantly exposed to demands and requirements, both internal and external,that trigger changing plans for their different production areas. Trigger examples are volumes increases, new product introductions, decreases in lead times ,improved visualisation and flow [1], etc. To be able to handle such triggers and change from a current stage to a future stage as optimal as possible, production logistics has to be high priority at companies of today. Its purpose is to ensure that each machine and workstation assembles the right product in the right quantity and quality at the right time [2]. Complexity, robustness and flexibility are three areas that have been identified by industry as important in order to decrease time and cost parameters [3]. This paper will focus on the areas flexibility, time and levels of automation. Flexibility is above all other measures of manufacturing performance, cited as a solution [4]. More flexibility in manufacturing operations means more ability to adapt to customer needs, respond to competitive pressures, and to be closer to the market. [4]. The types of flexibility discussed in this paper are defined as; • • •

Volume flexibility – The ability to handle a change in volume for a special unit [5]. Routing flexibility – The ability to continue manufacturing a product in spite of a tool breakdown [6]. Production flexibility – The ability to produce a multitude of products and handle changes in the production planning [7].

Developing rapid, dynamic and responsive manufacturing processes and systems is a core area of manufacturing system research. One powerful approach to achieve this is to create a more flexible and agile workforce in a production area [8]. Agile manufacturing is not simply concerned with being flexible and responsive to current demands. It also requires an adaptive

capability to be able to respond to future changes [9]. To achieve lead-time reduction or a time optimisation, Just-in-Time tools and philosophies from Toyotas Production System could be used [911]. The tools help to decrease the lead-time through the elimination of waste such as over production, wasted time, wasted operation motions, inventory and production of defect parts [10]. Time parameters such as through-put time is very important to focus on, reduce time and cost savings will come [12]. Companies has to adopt these tools and have some Lean awareness to be able to achieve the best result in time reduction [13]. The assembly system needs to have the “right” levels of automation i.e. an optimal mix between human and technology for each task and operation in the system. In 1958Daniels [14]tried to predict the future manufacturing need for automation , by increasing value in use of in-line conveyorised assembly system and semi-automatic bench- mounted machines. Furthermore, Dashchenko et al. proclaimed in 1995 [15] that the main features of a Factory of the Future are: high level of flexibility of technological processes and equipment, high degree of process automation, high productivity, and high quality of manufacturing products. Womak et al. stated in 1990 that “by the end of the 90s we expect that team assembly plants will be populated almost entirely by highly skilled problem solvers whose task will be to think continuously of way and means to make the system run more smoothly and productively.”, p. 102 [16] While Ohno [17] declared that "smart automation, is automation achieved with a human touch". In Fits' classical list of task allocation presented in 1951 [18] he showed which task machines and humans do best. However, the list is controversial and it is still debated what it means and how to make the task allocation [19]. In various times and context Human Centered Automation (HCA) is purported to mean: Allocating to the human the task best suited to the human, allocating to the automation the tasks best suited to it [19].

Frohm [20] defined levels of automation as: “The allocation of physical and cognitive tasks between humans and technology, described as a continuum ranging from totally manual to totally automatic” If the companies do not consider these three areas; Flexibility, Time, and Levels of Automation when changing the assembly system there is a risk for over or under automation, inflexible systems and operators that are over- or under-stimulated. Thus, our hypothesis can be stated as: Time parameters and Flexibility = f (LoA) To find correlation within the hypotheses, a method for analysing levels of automation [21] is used in six case studies. This paper will discuss and analyse if the companies in the case studies need to change LoA in order to achieve their goals in terms of triggers for change.

2 Method for analysing Levels of Automation The concept Levels of Automation (LoA) was described by Sheridan and Verplank in 1978 [22]. Their research was mostly focused on the areas teleoperation, telerobotics and supervisory control, in order to make humans work through machines within hazardous environment and control complex systems such as aircraft and nuclear power plants [23] e.g. primarily cognitive automation contexts. The manufacturing context consists of a mix between both mechanised (physical) and computerised (cognitive) tasks, both of these have to be measured in order to get a clear picture of the manufacturing system. An advantage of using the two LoA reference scales proposed by [20], shown in table 1, is that the levels of both mechanical and cognitive support can be assessed in the same taxonomy. Table 1 Levels of Automation Levels Mechanical 7 Totally automatic 6 Flexible workstation 5 Static work station 4 Automated hand tool 3 Flexible hand tool 2 Static Hand tool 1 Totally manual

Information Totally automatic Intervene Supervising Questioning Teaching Decision giving Totally manual

Mechanical LoA is level of automation for Physical support or mechanical activities, while the information LoA is the level of cognitive activities. But, do companies in general comprehend the term "levels of automation" and do they use it when designing or redesigning their assembly systems? Results from a Delphi study in 2005 [24] show that Swedish manufacturing companies are not acquainted with this term. A methodology called DYNAMO that was developed from 2004 to 2007 [20]. The aim of this method was to help companies to measure assessing LoA in order to find appropriate level of span of automation and, by that, maintain high productivity by reducing production disturbances [20]. The methodology was later validated in industry [25]. The validation group consisted of four people; two who developed the method and two that began looking at the methodology in 2006 as part of the ProAct project [26]. The group validated each step, except step 8 (analysing step) [25]. A further development with focus on the analysis step was done in 2008, DYNAMO++ [21]. The aim was to analyse whether the current systems’ LoA is too high, too low, or to static in order to fulfil the companies' triggers for change. The methodology, DYNAMO++ [21], consists of four different phases, seen in figure 1; • Pre-study • Measurement • Analysis • Implementation Each of these phases contains three steps, i.e. the methodology contains a total of twelve steps [21]. The first two phases are carried out in the current system, to get an accurate picture of today’s automation level, production-, information-, materialand resourse flow. The Level of Automation (LoA) is measured in each task within the operation with help of the reference scale. The final two phases are used as a step towards the future state and as an input for future improvements in the assembly system. The companies´ Triggers for Change (TfC), e.g. demands and requests, as well as the two first phases is used as an input for the two last. This paper focuses on the first three phases and the companies’ triggers for change.

Measuring and analysing Levels of Automation, DYNAMO++ PHASES Analysis Implementation - Step 10

PHASES Pre study Measurement

PHASE Implementation - Step 11

PHASE Implementation - Step 12

Future state Current state Triggers for changing the current assembly system: internal or external

Figure 1 Measuring and analysing Levels of Automation

New Current state

3 Research methodology The results presented in this paper are combinations of both inductive and deductive approaches. Theory (Production area with paradigms, performers , state of the art and already existing method, DYNAMO [20]) → Empirical data (case studies, interviews and observation to get new angles and understanding of the current stage and problems in the companies) → Theory ( formulation hypotheses, development and progression of the old method based on theory, own experience and the collected data, the DYNAMO++ methodology [21]) → Empirical data (validation of hypotheses and the developed methodology in terms of case studies, observations and semi structured interviews [27]) → Theory 4 Case studies The aim with the case studies is to analyse the companies’ current and future state with figure 1 as base to answer the questions; How does the current

stage look like? What are the triggers that make the companies change the system? What do the companies see as possible improvements that could be made in a near future? Furthermore to be able to prove the hypotheses, does time and flexibility affect change in LoA? The companies that have been analysed and measured in the case studies are six companies in different production areas, seen in table 2. 1.1. Pre-study phase In the Pre-study phase, data was gathered about flow, type of assembling and number of products in the measured area. Flow and time parameters were also documented. A measurement of the current stage’s Level of Automation was carried out; the value is based on the automation level that the operator used to perform the task. The information was gathered with help of observations and interviews. The result of the pre-study phase is illustrated in table 2. .

Table 2 Companies participating in the case studies; current stage Current stage Production area Type of flow Type of Assembling Type Assembling

Company A Engine parts U-cell ATO

Company B Chemistry [28] Line ATS

Company C Electronics

Batch

Batch

Number of 2 main Products 30 variants Number of Stations 4 1 Average LoAinformation (Used) 1 Average LoAmechanical (Used) ATO – Assemble-To-Order ATS – Assemble-To-Stock

Company D Cooling modules [29] Job Shop ATS

Company E Trucks

Company F Vessels [30]

Line ATO

U-cell ATO

Accord based

One piece flow

Batch

2 main

One piece flow 2 main

3 main

9 5

5 3

8 1

3 main Costume made 5 3

4 main 8 variants 9 1

5

5

1

-

1

U-cell ATO

1.2. Triggers of Change All the companies had conferred about their triggers of change that they wanted to investigate further in terms of investments and flow analysis to be able to meet the internal and external demands. A majority of the companies wanted to either increase the flexibility or decrease time parameters.

Investigations about the companies’ Lean awareness were also done before the analysis and results were presented. This was done to investigate if companies work on improvements in terms of waste reductions, machine layout, and visualisation etc before changing the level of automation. The result is seen in table 3.

Table 3 Triggers for change and Lean awareness Triggers for change

Company A Increase quality (Increase Cognitive LoA)

Company B Decrease throughput time

Company C Volume and product flexibility

Company D Wants to buy a robot (increase mechanical LoA)

Company E Simplify the information flow to the operators

Lean Awareness Middle None High None Middle (use of JIT tools [11]) High – The message had reached the operators and almost all the tools were implemented Middle – Started with the early-on tools [11], the implementation had stopped at the white-collar worker level None – have almost not heard of Lean Production

Company F Increase volume and product flexibility, visualise the flow Middle

1.3. Analysis phase In the analysis phase the triggers of change and the current situation were input. As one of the outputs some suggestions were presented to the companies, seen in table 4. To be able to determinate if tasks or entire operations in the assembly system needs to and can change LoA, and in what span of automation the companies should start investigating possible solutions or improvements, an analysis phase was developed. This phase contains of three different steps, illustrated in figure 2.

Figure 2 the steps of the analysis phase In the following sections a brief explanation of each step will follow, with case study A as an example of the different steps; Step 7 Work shop to decide Min and Max values The result and gathered information from earlier phases is used in this step in order to get an accurate picture of how the current stage of the assembly system looks and so that the people present in the work shop can discuss if they think that this gives a true picture. The measured value (marked M in figure 3) is based on observation of operators with different experience performing the tasks. Semi-structured interviews were also performed with people involved in the investigated area. The workshop starts with a short briefing on the earlier steps and then if the companies has some triggers of change to redesign their system, and if, what level of automation do they think is reasonable to have in the new system. In case study A the trigger of change were; need to decrease the redoing in the cell e.g. increase the quality of the products and to increase the information to operators on how to assemble so that they did right the first time (cognitive LoA).

demands on the future system. A result from one of the work shops is shown in figure 3. Step 8 Design a Square of Possible Improvements (SoPI) The result from the workshop is then transformed into the LoA matrix to illustrate and to be able to analyse the results. The min and max values form the boundaries for the Square of Possible Improvements (SoPI), shown in figure 4.

Figure 4 result from work shop illustrated in the LoA matrix Step 9 Analyse the SoPI The SoPI is the used in order to analyse if it is possible to do a task and/or an operation optimisation. The result from case study A was, as shown in figure 5, in need to increase the cognitive level of automation in almost all tasks, also a need to increase the mechanical LoA in some tasks.

Figure 5 Task and operation optimisation The result was that there were 18 possible improvements for task optimisation and 6 possible solutions if the whole operation should be improved. The companies then has to investigate what consequences this optimisation result in; in terms of achieving the future goal with this redesign, investments in terms of facilities, information, competence and resources? 5 Result The result from the case studies shows that it is not always the Level of Automation that has to be changed; some of the suggestions have to do with other issues, e.g. production logistics.

Figure 3 Result from Ws in case study A The minimum value is the value that they need to have in order to achieve good quality and reasonable working conditions. The maximum level is a look into the crystal bowl in terms of new technology (for the company) different system methodologies, product flows etc in order to fulfil the

Company A wanted to improve quality and decrease re-assembly rate, the suggestion for the future stage was to increase the information LoA in terms of digitalised and forced assembly instructions where the operators could choose the level of information shown, due to their competence and experience but at the same time have check-point that the operator had to trigger – this might increase the cycle-time but decrease the overall time

because it decrease the re-assembling. Design For Assembling (DFA) [31] was also a suggestion. The current products had many similar parts that could be squeezed into the wrong product, by designing them so that only the correct product fitted could also decrease the re-assembling. Company B had too much buffer capacity because they produced large batches. The company also had too many breakdowns, one reason is because the operators did not always see if the machine had stopped, furthermore the follow-up of the breakdowns did not exist. The company also had a lot of “homemade automation” which means that the machines was not optimised and worked poorly. This resulted in operators trying to fix what the machines did wrong and this decreased the number of operators needed in the system, decreased workload for the operators in terms of heavy lifting and dangerous material handling. The suggestions here were; o To increase the information LoA in terms of state lamps on the machines; this will increase the awareness of the operators when something abnormal is happening. o Better follow-up on the breakdowns and continuous maintenance so that the company knows what is wrong with the machines and fix the problems. This will decrease the MDT (Mean Down Time) and hopefully the MTBF (Mean Time Between Failures) in the future. Another suggestion was to redesign the whole system with better solutions in automation then today and educate the operators in machine control and continuous maintenance. Company C wanted to increase the volume and product flexibility if possible. The suggestion was to build an assembly system that could vary the LoA in terms of for example line replacement, redundancy or plug- and-produce [32], the system could also be constructed as module structured assembly system [33]. The operators were not involved in the maintenance of the machines which increased the MDT when machines stopped, one suggestion where to improve the competence of the operator to handle small problems. Company D wanted to invest in a more automated cell for the last assembly task to decrease the through-put time. After the three first phases in the DYNAMO methodology the outcome were no common possible improvements due to the LoA analysis. The solution was to start improve the flow and production logistics in the current stage to achieve the goal (decrease throughput time). The suggestions were presented in an automation stair [29] where increased level of automation where done stepwise. The first step was to improve the material handling to the assembly stations because it was a lot of waiting time for the operators on articles that were not in the buffer and the operators were expensive because they did specialised jobs. Furthermore to do some kind of kitting [34] or FIFO for these articles so that the operators could weld instead of doing the material handling.

