A Fuzzy Logic Based Approach to Explore Manufacturing System ...

Available online at www.sciencedirect.com

ScienceDirect Procedia CIRP 41 (2016) 3 – 8

48th CIRP Conference on MANUFACTURING SYSTEMS - CIRP CMS 2015

A fuzzy logic based approach to explore manufacturing system changeability level decisions Emmanuel Francalanzaa,*, Jonathan C. Borga, Carmen Constantinescub a

CERU, Department of Industrial and Manufacturing Engineering, Unviersity of Malta, Msida, MSD2080 Malta b Digital Engineering Group, Fraunhofer IAO, Nobelstraße 12, 70569 Stuttgart, Germany

* Corresponding author. Tel.: +356-2340-2060; E-mail address: [email protected]

Abstract In order to meet customer requirements, manufacturing companies need to cope with the challenge of constantly evolving product ranges. Breakthroughs in the analysis of big data will give us a better understanding of market trends and customer requirements, but there will always be some degree of uncertainty when foreseeing product evolution. Manufacturing system designers have to therefore take decisions under uncertainty, whilst considering other factors such as business, manufacturing and change strategies. In order to address this problem the factories need to be designed with a level of changeability that will allow them to be flexible or reconfigurable to produce future product platform evolutions. This paper thus contributes a novel fuzzy logic based approach to support manufacturing system designers in exploring changeability level decisions. Using this approach, this paper presents the system architecture used for an experimental implementation of an intelligent ICT tool for supporting the design of a changeable manufacturing system. © 2015 2015 The The Authors. Authors.Published PublishedbybyElsevier ElsevierB.V. B.V. This is an open access article under the CC BY-NC-ND license Peer-review under responsibility of the Scientific Committee of 48th CIRP Conference on MANUFACTURING SYSTEMS - CIRP CMS (http://creativecommons.org/licenses/by-nc-nd/4.0/). 2015. Peer-review under responsibility of the scientific committee of 48th CIRP Conference on MANUFACTURING SYSTEMS - CIRP CMS 2015 Keywords: Manufacturing system; Changeability; Fuzzy logic

1. Introduction As argued by Westkämper [1] factories are complex and long life products. The inherent nature of these long life systems means that they need to be capable of producing different product ranges throughout their life cycle [2]. Changing manufacturing systems and product requirements can trace their origin in customer behavior and evolving product requirements. This dynamic nature of customer requirements has been described as a constantly moving target [3], thus presenting a significant challenge for several aspects of product development. This means that when designing a new manufacturing system (MS) the factory planner has to ideally consider and analyze not only the current product range requirements but also how the products may evolve over time. In order to meet this challenge, methods and techniques have been developed which allow us to get an insight, or derive some knowledge, on how products and customer requirements will evolve over time. One can for instance use Foresight exercises [4] in order to construct improved planning models for new

products based on market scenario planning, product feature analysis and technological development analysis. Whilst still in its infancy big data analysis is also being explored for extracting insights into discovering what products to develop and send to production [5]. 1.1. Uncertainty in Changeable MS Design Breakthroughs in the analysis of big data and the use of foresight exercises will give us a better understanding of market trends and customer requirements. However, because of the difficulty in forecasting customer requirements, there will always be some degree of uncertainty when foreseeing product evolution. Changeable MS designers have to therefore take decisions under uncertainty, whilst considering other factors such as business, manufacturing and change strategies. One approach to address this dynamic problem is to employ changeable manufacturing systems. An important requirement for modern factories is the need to deal with product families and their evolution over time. In order to deal with these

2212-8271 © 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the scientific committee of 48th CIRP Conference on MANUFACTURING SYSTEMS - CIRP CMS 2015 doi:10.1016/j.procir.2015.12.011

4

Emmanuel Francalanza et al. / Procedia CIRP 41 (2016) 3 – 8

challenges, MS designers have to resort to changeability. Wiendahl et al [6] define changeability as “the characteristic to accomplish early and foresighted adjustments of the factory’s structures and processes on all levels to change impulses economically”. In order to develop and deploy changeability in industry, the degree or extent of changeability needs to be established and designed into the manufacturing system solution. 1.2. Motivation As explained by Francalanza et al in [7] during the synthesis decision making activity of changeable manufacturing systems, factory planners do not only have to select machines, layouts and services but they also need to select the changeability level and change enablers to implement. It can therefore be concluded that deciding this change level, i.e. how changeable will the MS be, and what type of changeability to adopt (Transformability, Flexibility, Reconfigurability), is a critical aspect of designing changeable manufacturing systems. The question therefore arises on how to select the change level and enablers required especially under uncertain conditions. In this paper, we argue that such decisions are based on (i) the change strategy adopted by the manufacturing company, and (ii) on the uncertainty of product evolution. Thus MS designers have to be effectively supported in selecting the degree of changeability and the changeability enablers which they are going to implement [8]. Therefore, the need arises to develop tools that support the changeable MS design decision making activity taking place under uncertainty due to evolving customer requirements. The design of changeable manufacturing systems poses many challenges to the stakeholders involved. As part of the Digital Factory [9] initiative this research aims to develop a digital support tool aimed at supporting stakeholders during the MS design activities. 2. Support for Changeable MS Design This section presents a review of means developed to support the MS design for evolving products and various extents of changeability. 2.1. Product Evolution Forecasting Kivi [10] develops an approach for planning and forecasting technology product evolution and the diffusion of new product features. The approach used by this research was to first isolate the underlying phenomena that governs the evolution process, and then formulate the process at the product category, product feature, and product model levels. The data generated from this analysis can be used in strategic and operational decision making. A framework for design of reconfigurable machine tools based on directed evolution is presented in [11]. The approach predetermines how the machine requirements will evolve and then the morphology of the machine modules is determined with upgradeability and adaptability in mind.

In [12] AlGeddawy investigates the co-evolution of both product design and manufacturing capabilities. A co-evolution model is formulated and interpreted using parsimony analysis of cladograms, inspired by natural species