They also had a push system and Assemble-ToStock (ATS) so the buffers were quiet big. Suggestions for this were to improve the production planning to be able to Assemble-To-Order (ATO) instead and move the ordering handling downstream to create a pull system [10]. The products where heavy (40 kg) and these were pushed on a non mechanical transport band, a suggestion where to increase the mechanical LoA for the transportation and also to redesign the assembly cells from a single station line to U-cells. The last suggestion was to increase the Lean awareness to at least middle level so that the whitecollar workers understood the importance of production logistic and reduction of waste or non value adding tasks in the system. Company E wanted to decrease the number of paper assembly instructions (today it is 40 000 paper a day printed out). Some of the reasons where • up-dates took a long time to reach the operators • it was too much information that was hard to understand on each assembly instruction the company had to print the assembly • instruction long beforehand and it cost a lot of both money and environment cost in terms of paper (forest) to print all these papers The suggestion was to increase the information LoA in terms of digitalise the assembly instructions an let the operators choose the information that he or she needed due to their competence and experience. Company F wanted to improve the product and volume flexibility. Today the assembly system is islands of assembly stations and it is hard to follow the product flow. Furthermore a lot of the assembly tasks are made by hand. If transportation lines and assembly stations got more structured and marked the visualisation could increase. This will result in a more manageable material and production flow. Furthermore to increase the mechanical LoA in term of a number of line based transportation and standardised module based assembly stations. This results in decreased set-up time between products, and to be able to assembly more then one product in each line e.g. product and volume flexibility. This also makes it easier to vary the mechanical LoA in the future to achieve higher volume flexibility. A summary of the suggested solutions is illustrated in gray in table 4. 6 Discussion Results from the case studies shows that companies with low lean awareness think of mechanical LoA when they want to increase automation. Companies with high mechanical LoA often forget to improve the cognitive automation. This result in long cycle times, hard for new operators to learn the assembly tasks, quality problems and longer down times.

So does Levels of Automation need to be changed in an assembly system? The result from the case studies shows that the companies often need to change either the mechanical or information LoA to achieve their trigger for change. One common proposal for company A, B, E were to increase the LoA information in order to get higher quality, decrease MDT and save time and money when digitalise the assembly instructions. Suggestion for company C, D, F were to increase the LoA mechanical in order to minimize the transport time between the assembly stations but

also to be able to vary LoA to achieve volume and product flexibility. Furthermore four of the six companies also got suggestion to increase the level of lean awareness. This was done in order to understand the importance of production logistics and to reduce the waste in the assembly system in order to decrease the through-put time. The biggest wastes were over production and operation motions. Company A and E have worked with the early-on lean tools [11] for more then a year and we think that it was time to time to evolve from lean tools to the lean philosophies.

Table 4 Suggestion for future stage Suggestions for future stage Production area

Company A

Company B

Company C

Company D

Company E

Company F

Engine parts

Chemistry [28] Line

Electronics

Cooling modules [29] Line with U-cells

Trucks

Vessels [30]

Line

ATO

Kitting and ATO if possible Batches if possible

ATO

4 Increase LoAinformation

ATO if possible Smaller Batches if possible 6 Increase LoAinformation

Line with Ucells ATO

Improve flow between stations and other products

Improve Maintenance and competence

Type of flow

U-cell

Type of Assembling

ATO

Type Assembling

Batch

Number of stations Suggestions LoA Other suggestions

U-cell

One piece flow

High

Middle

7

Bottleneck analysis Involve the operators more

Start with material handling and production logistics – “Automation stair”

Improve flow between stations and other products

Middle

Decrease non-value adding tasks High

7 Conclusions This paper has analysed the need for six industrial companies in different areas to change Levels of Automation (LoA) in order to achieve their goals in terms of triggers for change. As shown in table 3, the companies' triggers for change contain either flexibility or time parameters. In order to achieve the triggers for change, the majority of the companies needed to change LoA as shown in table 4; o LoAinformation (50 % of the companies) in terms of digital assembly instructions in different levels due to the operators’ competence and experience, or visualisation of the production in terms of state lamps. o LoAmechanical (33 % of the companies) in terms of conveyers (transport automation), and variable automation in terms of redundancy [13] or plug-and-play flows. Furthermore, it was shown that the hypothesis that LoA could be a function and DYNAMO++ a tool to demonstrate possible improvement in order to

High

achieve higher parameters.

5 Increase LoAinformation

Smaller Batches if possible 7 Increase LoAmechanical

5 Increase LoAmechanical

Design For Assembly

Lean Awareness (use of JIT tools)

One piece flow

flexibility

High

and

decrease

time

8 Acknowledgement The authors would like to express their deep gratitude to the students, researchers and industries that participated in case studies and in the projects SIMTER, MyCar and ProAct. This research was financially supported by the Swedish Foundation for Strategic Research (SSF) and by the Swedish Governmental Agency for Innovation Systems (VINNOVA). References [1] M. Bellgran and K. Säfsten, Produktionsutveckling - Utveckling och drift av produktionssystem. Lund, Sweden: Studentlitteratur, 2005. [2] H. Aronsson, B. Oskarsson, and B. Ekdahl, Modern logistik. Lund: Grahns Tryckeri AB, 2006. [3] Å. Fasth, J. Frohm, and J. Stahre, "Relations between Performers/parameters and Level of Automation," in IFAC

[4]

[5]

[6]

[7] [8]

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16] [17]

workshop on manufacturing modelling, management and control Budapest, Hungary, 2007. N. Slack, "The flexibility of manufacturing Systems," International Journal of Operations & Production Management vol. Vol. 25 No. 12, 2005. D. Gerwin, "A framework for Analyzing the Flexibility of Manufacturing Process," in Business Administration Wisconsin: University of Wisconsin, 1983. J. Browne, D. Dubios, K. Rathmill, S. P. Sethy, and K. E. Stecke, "Classification of Flexible manufacturing systems," The FMS Magazine, April, 1984. J.-E. Ståhl, Industriella tillverkningssystem - länken mellan teknik och ekonomi vol. 2. Lund: Lunds tekniska högskola, 2006. W. J. Hopp, E. Tekin, and M. P. Van-Oyen, " Benefits of skill chaining in serial production lines with cross-trained workers," Management Science, vol. Vol 50, pp. 83-98, 2004. A. Gunasekaran, "Agile manufacturing: enablers and an implementation framework," International Journal of Production Research, vol. 36, pp. 1223 1247, 1998. J. K. Liker, The Toyota Way: 14 Management Principles from the World's Greatest Manufacturer. USA: McGraw-Hill, 2004. P. Olsson and M. I. Johansson, "Changes in planning and control system during implementation of Lean Production," in PLANs forsknings och tillämpningskonferens - kundfokuserade varor och tjänster Jönköping, Sweden, 2007. H. Yamamoto, "Driving information, An industry Case to enhance manufacturing competitiveness," in The 41st CIRP conference on manufacturing systems Tokyo, Japan, 2008. Å. Fasth, J. Stahre, and K. Dencker, "Analysing changeability and time parameters due to levels of Automation in an assembly system," in The 18th conference on Flexible Automation and Intelligent Manufacturing - FAIM Skövde, Sweden, 2008 R. Daniels, "A Prediction of the Future Machinery of Automation," Production Techniques, IRE Transactions on, vol. 3, pp. 14-16, 1958. A. I. Dashchenko, D. Ludwig, and O. Dashchenko, "Assembly automation - the way to a factory of the future," in Emerging Technologies and Factory Automation, 1995. ETFA '95, Proceedings., 1995 INRIA/IEEE Symposium on, 1995, pp. 259267 vol.2. J. P. Womack, D. T. Jones, and D. Roos, The Machine that Changed the World. New York: Harper Collins, 1990. T. Ohno, Toyota Production System: Productivity Press, 1988.

[18]

[19] [20]

[21]

[22]

[23] [24]

[25]

[26]

[27] [28]

[29]

[30]

[31]

P. Fitts, "Human engineering for an effective air navigation and traffic control system." vol. research foundation report Columbus,OH: Ohio state university, 1951. T. B. Sheridan, "Human centred automation: oxymoron or common sense?," IEEE, p. 6, 1995. J. Frohm, "Levels of Automation in production systems," in Department of production system. vol. Doctorial Gothenburg: Chalmers University of technology, 2008. Å. Fasth, J. Stahre, and K. Dencker, "Measuring and analysing Levels of Automation in an assembly system," in The 41st CIRP conference on manufacturing systems Tokyo, Japan, 2008 R. Parasuraman, T. B. Sheridan, and C. D. Wickens, "A model for types and levels of human interaction with automation," IEEE transactions on system, man, and cybernetics - Part A: Systems and humans, vol. 30, pp. 286-296, 2000. T. B. Sheridan, Telerobotics, automation and human supervisory control. Cambridge Massachussetts: MIT Press, 1992. V. Lindström, "Individuella bedömningar av dynamisk automationsnivå," Product and production development, Gothenburg,sweden 2005. V. Granell, J. Frohm, J. Bruch, and K. Dencker, "Validation of the DYNAMO methodology for measuring and assessing Levels of Automation," in Swedish Production Symposium Gothenburg, 2007. K. Dencker, J. Stahre, P. Gröndahl, L. Mårtensson, T. Lundholm, J. Bruch, and C. Johansson, "PROACTIVE ASSEMBLY SYSTEMS- Relaizing the potential of human collaboration with automation," in IFAC-CEA, cost effective automation in networked product development and manufacturing Monterey, Mexico, 2007. R. Patel and B. Davidsson, Forskningsmetodikens grunder vol. 3. Lund: Studentlitteratur, 2003. K. Jonsson, "Flow analysis and simulation of new and old production plant at Eka Chemicals AB," in Dep. of product and production development. vol. Master Gothenburg: Chalmers University of Technology, 2007, p. 43. P. Ottosson and G. Grundström, "Flow analysis and simulation of the cooling module production at NIBE AB," in Product and production development. vol. Master Gothenburg: Chalmers University of Technology 2008 A. Bergmann, "Case study on varying mechanical Level of Automation in an assembly system at Bosch Rexroth," in Department of product and production development. vol. Bachelor Gothenburg: Chalmers, 2008 G. Boothroyd and P. Dewhurst, ”Product Design For Assembly”. Wakefield, RI, 1987.

[32]

[33]

[34]

K. Dencker, J. Stahre, Å. Fasth, P. Gröndahl, L. Mårtensson, and T. Lundholm, "Characteristic of a Proactive Assembly System," in CIRP tokyo, 2008. K. Tanaka, N. Nakatsuka, H. Hibino, and Y. Fukuda, "Module structured production system," in The 41st CIRP conference on manufacturing systems Tokyo, Japan, 2008. P. Jonsson and S. A. Mattson, Logistik : läran om effektiva materialflöden. Lund: Studentlitteratur, 2005.

PAPER 4 Fasth, Å. Reviewing methods for analysing task allocation in a production system (Accepted for publication), Journal of Logistic management

Journal of Logistics Management 2012; 1(1): 1-8 DOI: 10.5923/j.logistics.20110101.01

Reviewing Methods for Analysing Task Allocation in A Production System Åsa Fasth 1

Department of Product and Production Development, Division of Production systems, Chalmers University of Technology, Gothenburg, SE-41296, Sweden

Abstract This paper reviews 10 methods or models developed during the last twenty years for redesign, measuring or analysing a production system. Furthermore, a comparison is done between the methods and models based on four focus areas with the aim of putting the developed DYNAMO++ and concept model into perspective due to the other methods and models. A literature study is used in order to review the methods and the focus areas. The result shows that the DYNAMO++ and the Concept model could be a golden way between the most socio-cognitive models and the technical-physical models when measuring and analysing a production system. The model also takes into consideration both physical and cognitive Levels of Automation in a more delicate scale than the other methods and models which makes the task allocation more precise.

Keywords

Production, Assembly, Task allocation, Levels of Automation (LoA)

1. Introduction Current tradition for design and usage of assembly systems may not be adoptable to the future needs and challenges that production companies have to face. When companies adopt automated solutions, they need to determine the correct amount of automation. However, it might be suboptimal to just evaluate the technical part of the system. Completely automated systems almost always have a human operator somewhere, at some level[1] so Chapains’ dream in 1970, to automate everything you possible can towards autonomous systems remain a dream, forty years later. Jordan[2] argued that men and machines/technique should be seen as complementary, rather than conflicting, resources when designing a man-machine system. Automation aims to extend the physical and cognitive capacity of people to achieve what might otherwise be impossible[3]. Therefore it is also vital to evaluate the socio- part of the system when changing a production system. Moreover, it is imperative to understand why to change the system and, if possible, break down these triggers into measurable goals so a comparison after the change could be executed. Empirical studies show that companies often used informal and unstructured evaluations of the current system for why to change the system (conducted from fifteen case * Corresponding author: [email protected] (Åsa Fasth) Published online at http://journal.sapub.org/ logistics Copyright © 2012 Scientific & Academic Publishing. All Rights Reserved

studies[4, 5], from late 1990s). According to Fasth et al[6], a majority companies, knew why to change their system (results from ten case studies conducted in 2007-2008). However, in line with Säfsten, the evaluations was often informal and unstructured i.e. feelings rather than facts. Today, a lot of different methodologies and models exist in order to describe and improve production systems. This paper will therefore discuss the following question; Are the methodology DYNAMO++[7] and the concept model[8] filling any gaps regarding task allocation in a production system? In order to discuss this question, ten design and measurement methods or models regard to four focus areas connected to redesign, measuring and analysing a production system has been chosen. In order to get an historical perspective and to determine trends regarding focus area 4, the methods chosen has a time horizon of twenty years (development from 1990 to 2011). Design and measurement methods or models 1. DYNAMO++ [9] and Concept model[8] 2. TUTKA production assessment tool[10] 3. Systematic Production Analysis (SPA)[11] 4. Productivity Potential Assessment (PPA)[12] 5. Lean Customisation Rapid Assessment (LCRA)[13] 6. A model for types and levels of human interaction with automation[14] 7. Complementary Analysis and Design of Production Tasks in Socio-technical Systems (KOMPASS)[15, 16] 8. Cognitive Reliability and Error Analysis Method (CREAM)[17, 18] 9. Task Evaluation and analysis Methodology (TEAM)[15, 19]

2

Fasth., Å: Reviewing methods for analyzing task allocation in a production system

10. Taxonomy for Cognitive Work Analysis[20] Focus areas 1. What assessment scale and level of change within the production system is the main focus? 2. Assessment objectives i.e. what is the methods’ main measurement parameters? 3. Assessment methods i.e. qualitative or quantities methods? 4. Where within the dimensions of Socio-Technical and Physical -Cognitive is the methodology’s main focus?

2. Methodology review The following sections will provide a short summary of each methodology and a summary of focus area 1-3. DYNAMO++ methodology and concept model The DYNAMO++ method and the concept model (seen in figure 1) for task allocation were developed during 2007-2009. The main aim is to evaluate and analyze the current stage of assembly systems due to triggers for change i.e. the company’s internal or external demands. The analysis focuses on three main areas Levels of Automation (LoA), Level of Competence (LoC) and Level of Information (LoI). These areas are divided into direct task i.e. assembling or value adding tasks and indirect tasks i.e. planning, improvement – non value adding tasks. The main loop propose possible improvements for the future state of the system, in a more structured way by using the LoA matrix[7], illustrated in figure 2 as a quantitative tool, and the Square of Possible Improvements (SoPI) as a qualitative tool.

SoPI in the LoA matrix. For the indirect tasks in the current system, LoC can be described as the accumulated and combined knowledge of an operator group working in the system. An example of an indirect task is planning of assembling. LoI can be described as the indirect information needed in the assembly cell in order to handle the indirect tasks i.e. other then assembling products. The information flow is divided into carrier (HOW the information is presented, i.e. phone, paper, other operators, PDAs etc) and content (WHAT is presented i.e. orders, alarms etc). The LoA analysis is done mainly on a task, station or cell level and from an operator’s perspective and are considering the direct task. The measurement and analysis consider both the physical (WITH WHAT to assemble) and cognitive (HOW, WHAT to assemble) automation in a system and is a further development of a taxonomy described by Frohm[22], resulting in a LoA matrix. The matrix is 7 by 7, which means that there are 49 possible solutions that could be chosen for the future state.

Main Loop

Figure 2. LoA Matrix

Figure 1. Concept model, further developed from DYNAMO++[8]

The model considering the competence of the operator group (LoC) in the direct tasks in the assembly system could be described according to Rasmussen’s Skill-Role-Knowledge (SRK) behavior levels[21] and as an competence matrix where the tasks and the number of operators are listed and combined, This could also be transformed into the cognitive LoA and showed visually in the

The Square of Possible Improvements (SoPI) is an analysis tool that narrows the 49 solutions down depending on the companies’ triggers for change. The SoPI step could be seen as an attempt to find a balance between facts and feelings (experience) when improving the system. TUTKA production assessment tool [10] The TUTKA production assessment tool was developed during the end of 2000s. The main aim with the tool is to assess the current state of a production system and to identify potential and means for improvements. The tool is comparing the current state of the system with a desired state i.e. a well performed production system, by using 33 key characteristics, 6 decision areas and 6 production objectives. Systematic Production Analysis (SPA) [11] The SPA was developed in 2007-2008 with focus on

Journal of Logistics Management 2011; 1(1): 1-8

manufacturing processes such as machining. The main aim is to measure the existing production condition and to simulate [23] different outcomes regarding three main parameters i.e. Quality (Q), Down-time parameters (S) and Production speed/tact (P) in order to reduce cost. The methodology has also been used in assembly operations[24], focusing on capacity flexibility and part cost. Two levels of automation is used to describe the assembly stations (manual/ automatic). Productivity Potential Assessment (PPA) [12, 25] The PPA method was developed during 2005-2006 by the institute of innovation and management at Chalmers University of Technology, Sweden. The main aim is to show the improvement potential of productivity in Swedish manufacturing companies. The parameters forming the PPA Method are divided into different 4 levels; Level 1 is the core of the method, constituting two parameters for measuring efficiency in manual work and machine work respectively. Level 2 parameters affect productivity at corporate level, Level 3 parameters indicate the company’s ability to improve the production while maintaining a sound work environment. Level 4 treats the potential of improving productivity by improving the ―M‖ factor of equation 1 Four levels of (mechanical) automation are used; 1) ManManual, 2) Semi – Semi-automatic 3) Auto – Automatic 4) Proc – Process industry Lean Customisation Rapid Assessment (LCRA) [13] This method is a further development of the Rapid Plant Assessment (RPA) method, which was developed to help managers to fast determine if a factory was lean or not and discern the factory’s strength and weaknesses were [26]. The main aim with the further develop method, LCRA, is to provide support in the analysis and/or design of a production system or even en entire company for mass customisation [27]. This is done thru three evaluation sheets divided into costumer elicitation, engineering and manufacturing. A model for types and levels of human interaction with automation[14, 28] The model is primarily used to analyze ATC (Air Trafic Control) systems with the issue; given specific technical capabilities, which system functions should be automated and to what extent? The human performance consequences of specific types and levels of automation constitute the primary evaluative criteria for automation design using the model. Secondary evaluative criteria include automation reliability and the costs of action consequences. Such a combined approach—distinguishing types and levels of automation and applying evaluative criteria—can allow the designer to determine what should be automated in a particular system. The model does not prescribe what should and should not be automated in a particular system. Hence, the model provides a more complete and objective basis for automation design than approaches based purely on tech-

3

nological capability or economic considerations. Ten levels of automation of decision and action selection is used for task allocation. Complementary Analysis and Design of Production Tasks in Socio-technical Systems (KOMPASS) [15, 16] The main aim with the COMPASS method is to design production systems were human has control over technology i.e. automated systems. Expert analysis of existing systems is done based on three levels of analysis criteria; work system, human work tasks and human machine system. The method is built on the complementary principle[2] when designing a system i.e. humans and machines are fundamentally different and can therefore not be compared on a quantitative basis but complementing each other, performing tasks in a joint cognitive system[29] Cognitive Reliability and Error Analysis Method (CREAM) [17, 18] CREAM is a Human Reliability Analysis (HRA) method i.e. modeling cognitive errors and error mechanisms into the risk assessment processes. The basic notion is that of contextual control modeling, i.e., describing human cognition in terms of the competence for actions and the way in which the actions are controlled. CREAM can be used to identify the most likely cause of an observed event--either an accident or an erroneous action. The method can also be used in a predictive way to derive the likely consequences of specific erroneous actions. Task Evaluation and analysis Methodology (TEAM) [15, 19] The method was developed between 1994-1996. The main aim is to evaluate existing advanced manufacturing systems (AMS) from a user perspective in order to pinpoint efficiency problem areas. Further to provide support for humans to better interact with complex technology[15]. Task analysis is presented in an evaluation matrix, developed by Stahre [30], based on a combination between Sheridan’s supervisory control and Rasmussen’s human behaviour levels. Four factors are considered; work environment, work tasks, information flow and system performance[19]. The method should ideally be performed by multidimensional system design teams with at least one human factor specialist. Three levels is used for task evaluation; 1) generally difficult, 2) differentially difficult, 3) tasks known by few operators Taxonomy for Cognitive Work Analysis [20, 31] This taxonomy was first published in 80s and should be used for effective support of decision processes to create a work practise that suits the individual users’ cognitive resources[20]. A work domain should be represented at five levels of abstraction, representing goals and requirements, general functions, physical processes and activities, as well as material resources[20]. Any of these levels has a work function (what should be used) which can be seen both as a goal (why it is relevant) for a function at a lower level, and as a means for a function at a higher level (how this is real-

4

Fasth., Å: Reviewing methods for analyzing task allocation in a production system

ized)[32]. Moving from a lower level to a higher level of abstraction means a change in the representation of system properties.

3. Focus area 1-3 The following chapter will describe and define focus area 1-3 and show a summary of the focus areas regard to the different methods. Focus area 1: What assessment scale and level of change within the production system is the main focus? The assessment scale could be described as the deepness of the methodology in the production system. Figure 3 illustrates seven structuring levels and two views; the resource view proposal by Westkämper[33] and the space view proposal from Nyhuis[34] based on H-P Wiendahl[35]. The resource view looks for the technical and human resources, which maintain the processes whereas the space view considers the architectural objects which have to be designed in accordance with these resources. The resource view is used in this paper to describe the deepness of the methods and models, seen in table 1.

Focus area 2: Assessment objectives i.e. what is the methods’ main measurement parameters? As been said in the introduction, it is important to know why to change a system and to have parameters to compare the current state and the system after the changes in order to see if the goals with the change have been achieved. In this paper these parameters are divided into two different types; PDM -Parameters that are direct measurable (quantitative) i.e. time, cost PIDM -Parameters that are indirect measurable (qualitative) i.e. Flexibility, Complexity, Proactivity Focus area 3: Assessment methods i.e. qualitative or quantities methods? Different assessment methods could be used in order to collect the data needed for analysing the system. In this paper the methods are divided into qualitative and quantitative approaches. The differences with the approaches could be described as; Qualitative research is to understand the meaning of a certain phenomenon or discovery, whereas quantitative research dissects the phenomenon to explore its components, which later become the studied variables[38]. These approaches also have a relation due to the different dimensions that will be described in focus area 4. If the method or model is more focused on a socio-dimension tendencies are to used qualitative methods i.e interviews, observations etc, while if the method investigates more of a technical approach it tends to be more quantitative methods i.e. measurements and pre-defined criteria. Parasuraman et. al[14] argues that automation design is not an exact science i.e. solely quantitative, however, neither does it belong in the realm of the creative arts. Table 1. Comparison between the different models and methods regarding focus area 1-3

Figure 3. tructuring levels and views of a factory [36], edited.

Tasks within stations has been added as a level in the model, in figure 3[33, 35] [34]. This is done to be able to in count task allocation in the model e.g. the assessment scale in this paper is a maximum of seven (not including the processes which are the resources, machines and/or humans working in the different levels[36]) start counting from task level and up. The level of change could be described as a two degree change according to Porras and Robertsson[37]; 1st degree – Changes or improvements in the current system and 2nd degree - Redesigning the system.

Journal of Logistics Management 2011; 1(1): 1-8

4. Focus area 4 The forth focus area has a more holonic perspective and are looking at the methods to see what kind of focus they have regarding the ―socio and technical schools‖, but also to identify trends and similarities between the methods. Where within the dimensions of Socio-Technical and Physical-Cognitive is the methodology’s main focus? The dimensions seen in figure 4 have been chosen because most production and manufacturing system can be described within these.

5

develop multiple skills in the individual and immensely increase the response repertoire of the group. 5. The socio-technical principle valued the discretionary rather than the prescribed part of work roles[45] 6. The principle treated the individual as complementary to the machine rather than as an extension of it[2] 7. It was variety-increasing for both the individual and the organization rather than a variety in the bureaucratic mode Out of the dimensions illustrated in figure 4, four areas could be discussed. The following sections will bring up some definitions of these areas and position the models and methods in within these areas;

Socio- Physical

Figure 4. The dimensions of Socio-Technical and Physical-Cognitive

Socio-technical dimension The Socio-Technical viewpoint of manufacturing systems emerged from two different schools with the idea to combine technology, organization and human growth in order to maximise the system performance. The theory was first formed in 1950 at the Tavistock institute of human relations, with the beginning of the well known empirical analysis by Trist and Bamford at the English coal mines[39]. The school of human relation[40] which also could be described as the social sub system handle members of the organisation, individual demands and qualifications and group specific demands[41]. The second school, scientific management[42] – rational production engineering, also described as the technical sub system[41] handle resources, technologies and methods. These thoughts led to a new paradigm of work with seven principles that differ from the old way of thinking[43]; 1. The work system, which comprised a set of activities that made up a functioning whole, now became the basic unit rather than the single jobs into which it was decomposed 2. Correspondingly, the work group became central rather than the individual job-holder 3. Internal regulation of the system by the group was thus rendered possible rather than the external regulation of individuals by supervisors 4. A design principle based on the redundancy of functions rather than the redundancy of parts[44] characterized the underlying organisational philosophy which tended to

A composite class of sciences which intersect in a fundamental way both to the physical and social science could be called socio-physical sciences. Furthermore, socio-physical sciences must be vitally concerned with human behaviour and objectives. An example is industrial engineering. The eventual extensive use of automation will bring into the class many additional areas now considered components of social or physical sciences. The necessity of considering a technology directly in terms of goals of the social groups involved in or affected by that technology is what ultimately distinguishes this class of sciences[46]. The methodologies or models place in this quadrant are models with the main focus on strategies and organisation theory. None of the chosen methods is put here. Keywords: Industrial engineering, Industrial sociology, economics and psychology, organization and communication

Socio-Cognitive

Socio-cognitive engineering aims to analyze the complex interaction between people and computer-based technology and then transform this analysis into usable, useful socio-technical systems, a dialectical relationship to user-centered design[47]. Humans developed a unique socio-cognitive ability to cognitively create information that they then store, organize, retrieve and used. There is a critical need to understand and incorporate information behavior into an evolutionary and life-span understanding of human behavior[48]. The methodologies or models place in this quadrant are models with the main focus on Humans-in-control and human based systems further the assessment methods are

6

Fasth., Å: Reviewing methods for analyzing task allocation in a production system

often qualitative, i.e. CREAM,TEAM, COMPASS, A model for types and levels of human interaction with automation and Taxonomy for Cognitive Work Analysis

An explanation to why none of the methods or models was place in the socio-physical or technical-cognitive is that the methods is chosen due to task allocation and assessments mainly on a shop-floor level. These two quadrants are on a more strategic level.

Technical - Cognitive

Cognitive Technical Systems differ from other technical systems in that they perform cognitive control and have cognitive capabilities[49, 50]. A cognitive system is goal directed; it uses knowledge about itself and its environment to monitor, plan, and modify its actions in the pursuit of goals; it is both data and concept-driven. The development and application of Cognitive Technical Systems (CTS) aims at an integrated approach for the planning and execution, as well as the continuous learning and adaptation of processes in technical systems under unpredictable circumstances[50]. Furthermore it could be described as a single, integrated system composed of both human and artificial cognitive systems[51]. Advances in computation AI technology have greatly expanded the potential for the support of human cognitive activities and for the development of artificial cognitive systems--i.e., systems that perform tasks normally associated with human cognition[52]. Methodologies or models place in this quadrant are models with the main focus on autonomous systems. None of the chosen methods are placed here, But one example of such a method or model could be the cognitive factory[50]. Keywords; AI, evolvable systems, autonomous systems, cognitive agents, cognitive engineering

Figure 5. A summary of the methods and models placed in the different dimensions

Furthermore the figure shows that the methods and models conducted in the 1990s were more focused on the socio-cognitive aspects in the system than the methods developed after 2000. A trend towards more quantitative and techniocal-physical focus of the methods is clear. The aim with the DYNAMO++ and the concept model are therefore to keep the socio-cognitive aspect, but also to be more concreate and quantitative. One example of this is to see people as possible resources in the design face and not to let them take the ―left over automation‖ tasks when the system has been designed. By thinking and optimising the system to both humans and machines, DYNAMO++ and the concept model could be seen as a method that is placed in the middle of socio-cognitive and technical-physical squares.

5. Conclusions Technical-Physical

This quadrate could be described as the most technical or mechanized way of describing a production system or rational production engineering. The methods often use quantitative methods such as PDM: s or pre-defined criteria which are assessed with the system. For example; in the PPA method a questionnaire with 40 yes or no questions (or predefined criteria of a productive system) is used to determine the level of productivity within the system (above 35 yeses is seen as a well functional productive system) Out of the ten methods, the assessment tools (PPA, LCRA and TUTKA) and the SPA method is placed here. Keywords: Production engineering, System design, Physical automation, scientific management MARY OF FOCUS AREA 4 A summary of the chosen methods and models placed in the different areas is shown in figure 5.

The result shows that the DYNAMO++ and the Concept model could be a golden way between the most socio-cognitive models and the technical-physical models when measuring and analysing a production system based on two main reasons. The model takes into consideration both physical and cognitive Levels of Automation in a more delicate scale than the other methods and models which makes the task allocation measurements and analysis more precise. Furthermore, the model also considers the social aspects in terms of competence within the operator group and the information flow to and from the cell or station.

ACKNOWLEDGMENTS The authors want to express their deep gratitude to VINNOVA (The Swedish Governmental Agency for Inno-

Journal of Logistics Management 2011; 1(1): 1-8

vation Systems) for funding this research.

REFERENCES [1]

S. Dekker and D. Woods, "MABA_MABA or ABRAKADABRA? Progress on HUman-Automation Co-ordination," Cognition, Technology and Work, vol. 4, pp. 240-244, 2002.

[2]

N. Jordan, "Allocation of functions between human and machine in automted systems," Journal of applied psychology, vol. 47, pp. 161-165, 1963.

[3]

J. D. Lee, "Review of a pivotal Human Factors Article: "Human and Automation: Use, Misuse, Disuse, Abuse"," Golden anniversity special issue of Human Factors, vol. 50, pp. 404-410, 2008.

[4]

[5]

[6]

[7]

K. Säfsten and E. Aresu, "Vad är bra monteringssystem?: En studie av utvärdering och utformning på 15 industriföretag i Sverige," Linköpings universitet, Linköping2000. M. Bellgran and K. Säfsten, Produktionsutveckling - Utveckling och drift av produktionssystem. Lund, Sweden: Studentlitteratur, 2005. Å. Fasth and J. Stahre, "Does Levels of Automation need to be changed in an assembly system? - A case study," in Proceedings of the 2nd Swedish Production Symposium (SPS), Stockholm, Sweden, 2008. Å. Fasth, J. Stahre, and K. Dencker, "Measuring and analysing Levels of Automation in an assembly system," in Proceedings of the 41st CIRP conference on manufacturing systems Tokyo, Japan, 2008

7

model for types and levels of human interaction with automation," IEEE transactions on system, man, and cybernetics Part A: Systems and humans, vol. 30, pp. 286-296, 2000. [15] T. Wäfler, A. Johansson, G. Grote, and J. Stahre, "Complementary interaction of humans and machines in highly automated production systems - KOMPASS and TEAM," in 13th Triennial Congress of the international Ergonomic Assosiation, Finnish institute of occupational health, 1997, pp. 227-279. [16] G. Grote, "Uncertainty management at the core of system design," Annual Reviews in Control, vol. 28, pp. 267-274, 2004. [17] E. Hollnagel, "Reliability analysis and operator modelling," Reliability Engineering and System Safety pp. 327-337, 1996. [18] E. Hollnagel, "HRA -- The First Generation," in Cognitive Reliability and Error Analysis Method (CREAM) Oxford: Elsevier Science Ltd, 1998, pp. 120-150. [19] A. Johansson, "Decision support tools in manufactoring systems: a task evaluation method," in European conference on integration in manufactoring, Amsterdam, 1994. [20] J. Rasmussen, A. M. Pejtersen, and K. Schmidt, Taxonomy for Cognitive Work Analysis. Roskilde, Denmark: Grafisk service, 1990. [21] J. Rasmussen, "Skills, Rules, Knowledge, Signals, Signs and Symbols and other Distinctions Human Performance Models," IEEE Transactions on Man, Systems and Cybernetics, pp. 257-266, 1983. [22] J. Frohm, V. Lindström, M. Winroth, and J. Stahre, "Levels of Automation in Manufacturing," Ergonomia IJE&HF, 30:3, 2008.

[8]

Å. Fasth and j. Stahre, "Concept model towards optimising Levels of Automation (LoA) in assembly systems," in Proceedings of the 3rd CIRP Conference on Assembly Technologies and Systems, Trondheim, Norway, 2010.

[23] M. Jönsson, C. Andersson, and J.-E. Ståhl, "Implementation of an Economic Model to Simulate Manufacturing Costs," in The 41st CIRP conference on manufacturing systems Tokyo, Japan, 2008.

[9]

Å. Fasth, J. Bruch, K. Dencker, J. Stahre, L. Mårtensson, and T. Lundholm, "Designing proactive assembly systems (ProAct) - Criteria and interaction between automation, information, and competence " Asian International Journal of Science and Technology in production and manufacturing engineering (AIJSTPME), vol. 2 (4), pp. 1-13, 2010.

[24] [24] C. Andersson, M. Jönsson, and J.-E. Ståhl, "Capacity flexibility and part cost," in Proceedings of the 3rd Swedish Production Symposium (SPS), Stockholm, sweden, 2009.

[10] M. Koho, "Production System Assessment and Improvement - A tool for MTO and ATO companies," in Production engineering. vol. Doctorial Tampere: Tampere university of technology, 2010.

[25] P. Almström and A. Kinnander, "PRODUCTIVITY POTENTIAL ASSESSMENT OF 30 SUPPLIERS TO THE AUTOMOTIVE INDUSTRY," in Proceedings from the 3rd Swedish Production Symposium, Stockholm, Sweden, 2009 [26] R. Goodson, Read a plant - fast: Harvard business review, 2002.

[11] J.-E. Ståhl, Industriella tillverkningssystem - länken mellan teknik och ekonomi vol. 2. Lund: Lunds tekniska högskola, 2007.

[27] M. Comstock and S. Bröte, "Beyond 'Read a plant - fast' (for lean): read an enterprise for mass customization?," in Proceedings of the 2nd international conferens on mass costumization and personalization Munich, Germany, 2003.

[12] P. Almström and A. Kinnander, "PRODUCTIVITY POTENTIAL ASSESSMENT OF THE SWEDISH MANUFACTURING INDUSTRY," in Proceedings from the 1st Swedish Production Symposium, Gothenburg, Sweden, 2007

[28] R. Parasuraman and C. D. Wickens, "Humans: Still Vital After All These Years of Automation," Golden anniversity special issue of Human Factors, vol. 50, pp. 511-520, 2008.

[13] M. Comstock, "Production systems for mass costumization bridging theory and practice," in Dep. of mechanical engineering vol. Doctorial Linköping: Linköping university, 2004. [14] R. Parasuraman, T. B. Sheridan, and C. D. Wickens, "A

[29] E. Hollnagel and D. D. Woods, Joint cognitive systems foundation of cognitive system engineering: CRC press, Taylor and Francis group, 2005. [30] J. Stahre, "Evaluating human/machine interaction problems in advanced manufacturing," Computer Integrated Manu-

8

Fasth., Å: Reviewing methods for analyzing task allocation in a production system

facturing Systems, vol. 8, pp. 143-150, 1995. [31] J. Rasmussen, "The Role of Hierarchical Knowledge Representation in Decisionmaking and System Management," IEEE Transactions on Man,Systems and Cybernetics, vol. SMC-15/2, pp. 234-243, 1985. [32] J. Rasmussen, A. M. Pejtersen, and L. P. Goodstein, Taxonomy for work analysis. Design of work and development of personnel in advanced manufactoring. New York: John Wiley & Sons,inc p. 41-48, 1994. [33] [33] E. Westkämper, "Digital Manufacturing in the global Era," in Proceedings of the 3rd International CIRP Conference on Digital Enterprise Technology, Setúbal, Portugal, 2006. [34] P. Nyhuis, Kolakowski, and H. M., C. L, "Evaluation of Factory Transformability," in 3rd Interna-tional CIRP Conference on Reconfigurable Manu-facturing Ann Arbor, USA, 2005. [35] H.-P. Wiendahl, "Wandlungsfähigkeit: Schlüs-selbegriff der zukunftsfähigen Fabrik (Transformabi-lity: key concept of a future robust factory)," in Werkstattstechnik online. vol. 92/4, 2002. [36] H. P. Wiendahl, H. A. ElMaraghy, P. Nyhuis, M. F. Zäh, H. H. Wiendahl, N. Duffie, and M. Brieke, "Changeable Manufacturing - Classification, Design and Operation," CIRP Annals - Manufacturing Technology, vol. 56, pp. 783-809, 2007. [37] J. I. Porras and P. J. Robertsson, Organizational Development: Theory, Practice and Research in Handbook of industrial and organisational psycology. California: Consulting Psychologist Press inc, 1992.

[41] E. Ulich, Arbeitspsychologi. Sturtgart, 1994. [42] F. W. Taylor, Principles of Scientific Management, 1911. [43] E. L. Trist, "Evolution of socio-technical systems - a conceptual frame work and an action research program," in Conference on organizational design and performance, University of Pennsylvania, 1980. [44] F. E. Emery, "The next thirty years: Concepts, methods and anticipation," Human relations, vol. 50, (20), pp. 199-237, 1997, (1967). [45] E. Jaques, Measurement of responsibility: A study of Work, Payment and individual capacity. New York: Dryden, 1956. [46] A. Abruzzi, "Problems of interference in the socio-physical scienses," the journal of philosophy, vol. 19, pp. 537-549, 1954. [47] M. Sharples, N. Jeffery, J. B. H. d. Boulay, D. Teather, B. Teather, and G. H. d. Boulay, "Socio-cognitive engineering: a methodology for the design of human-centered technology," APPLIED COGNITIVE SCIENCE AND HUMAN CENTERED SYSTEMS, pp. 67-73, 1999. [48] A. Spink and C. Cole, "Information Behavior: A Socio-Cognitive Ability," Evolutionary Psychology vol. 5, pp. 257-274, 2007. [49] M. Beetz, M. Buss, and D. Wollherr, "Cognitive Technical Systems - What Is the Role of Artificial Intelligence? ," in Proceedings of the 30th German Conference on Artificial Intelligence, 2007.

[38] S. B. Merriam, Fallstudien som forskningsmetod. Lund: Studentlitteratur, 1994.

[50] M. F. Zäh, M. Beetz, K. Shea, G. Reinhart, K. Bender, C. Lau, M.Ostgathe, W. Vogl, M. Wiesbeck, M. Engelhard, C. Ertelt, T. Rühr, M. Friedrich, and S. Herle, Cognitive factory, chapter 20: Springer, 2009.

[39] E. L. Trist and K. W. Bamforth, "Some Social and Psychological Consequences of the Longwall Method of Coalgetting," Human Relations, vol. 4, p. 35, 1951.

[51] E. Hollnagel and D. D. Woods, " Cognitive systems engineering: New wine in new bottles," International Journal of Man-Machine Studies,, vol. 18, pp. 583-600, 1983.

[40] G. Boalt and G. Westerlund, Arbetssociologi. Stockholm: Tidens Förlag, 1953.

[52] D. D. Woods, "Cognitive Technologies: The Design of Joint Human-Machine Cognitive Systems," The AI magazine, vol. 6, 1985.

PAPER 5 Fasth, Å. Bruch, J., Dencker, K., Stahre, J., Mårtensson, L. and Lundholm, T. (2010) Designing proactive assembly systems (ProAct) - Criteria and interaction between automation, information and competence, Asian International Journal of Science and Technology in production and manufacturing engineering (AIJSTPME), vol 2 issue 4, pp.1-13

AIJSTPME (2009) 2(4): 1-13

Designing proactive assembly systems (ProAct) – Criteria and interaction between automation, information, and competence Fasth Å. Chalmers University of Technology, Division of Production Systems, SE-41296 Stahre J. Chalmers University of Technology, Division of Production Systems Bruch J. Chalmers University of Technology, Division of Production Systems and Jönköping University, Department of Industrial Engineering and Management Dencker K. Royal Institute of Technology, Department of Production Engineering and Swerea IVF, Industrial Research & Development Corp Lundholm T. Royal Institute of Technology, Centre for Design and Management of Manufacturing systems Mårtensson L. Royal Institute of Technology, Department of Industrial Economics and Management Abstract Production companies of today face extreme challenge to meet the rapid changes and increased flexibility that mass customization require. More and more customers are requiring the product to suite specific needs such as design, function and sustainability. These requirements results in increasing demands for the developers of the product but also for the personnel who will assemble the products in the final assembling. This paper suggests the need for further development, primarily addressing time parameters in dynamically changing assembly systems. We propose proactivity as a vital characteristic of semi-automated assembly systems, to increase fulfilment of customer demands and decrease non value-adding tasks. In proactive assembly systems, the potential of human operators and technical systems is utilised. Criteria for proactivity are reviewed from automation, information, and competence perspectives. Empirical data have been collected from five production companies in Sweden. Keywords: Assembly systems, Automation, Information and Competence 1 Introduction Current tradition for designing and developing assembly systems may not be adaptable to the needs and future challenges that production companies have to face. Frequent demands and requirements, both internal and external trigger a plan for change in different production areas. Extreme flexibility will be needed in future assembly systems to be able to

handle rapid changes as a result of an increasing number of product variants e.g. mass customisation. This will result in smaller batches and shorter time limits for set-up between products. As a result, companies have to find more flexible methods for assembling their products and become more proactive in the assembly system itself. Indentifying

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

1

Fasth Å. et al. new strategies to reduce time parameters in a system e.g. cycle-time, set-up time, throughput time and non value-adding tasks becomes vital. According to Frese and Fay [1], research in general focuses on reactive concepts [2] e.g. the operators have to perform already given tasks and solve problems which have already occurred. We believe that the operator group within the assembly system should have the ability to work proactively and that the operators‟ job flexibility or action space should increase so that problems can be prevented and thereby decrease the downtime in the system. The work situations an operator will face in the future are unlikely to be exactly identified by work instruction sheets. It is assumed that assembly work settings enabling proactive behaviour of the human operators will be important to handle the increasing uncertainty of work contexts [3]. The objective of the ProAct project [4] is to identify proactive solutions for time minimisation at the operational shop floor level in assembly systems [5]. The approach is based on the concept of proactive behaviour; as the ability of operators to control a situation by taking action and effectuating changes of the work situation in advance in order to create a favourable outcome [6]. A proactive behaviour can be characterised by three main areas [7]; • Anticipation of problems related to change • Initiation of activities that lead to solution of change related problems and improvements in work, and • Resolution of change related problems. Proactive operator work was defined by Griffin et al [3] as “the extent to which the individual takes selfdirected action to anticipate or initiate change in the work system or work roles”, Operators should then have correct situation awareness, longer time horizon and an anticipative perspective on continuous improvements of their work places [8]. This could lead to minimising of predicted and unpredicted disturbances and thereby, the availability of the assembly system will be increased [9]. Proactivity is not used as a competition benefit by companies today, even though proactive behaviour can result in increased organisational effectiveness [10]. Ideas similar to proactivity have been presented. The concept that operators of complex systems can perceive, comprehend, or project system behaviour, to increase control ability and precision, was defined as situation awareness [18]. Our assumption is that proactivity is highly influenced by three areas of

2

assembly systems design and operation [11] i.e. the Levels of • Automation (LoA) • Competence (LoC) and • Information (LoI) It is vital to find criteria and interaction between these and to combine the right levels of the suggested parameters in order to maximise the systems flexibility and the operators‟ action space. This paper will address the criteria and interaction between the three areas: automation, information, and competence. Further the paper will describe a Meta methodology (the ProAct loop), for design of a proactive system in a structured way with the production goal to increase flexibility and achieve time minimisation. 2 Theoretical framework This section will briefly introduce three concepts: levels of automation, information and competence. Further, different methodologies used will be presented. 2.1 Level of automation (LoA) There is no uncomplicated way to make automation human-oriented that is applicable across all domains and types of work. Different processes and domains may put different emphasis on precision, stability and/or speed of production. Human-centeredness is therefore a reminder of the need to consider how the system can remain in control, rather than a goal in itself. Each case requires its own considerations of which type of automation is the most appropriate, and how control can be enhanced and facilitated via a proper design of the automation involved [12]. Smart automation is defined by Ohno [13] as the human aspect of 'autonomation' whereby automation with a human touch is achieved. Fitts MABA-MABA list (Men-Are-Better-At – Machines-Are-Better-At) [14], from 1951, gives an indication of how to allocate tasks between humans and machines in technical systems, e.g. assembly systems. Sheridan [15] argues that if tasks in which machines are better are automated and operators are still required to monitor such automation and maintain full situation awareness [16] we might lose more than we gain. Sheridan and Verplank (1978) proposed a scale of levels of automation (LoAs) involving automation of decision making and action [17]. However, automation also includes information gathering and analysis. Parasuraman et. al [18] accordingly

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

Designing proactive assembly systems (ProAct) – Criteria and interaction between automation, information, and competence proposed an extension of the LoA concept to four information-processing stages: (a) information acquisition, (b) information analysis, (c) decision making, and (d) action, with each stage having its own LoA scale (for similar scales, see [19];[20]). Table 1 provides a historical review of levels of automation research. The levels of automation concept, used in this paper was defined by Frohm [21] as; “The allocation of physical and cognitive tasks between humans and technology, described as a continuum ranging from totally manual to totally automatic” Frohm [21], defines physical tasks as the level of automation for mechanical activities, mechanical LoA, while the level of cognitive tasks is called information LoA. Further it could be explained as; mechanical LoA is WITH WHAT to assemble and cognitive LoA is HOW to assemble on the lower levels (1-3) and situation control on the higher level (4-7). Table 1: Definition of Levels of Automation [22] Year 1958

1962

1980

1981

1985

1997

Definitions of Levels of Automation Divides the levels depending on who is initiating control; the human (1-4), the human together with automation (5-8) or the automation (9-17), (Bright, [23]) The extent to which human energy and control over the production process are replaced by machines (Amber & Amber, [24]) The level of automation incorporates the issue of feedback, as well as relative sharing of functions in ten stages (Sheridan, [25]) Automaticity is defined in six levels from conducting the tasks manually, without any physical support, to fully automated congnition with computer control (March & Manari, [26]) The degree of mechanisation is defined as the technical level in five different dimensions or work function (Kern & Schumann, [27]) The level of automation goes from direct manual control to largely autonomous operation where the human role is minimal (Billings, [28])

1997

1998

2000

2001

2002

2008

The level of automation in the context of expert systems is most applicable to cognitive tasks such as ability to respond to, and make dicisions based on, system information (Endsley, [29]) The level of automation is defined as the sharing between the human and the machines with different degrees of human The interaction and task division between the human and the machine should instead be viewed as a changeable factor which can be called the level of automation (Parasuraman et. al, [18]) Level of mechanisation can be defined as the manning level, with focus on operating of the machines (Groover, [31]) ‘Manual’ tasks as being those in which humans are responsible for conducting the task. „Semi-automatic‟ is a higher level of automation and involves automated alignment and application of epoxy by a robot. Material handling, on the other hand, is still conducted by humans, unlike „automatic‟, where material handling is also automated. (Ducheon, [32]) The allocation of physical and cognitive tasks between humans and technology, described as a continuum ranging from totally manual to totally automation (Frohm, [22])

2.2 Dynamo++ The DYNAMO++ methodology was developed [33] based on previous research [21]. A taxonomy for measuring both cognitive (information) and physical (mechanical) Level of Automation (LoA) in a current stage of a system was defined [22]. The aim was to measure accessible LoA in order to find an appropriate span of levels of automation and, by that, maintain high productivity by reducing production disturbances [21]. The focus of further development is on the analysis and clarifies whether the current systems‟ LoA is too high, too low, or too static in order to fulfil the companies' triggers for change e.g. internal or external demands. Further to redesign that system in a structured way with the most advantageous LoA.

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

3

Fasth Å. et al. The stakeholders for this methodology are firstly production engineers, who will have a structure to work with when choosing type and amount of automation for a specific part of a production system. The DYNAMO++ methodology is divided into three different states; current, future and new current stage, as illustrated in Figure 1. These stages contain four different phases; 1) Pre study, 2) Measurement, 3) Analysis and

4) Implementation [33, 34]. These phases or methods are used to answer different questions in order to redesign the system in a structured way and avoid over or under automated systems. Focus in this paper is within the current stage and the future stage. A brief description of the methods used is given in the sections below.

Figure 1: The DYNAMO++ methodology [34] 2.1.1 Current stage Pre study phase • “Go see”, Genchi Genbutsu, e.g. observe the production floor without preconceptions and with a blank mind [13, 35], and chose a part of the system where must improvements are needed. • Get an accurate overview of the current system (number of operations) and the different flows (information, material and product) in the chosen area by using for example VSM (Value Stream Mapping [36]). Measurement phase • Divide the chosen operations into main and sub tasks by using Hierarchical Task Analysis (HTA). The HTA is a method for description of activities under analysis in terms of a hierarchy of goals, sub-goals, operations, and plans [37]. • Measure the current mechanical and cognitive LoA in the subtasks, based on the reference matrix presented in Figure 2.

4

Figure 2: The LoA matrix 2.1.2 Future stage Analysis phase The measured LoA value and relative minimum and maximum values for each task are illustrated in a matrix. This results in a Square of Possible Improvements (SoPI), with both physical and cognitive automation possibilities illustrated in Figure 3.

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

Designing proactive assembly systems (ProAct) – Criteria and interaction between automation, information, and competence

Figure 3: The LoA matrix with a SoPI [33] The span of different automation solutions contribute to a larger span of possible solutions for replacement and in real time optimisation [5, 38]. Equipment for automation for mechanical tasks, cognitive tasks and information tools can be evaluated with the criteria anticipation, initiation of activities, and resolution of change-related problems [11]. 2.3 Level of competence (LoC) and the operator role in the assembly system According to ISO SS 62 40 70 [39] competence is defined as “the ability and willingness to carry out a task by applying knowledge and skills”. When defining competence in the ProAct project the following implications were made: Ability – experience, comprehension and judgment to use knowledge and skills in practice, where willingness is the attitude, commitment, courage and responsibility; knowledge means facts and methods – to know and skills is to carry out in practice – to do [5]. A model frequently used for human supervisory control of highly automated workplaces was developed by Sheridan [40], primarily for application in process and aerospace industry. The model was further developed and adapted for manufacturing industry by Stahre [41]. The model includes five main operator roles i.e. plan, teach (programming), perform, intervene and learn. Plan means that the operator decides what to do in the system. The teaching step describes how to transfer the proposed plan to the technical system through programming but also to assure that the correct tools and materials are available to fulfil the plan. Monitoring means that the operator starts the process and controls the products when deviation occurs. The operator has to intervene when different disturbances occurs. The operator learns from every new task, product or disturbance and can use this when planning the next batch. A further development of Sheridan’s five operator roles with a combination

of human needs, phrased by Maslow (1954) and the socio-technical school [42, 43] was done by Mårtensson [44]. The six requirements and related criteria are general and applicable to all kinds of work; • A versatile work content (e.g., the individual should plan, perform, and monitor the production task) • Responsibility and participation (e.g., responsibility for the complete work task, participation in the design process) • Information processing (e.g., planning one‟s work, cognitive activity in new situations, problem solving at disturbances followed by decision-making) • Influence on physical work performance (e.g., choosing an automatic or manual method if possible, physical mobility, temporarily leaving the work place for a short while) • Contact and cooperation with colleagues (e.g., verbal and visual contact with at least one person, contacts with programmers or assembly department, cooperation in teams) • Competence development (e.g., to the individual acceptable skill level, competence being used in more qualified tasks, continuous training) [45]. The list could be used as a checklist when designing the work organisation in any human – machine system and when designing decision support for system operators. The need for cognitive support can be divided into three separate levels, skill-, rule- and knowledgebased, described by Rasmussen as behaviour levels [46]. At the skill-based level activities are carried out with very low need for conscious attention or control. Many tasks are skill- based; they are performed more or less “automatically” by the operator with minimal thinking and can be seen as LoA cognitive= 1 e.g. it can be connected to previous physical practice and experience. If a problem or variation between anticipated and real outcome in the performed task occurs, the operator must increase the mental effort to the rule-based level. If tasks are standardised, e.g. by having tools or techniques [47, 48] to give decision or teaching the operator how to perform a task in the assembling, it becomes rule based and could be transformed into LoAcognitive= 2-4. If an unforeseen situation occurs, or if no rules are applicable to the specific task, the operator's mental effort is raised to the knowledgebased level. Arguments are then made from detailed analyses and there is a need for conscious and

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

5

Fasth Å. et al. focused operator attention. A combination between Sheridan’s supervisory control and Rasmussen’s human behaviour levels was developed by Stahre [41] as an evaluation matrix, used for task analysis, illustrated in Figure 4.

The list is used as a reference when gathering empirical data in case studies in order to investigate and map the operators‟ action space e.g. involvement in these areas which will be brought up in section 3.1. 2.4 Level of information (LoI)

Figure 4: Evaluation matrix, combining supervisory control roles and human behaviour levels [41]

Information could be described as a collection of facts organised in such a way that they have additional value beyond the value of the facts themselves. An information system is a set of interrelated components that collect, manipulate and disseminate data and information and provide a feedback mechanism to meet an objective [49]. A well functioning information flow throughout the chain is a powerful and necessary prerequisite [50]. A modern information system constitutes a network of multi-media work-stations, designed to present an operator with an information environment. This enables the user, within the constraints posed by the work requirements to create actively a work practice that suits the individual users‟ cognitive resources and subjective preferences [51]. Frequent exchange of information between different company levels is crucial to create quick “decisionloops” which are correctly connected and updated with real manufacturing conditions [8]. To achieve efficiency in assembly systems, the information flow has to fulfil several qualitative criteria defined by [52] as; • Relevance - Users benefit in their decision or action because of it • Timeliness - information is available in time • Accuracy - information is free from error • Accessibility - information is readily available • Comprehensiveness - information is free from omissions and redundant data • Format - effectiveness with which information is perceived In order to perform a specific task within the operator group and to make the information flow more efficient, the operator will need information on a specific information level, as shown in Figure 6.

Figure 5: Operator roles and work tasks in an assembly system The operators’ roles in this paper are a mix between the evaluation matrix and the work tasks in the automatic assembly system. Together they form a 15 point list, shown in Figure 5.

6

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

Designing proactive assembly systems (ProAct) – Criteria and interaction between automation, information, and competence Level of Automation (LoA). The technical system, which could be connected to mechanical LoA and in some cases LoI in terms of technical solutions (information carriers).

Figure 6: Why, What and How in the abstraction hierarchy, from [17], edited by [2] The information must be able to answer three questions; what should be used (work function), why the information is relevant, this can be seen both as a goal for a function at a lower level, and the third question how the task is realized as a means for a function at a higher level [53]. This does not imply a removal of detailed information about the product or material flow; rather information is added on a higher level in terms of describing the co-functioning of the different tasks and information flow between different resources [2, 54]. The information flow to and from the assembly unit as well as within the assembly unit was identified by means of a process map. The process map displayed the way work was currently done including information support. Furthermore, it revealed with whom the operator exchanged information and also the source of information. A process map identifies, in sequence, the steps that a department performs to transform inputs into outputs for a specific process [55]. The process map was the basis for the analysis to identify the information required to enable extension of the decision latitude of the assembly operator in order to facilitate or support proactivity. The information flow was also categorised on operation level based on four levels of information; 1. Insufficient 2. Sufficient for expert operator 3. Sufficient for novice operator 4. Too much 3 Analysis towards proactivity The socio-technical school [42, 43] could be seen as an alternative in order to expand the operator action space, see Figure 7. Further, to find the interaction between the three areas. The social system could be related to the operators’ roles, Level of Competence (LoC), Level of Information (LoI) and the cognitive

Figure 7: Action space in relation to the three areas 3.1 Operator roles in the assembly system To be able to map the current state of the operators action space a mapping of the operation roles and work tasks, as listed in Figure 5, was executed. The involvement was mapped on three different levels; full, partly or no participation in the task. Empirical data, illustrated in Figure 8 show that the operators have the main responsibility for less than 20 % of these tasks in all case companies.

Figure 8: Task allocation between different roles In most of the cases, operators had no participation in planning and maintenance (only in minor disturbance handling). Partly involvement differs from 40 to 65 % between the companies. This means that the information flow between operators and other

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

7

Fasth Å. et al. sources becomes more important in the companies with low operator involvement. Braverman wrote in 1974 [56]; “The divorce of mental and physical work reduces, on every production level, the need of workers that are directly engaged in production because this relieves these workers from time consuming mental functions, these tasks are placed in other departments within the company”. In order to create a proactive assembly system we believe that the involvement of the operators in the work tasks has to increase i.e. a remarriage of the mental and physical work of the operators. 3.2 Meta methodology, the ProAct loop The ProAct-loop, shown in Figure 9, is a Meta methodology connecting the three areas by combining the theory and methods described in previous sections. The loop is used in two iterations, the current stage mapping and the future stage analysis. The methodology was tested and validated in five Swedish production companies in 2007 and 2008. A web based tool has been developed in order to perform the methodology in a structured way, as illustrated in Figure 10.

Figure 9: Methodologies used in the ProAct Loop

3.2.1 1st Loop - current stage mapping In the current state mapping, four different types of operations were identified; pre assemblingassembling-test/control-packing. The operations were analysed in each case study, in accordance with the three areas. Figure 11 shows the average LoA levels identified in the different operations. Generally, most manual work was performed in the packing stations while the control and testing stations were the most automated. The most common reason for highly automated control and test stations where the requirement on precision i.e. adjustments of air/oil pressure or power. This is in line with one of the three areas for automation proposed by Hollnagel [12]. The information analysis contained two different mappings, process mapping of the information carriers and type of information to and from the operations [57]. The second mapping made, illustrated in Figure 12, shows what information and with whom the operator communicated with in his/her daily work.

8

Figure 10: Structure of the web based tool, 1st loop

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

Designing proactive assembly systems (ProAct) – Criteria and interaction between automation, information, and competence

Figure 11: Current stage LoA in the case studies

Figure 12: Current stage of communication between operator and other personnel [57] © King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

9

Fasth Å. et al. The arrows in Figure 12 identify one-way or twoway communication. The filled boxes indicate people who had a supporting role for operator groups [57]. 3.2.2 2nd Loop- future stage analysis The second loop of the methodology contains an analysis of to the current stage and the company‟s goal functions and triggers for change [33]. One of the results from the analysis is the Square of Possible improvements (SoPI), as shown in Figure 13. This square represents the tasks or operations´ action space in which automation solutions for the future assembly system may vary. Depending on how movement in this square is done, the effort to change to other technical solutions, is reflected on both the level of information and the level of competence.

Figure 13: Square of Possible improvements and future stage analysis In proactive assembly systems, operators are encouraged to take more responsibility, far beyond managing operation and disturbances. Requirements on frequent reconfiguration, either initiated by explicit demands, or by changes due to proactively foreseeable problems or requirements, are normal. Correct amount, quality, and accessibility of information is required to provide the assembly system and its operators with abilities to handle changes within the assembly systems local layout and disturbances during operation. This requires a large amount of information to be exchanged in a short period of time. Moreover, the purpose is to ensure that the operator‟s perspective is taken into account with respect to applicable information and future decision support systems in assembly. This means that in the future there is a need to take operator abilities and limitations into consideration, as well as the operator‟s various ways of using information and

10

making decisions in different working situations [58]. Continuous iterative development of the system design increases usability from physical and a cognitive perspective. Through cyclic development of physical and cognitive user interfaces operators may refocus their energy from comprehension of the process to problem-solving, projection of system behaviour, and refining of the system. One criterion is to support correct situation awareness for the operator team. This puts requirements on the information to have ability both to provide correct information about the system, the resources (operators and machines) available in the system and the system as a whole e.g. from higher abstraction levels, as shown in Figure 14.

Figure 14: Time perspective in the Why, What, How matrix [57] The goal for proactivity is to be on the middle diagonal. • Below the diagonal - less information about intentional features • Above the diagonal - less information about functional and material features Information also needs to be presented in a way that supports the operator team to achieve the high situation awareness in a short time. The competence of the operator team is crucial because it realises the system’s built in proactivity. Competence is not entirely replaceable with information but it could be reinforced with the right Level of Automation. Without correct competence it is not possible to achieve appropriate situation awareness in a short time. Competence includes the ability to take the right action but also to predict the consequences of the action taken. The operator has to have competence within the group to be “aware of the environment within a

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

Designing proactive assembly systems (ProAct) – Criteria and interaction between automation, information, and competence volume of time and space, and to understand their meaning and the assembly systems status in near time and in the future” [19]. 4 Discussion and conclusion The paper presents the criteria and interaction between three areas; automation, information, and competence in order to design a proactive assembly system. Through a deep understanding of this interaction and optimising the separate areas in a structured way using the proposed Meta-method (the ProAct-loop), the companies can avoid over- or under-automated systems. Also, excessive information and incorrect competence levels can be avoided. Further, companies will be able to balance the three areas and optimise where it is most needed. For example, low competence level can to some extent be compensated by higher levels in the information and/or automation area and vice versa. Furthermore, the system operators' ability to participate and take part of information within the work tasks, as illustrated in figure 4 and the different abstract levels in figure 9 may be radically increased. The operators' situation awareness will improve, thus increasing their ability to act proactively. Acknowledgement The authors want to express their deep gratitude to the Swedish Governmental Agency for Innovation Systems (VINNOVA) for funding the ProAct project. We also want to express our gratitude to our ProAct project colleagues. Further, our thanks go to the participating companies: Parker Hannifin, Global Garden Products, Bosch Rexroth, Siemens Building Technologies, Stoneridge Electronics and Electrolux Home Products Reference [1] Frese M. and Fay D., 2001. Personal Initiative: An Active Performance Concept for Work in the 21st Century, in Research in Organizational Behavior, Edited by Staw,B.M., Sutton, E.L., Elsevier Science, 133-187. [2] Bruch J., Karltun J., Johansson C. and Stahre J., 2008. Towards a Methodology for the Assessment of Information Requirements in a Proactive Assembly Work Setting, in Proceedings of 2nd Swedish Production Symposium (SPS) Stockholm, Sweden.

[3] Griffin M. A., Neal A. and Parker S. K., 2007. A New Model of Role Performance: Positive Behavior in Uncertain and Interdependent Contexts, Academy of Management Journal, 50: 327-347. [4] ProAct, “http://proact.iip.kth.se/,” 2006-2009. [5] Dencker K., Stahre J., Fasth Å., Gröndahl P., Mårtensson L. and Lundholm T. 2008. Characteristic of a Proactive Assembly System, in Proceedings of the 41st CIRP conference on manufacturing systems, Tokyo, Japan. [6] Bruch J., Karltun J. and Dencker K., 2008. A proactive assembly work setting – Information requirements, in Proceedings of the 41st CIRP Conference on Manufacturing Systems, Tokyo, Japan. [7] Sherehiy B., Karwowski W. and Layer J. K., 2007. A review of enterprise agility: Concepts, frameworks, and attributes, International Journal of Industrial Ergonomics, 37(5): 445460. [8] Bruch J., Johansson C., Karltun J. and Winroth M., 2007. Considering design demands of a proactive assembly system - A position paper, in Proceedings of the 1st Swedish Production Symposium (SPS), Gothenburg. [9] Flegel H., Jovane F., Williams D., Westkämper E., Bueno R., Richet R., Hanisch C., Groothedde R., Chlebus E. and Mendonça J. M., ManuFuture 2005. Platform-STRATEGIC RESEARCH AGENDA, assuring the future of manufacturing in Europe, in Manufuture Platform-Executive Summary, Manufuture Conference, Making it in Europe, Rolls Royce, Derby, UK. [10] Crant J. M., 2000. Proactive behavior in organizations, Journal of Management, 26: 435462. [11] Dencker K., Stahre J., Mårtensson L., Fasth Å., and Akillioglu H., 2009. PROACTIVE ASSEMBLY SYSTEMS- Realizing the potential of human collaboration with automation, Annual Reviews in Control, 33:3 IFAC/Pergamon, ISSN: 1367-5788. [12] Hollnagel E., 2003. Handbook of Cognitive Task Design. Mahwah, New Jersey: Routledge. [13] Ohno T., 1988. Toyota Production System: Productivity Press. [14] Fitts P., 1951. Human engineering for an effective air navigation and traffic control system. v research foundation report Columbus,OH: Ohio state university.

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

11

Fasth Å. et al. [15] Sheridan T. B., 1995. Human centred automation: oxymoron or common sense?, in Intelligent Systems for the 21st Century, IEEE international: Proceedings of: Systems,Man and Cybernetics, [16] Endsley M. R., 1988. Design and evaluation for situation awareness enhancement, in Human Factors Society 32nd Annual Meeting, Santa Monica, USA. [17] Parasuraman R. and Wickens C. D., 2008. Humans: Still Vital After All These Years of Automation, HUMAN FACTORS - Human Factors and Ergonomics Society, 50: 511-520. [18] Parasuraman R., Sheridan T. B. and Wickens C. D., A 2000. Model for types and levels of human interaction with automation, IEEE transactions on system, man, and cybernetics Part A: Systems and humans, 30: 286-296. [19] Endsley M. and Kaber D., 1999. Level of automation effects on performance, situation awareness and workload in a dynamic control task, Ergonomics, 42: 462-492. [20] Endsley M. and Kiris E., 1995. Out-of-the-loop performance problem and level of control in automation, HUMAN FACTORS - Human Factors and Ergonomics Society., 37: 381-394. [21] Frohm J., 2008. Levels of Automation in production systems, in Department of production system, Doctorial thesis, Gothenburg: Chalmers University of technology. [22] Frohm J., Lindström V., Winroth M., and Stahre J., 2008. Levels of Automation in Manufacturing, Ergonomia IJE&HF, 30:3. [23] Bright J., 1958. Automation and Management. Boston, USA. [24] Amber G. H. and Amber P. S., 1962. Anatomy of Automation, Prentice Hall. [25] Sheridan T. B., 1980. Computer control and Human Alienation, Technology Review, 83: 6073. [26] March R. and Mannari H., 1981. Technology and size as determinants of the organizational structure of Japanese factories, Administrative science quarterly, 26: 33-57. [27] Kern and Schumann, Das Ende das Arbeitsteilung Verlag Beck, 1985. [28] Billings C., 1997. Aviation Automation: The search for a Human-centered approach. Mahwah, New Jersey, USA: Lawrence Erlbaum Associates. [29] Endsley M. R., 1997. Level of Automation: Integrating humans and automated systems, in 12

Proceedings of the 1997 41st Annual Meeting of the Human Factors and Ergonomics Society. Part 1 (of 2), Albuquerque, NM, USA, Santa Monica, CA, USA. [30] Satchell P., 1998. Innovation and Automation. Ashgate, London, England. [31] Groover M. P., 2001. Automation, production systems, and computer-integrated manufacturing, 3 ed. Upper Saddle River, N.J.: Prentice Hall,. [32] Duncheon C., 2002. Product miniaturization requires automation - but with a strategy, Assembly Automation, 22:16-20. [33] Fasth Å., Stahre J., and Dencker K., 2008. Measuring and analysing Levels of Automation in an assembly system, in Proceedings of the 41st CIRP conference on manufacturing systems, Tokyo, Japan. [34] Fasth Å., 2009. Measuring and Analysing Levels of Automation in assembly systems - For future proactive systems, in Product and production Development, Production systems, Licentiate thesis, Gothenburg: Chalmers University of Technology. [35] Liker J. K., 2004. The Toyota Way: 14 Management Principles from the World's Greatest Manufacturer. USA: McGraw-Hill. [36] Shigho S., 1981. A study of the Toyota production system from an industrial engineer viewpoint. New York, USA: Productivity press. [37] Stanton N. A., Salmon P. M., Walker G. H., Baber C. and Jenkins D. P., 2005. Human Factors Methods - A Practical Guide for Engineering and Design, Ashgate. [38] Fasth Å., Stahre J., and Dencker K., 2008. Analysing changeability and time parameters due to levels of Automation in an assembly system, in Proceedings of the 18th conference on Flexible Automation and Intelligent Manufacturing - FAIM, Skövde, Sweden. [39] 2003. SIS/SS/624070, Competence management systems - requirement. ICS 03.120.10, ISO, Ed. [40] Sheridan T. B., 1992. Telerobotics, automation and human supervisory control. Cambridge Massachusetts: MIT Press. [41] Stahre J., 1995. Evaluating human/machine interaction problems in advanced manufacturing, Computer Integrated Manufacturing Systems, 8:143-150. [42] Einar T. and Emery F. E., Medinflytande och engagemang i arbetet. Stockholm: Utvecklingsrådet för samarbetsfrågor, 1969.

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

Designing proactive assembly systems (ProAct) – Criteria and interaction between automation, information, and competence [43] Trist E. L. and Bamforth K. W., 1951. Some Social and Psychological Consequences of the Longwall Method of Coalgetting, Human Relations, 4:35. [44] Mårtensson L., Requirements on work organisation - from work environment design of 'Steelworks 80' to human-machine analysis of the aircraft accident at Gottröra, in Department of Environmental Technology and Work Science, Doctorial thesis Stockholm: The Royal Institute of Technology, TRITA-IMA 1995:2, pp. ISSN 1104-2656, ISRN KTH/IMA/R-95/2-SE [45] Mårtensson L. and Stahre J., 2003. Skill and role development in Swedish industry, in Proceedings of the 8th IFAC Symposium on Automated Systems Based on Human Skill and Knowledge. [46] Rasmussen J., 1983. Skills, Rules, Knowledge; Signals, Signs, and Symbols and Other Distinctions Human Performance Models, IEEE Transactions on Man, Systems and Cybernetics, 257-266. [47] Stahre J., 1995. Towards human supervisory control in advanced manufacturing systems, in Human factors engineering, Doctoral thesis, Gothenburg: Chalmers university of technology. [48] Johansson A., 1999. On optimizing manufacturing systems - a user perspective, in Dept of Human factors engineering, Doctoral thesis Gothenburg: Chalmers University of Technology. [49] Stair R. M. and Reynolds G. W., 2001. Principles of Information systems, 5th ed. USA: Course Technology - ITP. [50] Lumsden K., 1998. Logistikens grunder. Gothenburg: Lund: Studentlitteratur.

[51] Rasmussen J., Pejtersen A. M. and Schmidt K., 1990. Taxonomy for Cognitive Work Analysis. Roskilde, Denmark: Grafisk service. [52] Kehoe D. F., Little D., and Lyons A. C., 1992. Measuring a Company Information Quality, Factory 2000 - 3rd International Journal of Flexible Manufacturing Systems, 173-178. [53] Rasmussen J., Pejtersen A. M., and Goodstein L. P., 1994. Taxonomy for work analysis. Design of work and development of personnel in advanced manufacturing, New York: John Wiley & Sons, Inc 41-48. [54] Rasmussen J., 1985. The Role of Hierarchical Knowledge Representation in Decisionmaking and System Management, IEEE Transactions on Man,Systems and Cybernetics, SMC-15/2, : 234-243. [55] Rummler G. A. and Brache A. P., 1995. Improving Performance: How to Manage the White Space on the Organization Chart. San Francisco, CA: Jossey-Bass Publishers. [56] Braverman H., 1974. Labor and Monopoly Capital. New York: Monthly Review Press, Inc. [57] Bruch J., 2009. Information requirements in a proactive work setting, in Product and production Development, Licentiate thesis, Gothenburg: Chalmers University of technology. [58] Dencker K., Stahre J., Gröndahl P., Mårtensson L., Lundholm T., Bruch J., and Johansson C., 2007. PROACTIVE ASSEMBLY SYSTEMSRealizing the potential of human collaboration with automation, in Proceedings of the IFACCEA, cost effective automation in networked product development and manufacturing, Monterey, Mexico.

© King Mongkut’s University of Technology North Bangkok Press, Bangkok, Thailand

13

PAPER 6 Fässberg, T., Fasth, Å, Hellman, F., Davidsson, A., and Stahre, J (2012), Interactions between complexity, quality and cognitive automation, 4th CIRP Conference On Assembly Technology Systems, Ann Arbor, USA

Interaction between Complexity, Quality and Cognitive Automation T. Fässberg1, Å. Fasth1, F. Hellman2, A. Davidsson3 and J. Stahre1 1

Department of Product and Production Development, Chalmers University of Technology, Gothenburg, Sweden 2 Volvo Trucks, Gothenburg, Sweden 3 Volvo Cars Corporation, Gothenburg, Sweden

Today’s assembly systems are adapting to the increased mass customisation. This means shorter cycle times, more variants and a more complex environment for the operators. An industrial case study has been performed in order to describe relations between complexity, quality and cognitive automation. This article use quantitative methods to describe the complex environment. This is done in order to create a better understanding for the importance of using cognitive automation as an enabler in order to create a more competitive assembly system for the future. Assembly, Complexity, Quality, Cognitive automation, LoA

1. Introduction The future holds a more customized market. Complexity related to the increased number of product variants induced by mass customization has huge effects on the final mixed model assembly lines in modern factories. This leads to more unique products and a more complex work environment for the operator who will assemble the products, case studies show that 90 % of final assembly tasks are still performed by humans [1]. One definition of complexity is by Weaver [2] whom defines complexity as the degree of difficulty to predict the system properties, given the properties of the systems parts. Schleich means that a driver for assembly system complexity is the high variety of products and parts [3]. Similar ideas can be found by Urbanic et al. which, presents a model of complexity were quantity, diversity and content of information is direct associated with complexity [4]. The focus in this paper is the complexity related to mass customization i.e. caused by an increase of number of products and parts to assemble (increased amount of information). To meet requirements from mass customization, many assembly systems are using a mixed-model assembly approach as an enabler for the high variety of products. Although mixed model assembly is an enabler for high variety, such systems tend to get very complex as variety increase [5]. An important aspect of complexity is the “perceived complexity”. From an operator point of view this is a subjective factor such as competence and information [6]. Cognitive help tools are seen to reduce the perceived complexity by supporting competence and information. The increased task complexity in assembly needs to be handled otherwise the quality of the product and productivity in the system could be affected. In order to maintain high quality and reduce the complexity, one solution could be to consider cognitive automation for the operator e.g. technical support to know how and what to assemble and to be in situation control. An industrial case study has been executed in order to investigate the effects cognitive automation have on quality, in terms of assembly errors, in a complex final assembly context.

The aim of this paper is to: Investigate if cognitive automation can be used to increase quality in a complex final assembly context. An industrial case study has been executed to test if there is a relation between cognitive automation, quality and quantitative (objective) station complexity. 2. Case company Volvo Car Corporation manufacturers around 400 000 cars per year. The two main assembly plants are located in Gent, Belgium and in Torslanda, Sweden. In the Torslanda plant five models; V70, XC70, S80, XC90 and V60 are produced with a total volume of 136 323 cars for the year 2010. The five models are based on three different platforms. One serial flow mixed-model assembly line is used for all five models in the final assembly plant. The assembly line is divided into different line segments, which can have buffers in between. A driven conveyor line continuously paces a majority of the assembly line segments. The assembly is characterized by short cycles at each station and a high division of work. The current tact time (September, 2011) is about 66 seconds but can vary between the different line segments. To some extent subassembly lines are also used at different parts of the line. At the sub-assemblies other tact times may be used. 2.1 Selected area In order test the aim of the paper, an area of interest was selected. The area is one of the most complex in the final assembly with a very high product variety and a large number of parts. The chosen area consists of a total number of sixteen stations were seven have been studied within this project (the grey operators in figure 1 represents the chosen stations). The chosen stations are a part of the pre-assembly area for the preparation line of engines. In the line the engines are customised with correct driveshaft, cables etc. The engines assembled are used in all models and variants on the main assembly line. There

are three areas for the pre-assembly of the engines and this is the second area, Power Pack 2 (PP2).

Equation 1 is used to calculate the operator choice complexity for each of the seven selected stations. Input to the equation is the number of variants that occurs at each station and the demand for each variant based on 3835 cars produced during one week. The probability P is calculated for each variant j and for each station i and the total operator choice complexity is calculated with the entropy equation 1. The result for each station i is presents in figure 2. The unit scale in the figure is bit.

Figure 1. Selected area, total number of stations and selected stations

The layout of the pre-assembly area is organized as a serial flow assembly line without buffers between the stations. A driven assembly line conveyor paces the line. The assembly line is characterised by short tact times, currently 63,2 seconds, and a high number of different product variants and a large number of different parts. There is one operator working at each station. Both sides of the line are used resulting in that two stations can use the same line range but on different side of the assembly object. Some stations utilize both side of the line, which for instance can be due to large size components. The work organization is designed so that one team is responsible for 6-8 stations. There is one team leader within each team. The operators rotate between the stations in the team. For the stations chosen in this study a total number of two teams are involved as seen in figure 1. All of the investigated stations are considered to be complex due to the large number of variants and parts. 3. Quantitative methods used

Figure 2. Operation choice complexity result in the chosen stations

Other complexity parameters such as number of tools, parts and tasks to perform have been gathered, seen in table 1. These parameters is used as a complement to the OCC measure. The parameters have either been collected through direct observations at the assembly line or the balancing and sequencing system used at Volvo Car Corporation. Table 1. Complexity parameters related to each station

Three different measurements have been used to verify the hypothesis of this paper namely; operator choice complexity, assembly errors (quality) and cognitive automation. How these measurements have been gathered is explained in the following sections. Data have been gathered in the final assembly plant at Volvo Cars in Torslanda, Sweden during the summer and autumn of 2011.

30

8

9

10

11

13

23

Number of Parts*

15

23

20

14

18

15

12

Number of Tools

2

3

3

4

5

3

2

Number of Tasks**

14

22

26

17

15

25

17

* Sequenced parts seen as one ** Mean value of two products

3.1 Operator Choice Complexity The complexity in mixed-model assembly caused by the high levels of variety is by Hu et al. called “Operator Choice Complexity” (OCC), which concerns all choices that the assembly operator can make and the risk for error associated with these choices [5]. The measurement of complexity at each station that is used for comparison in this paper is the operator choice complexity proposed by Hu et al. [5] and Zhu et al. [7]. The model can be used to calculate a complexity measure for mixedmodel assembly lines. The complexity model is based on entropy function. A definition of the operator choice complexity that is induced by product variety is given as follows: Complexity is the average uncertainty in a random process i of handling product variety, which can be described by entropy function Hi in the following form: Hi (Pi1, Pi2, ..., PiMi) = -C∑Mi j=1PijlogPij

Station number

(Eq 1)

where Pij is the occurrence probability of a state j in the random process i, j  [1 Mi], C is a constant depending on the base of the logarithm function chosen. If log2 is selected, C = 1 and the unit of complexity is bit [5].

3.2 Assembly errors Assembly errors are discovered at control stations or directly by the operators at the assembly stations. All errors are reported by team leaders or by responsible quality personnel. The errors are connected to the product architecture. This means that even if a problem is discovered downstream from where it actually occurred it can be traced back to station, responsible team and individual operator causing the error. Errors reported to the internal quality system have been extracted for a time period of 16 weeks from the system and sorted by station. The results are presented in table 2. The errors have been cleared from errors caused by material and parts defects i.e. only assembly errors are included. The errors found had the following characteristics:

Table 2. Type and number of errors Error type

Number of errors

Percentage

Not Connected

106

30%

Incorrectly fitted

83

24%

Missing

51

14%

Not tightened

38

11%

Total

278

79 %

Assembly errors are categorized in eleven categories, were the top four categories accounts for (278) 79% of the total number (353) of errors. These categories are associated with errors were parts not have been connected properly, incorrectly assembled or that the parts are missing or not tightened correctly. Other quite common errors are that parts are loose or that e.g. plastic covers have not been dismantled. 3.3 Cognitive Automation In order to measure the cognitive level of the station a components of a method called DYNAMO++ was used. The DYNAMO++ method [8] and a concept model [9] for task allocation were developed during 2007-2009. The main aim is to evaluate and analyse changes in an assembly system due to triggers for change i.e. the company’s internal or external demands and Levels of Automation (LoA). The LoA analysis is done at working place level [10] i.e. on task, in stations [11, 12] and from an operator’s perspective. The measurement parameters used for task allocation is a seven by seven matrix [8], seen in figure 3, further developed from Frohm’s taxonomy [13].

Figure 4. Illustration of Levels of Automation (LoA) measured at all the seven stations

Results show that 62 percent (H) and 64 percent (C) were made with LoA level= (1;1) i.e. by hand and with own experience. The fact that so many tasks are done without cognitive support could have an impact on quality. Further, 25 percent (H) and 24 percent (C) is done with LoAcog =5 (often Pick-By-Light or indicators of what bit to use for the pneumatic screwdrivers). These are examples of tools, which are used to guide the operator to make a correct action and avoid errors. Examples of the different cognitive support tools (levels of cognitive automation) used at the stations The parts are presented to the operators in the material facade in bulk packages or by sequence racks (which could be seen as cognitive automation because they are sorted i.e. LoAcog=2=working order). Many different sizes of the bulk packages are used in the façade. Poka-yoke solutions such as Pick-By-Light systems, illustrated in figure 5, are used for some parts but not all, see table 3.

Figure 3. LoA matrix [14]

The LoA measure was made from direct observations and from standardised assembly instructions. An advantage of the use of two sources of information is that the standardised assembly instruction does not always correspond with the reality, which we wish to capture. Two models were assessed for each station, the most common model (C) regarding demand and the heaviest model (H) to produce regarding time. The distributions of the tasks for the two models are presented in the matrix illustrated in figure 4.

Figure 5. Pick-By-Light, an example of cognitive automation, LoAcog=5

Figure 7. Relation between operator choice complexity and assembly errors

Operators are supported in their work by screens, which show current model and variant and status of required tightening operations. The operators are also provided with feedback from the Pick-By-Lights and haptic feedback from some of the tools used. Operator instruction sheets are available at every team area gathered in binders. Both manual tools and automated tools are used in the assembly work. Table 3. Nr of tools and cognitive support used at the stations

4.2 Relation 2; between assembly errors and cognitive (and physical) automation The station that sticks out is station 11, if assembly errors had a correlation with OCC the anticipated number errors found would have been approximately the half, why is it so high? Due to the fact that over 60 percent of the tasks are done with own experience and that “incorrectly fitted” and “not connected” has the highest assembly errors in the further investigated stations (11,13 and 23), a summary of the assembly errors is shown in table 4, could be an indicator that there is a need for more cognitive support within these stations.

Station

30

8

9

10

11

13

23

Number of Tools

2

3

3

4

5

3

2

Number of PBL

7

15

13

0

6

3

5

Table 4. Summary of errors at station 11, 23 and 33

Sequenced articles

0

0

0

1

1

1

1

Error Code

4. Relations between the three areas In order to answer the hypotheses an investigation between four relations (illustrated in figure 6) has been done and is discussed in following sections.

Incorrectly fitted Not Connected Not Tightened Missing Total

Nr. of errors (station 11) 4

Nr. of errors (station 13) -

Nr. of errors (station 23) 54

Total nr of errors 58

38

10

5

53

2

-

19

21

14 60

10

5 91

19 161

Station 11 A total of 241 assembly errors were found during the investigated time period. 181 assembly errors were excluded due to that these errors were associated with errors from a supplier and not the assembly operation. Leaving the total number of errors for the investigated time period to 60 errors. 63% (38) of the errors were classified as “not connected” and one single part and task accounted for 38 % (23) of the total errors. The LoA of this specific task was (1, 1)1. Meaning that the operation was performed without any support. Figure 6. Overview of the three investigated areas

4.1 Relation 1; between operator choice complexity and assembly errors The first relation between the operator choice complexity and the assembly errors is illustrated in figure 7. As seen there is a relation between the lowest complexity and the lowest number of assembly errors (Station 13) and vice versa (Station 23). Station 11 differs the most from the pattern that the assembly errors follows the measure of OCC. Therefore these three stations have been further compared in the other relations.

Station 23 A total of 91 assembly errors were found during the investigated time period. 60% (54) of the errors were classified as “incorrectly fitted”. One single part and task accounted for 51 % (47) of the total errors. The part was either placed in wrong position or missed. The LoA of this specific task was (1,1). Meaning that the operation was performed without any support. Station 13 A total of 10 assembly errors were found during the investigated time period. They were all classified as “not connected”. The low error rate at this station could be explained by that most operations at the station were associated with a high LoA. Part assurance was made with a hand scanner and tightening operations were counted by the system to match the number of tasks supposed to be performed. 4.3 Relation 3; between operator choice complexity and cognitive automation The choice complexity is directly influenced by the choices and variance of solutions. An increased number of models, parts, tools etc. will result in an increase of choice complexity. Table 5 1

Not observed during the LoA assessment assessed afterwards

shows the number of variants and the demand of the most common variants. The choice complexity measure cannot directly be reduced by cognitive automation. However, introducing cognitive automation can reduce the perceived complexity caused by the increased choice complexity.

program Production Strategies and Models for Product Realization. This work has been carried out within the Sustainable Production Initiative and the Production Area of Advance at Chalmers. The support is gratefully acknowledged.

Table 5. Complexity elements

References

Objective Complexity Elements OCC Number of tools Number of variants Demand for each variant

Station 11

Station 13

Station 23

3,9 5

3,8 3

4,5 2

31

27

51

8 variants accounted for 77 percent

9 variants accounted for 78 percent

6 variants accounted for 51 percent

At the investigated stations decision support is given by Pick-By-Light and process support is given by monitors and tools associated with tightening operations. Tightening tasks are easy to control and restrict while assembly operation, which is done without any use of a tool, are very hard to monitor and control. Many manual tasks on the stations were to connect electrical connections. However neither decision nor process support was given when performing contact operations. The information regarding these operations was to be found in binders at the stations. If the contact operations are missed or badly performed the error is not acknowledged until on later control stations while tightening operations are controlled within the station boundaries by a control system connected to the tools. 4.4 Relation 4; is there a relation between cognitive automation, quality and quantitative (objective) station complexity? Earlier empirical results [15] show that in general, system complexity, does affect performance negatively and that training and that man/machine interface plays important roles in minimizing the negative effect of system complexity on performance. Results from previous sections show that relations could be made between quality, complexity and cognitive automation. Believes are that cognitive automation can be used as a mean to reduce the negative effects of choice complexity in terms of quality. 5. Conclusion This paper shows that it is possible to use quantitative measures in order to show relation between station complexity, quality and cognitive automation. These methods could be further used in order to improve both the resource efficiency and resource allocation in order to get an effective assembly system. Then, the operators’ competence and experience should also be taken into consideration, which is not fully covered by using the three methods. The main conclusion is that there is evidence that cognitive support is needed in final assembly to minimize the negative effects of complexity. Acknowledgment The authors like to express their gratitude to all involved personal at the case study and the collages within the COMPLEX project. This work has been funded by Vinnova within the

[1] Fasth Å, Stahre J. Does Levels of Automation need to be changed in an assembly system? - A case study. Proceedings of the 2nd Swedish Production Symposium (SPS). Stockholm, Sweden, 2008. [2] Weaver W. Science and Complexity. American Scientist 1948;36:536. [3] Schleich H, Schaffer L, Scavarda F. Managing Complexity in Automotive Production. 19 th International Conference on Production Research, 2007. [4] Urbanic RJ, Waguih H, ElMaraghy HA. Modeling of Manufacturing Process Complexity, 2006. [5] Hu SJ, Zhu X, Wang H, Koren Y. Product variety and manufacturing complexity in assembly systems and supply chains. CIRP Annals - Manufacturing Technology 2008;57:45-48. [6] Gullander P, Davidsson A, Dencker K, Fasth Å, Fässberg T, Harlin U, Stahre J. Towards a Production Complexity Model that Supports Operation,Re-balancing and Man-hour Planning. Proceedings of the 4th Swedish Production Symposium (SPS), 2011. [7] Zhu X, Hu SJ, Koren Y. Modeling of Manufacturing Complexity in Mixed-Model Assembly Lines. Journal of Manufacturing Science and Engineering 2008;130. [8] Fasth Å, Stahre J, Dencker K. Measuring and analysing Levels of Automation in an assembly system. Proceedings of the 41st CIRP conference on manufacturing systems Tokyo, Japan, 2008 [9] Fasth Å, Stahre J. Concept model towards optimising Levels of Automation (LoA) in assembly systems. Proceedings of the 3rd CIRP Conference on Assembly Technologies and Systems. Trondheim, Norway, 2010. [10] Westkämper E. Digital Manufacturing in the global Era. Proceedings of the 3rd International CIRP Conference on Digital Enterprise Technology. Setúbal, Portugal, 2006. [11] Nyhuis P, Kolakowski, M. H, C. L. Evaluation of Factory Transformability. 3rd International CIRP Conference on Reconfigurable Manufacturing. Ann Arbor, USA, 2005. [12] Wiendahl H-P, ElMaraghy HA, Nyhuis P, Zäh MF, Wiendahl H-H, Duffie N, Brieke M. Changeable Manufacturing Classification, Design and Operation. Annals of the CIRP 2007;56:783-809. [13] Frohm J, Lindström V, Winroth M, Stahre J. Levels of Automation in Manufacturing. Ergonomia IJE&HF, 30:3 2008. [14] Fasth Å, Lundholm T, Mårtensson L, Dencker K, Stahre J, Bruch J. Designing proactive assembly systems – Criteria and interaction between Automation, Information, and Competence. Proceedings of the 42nd CIRP conference on manufacturing systems Grenoble, France, 2009 [15] Guimaraes T, Mårtensson N, Stahre J, Igbaria M. Emperically testing the impact of manufacturing system complexity on performance. International Journal of Operations & Production Management 1999;19:1254-1269.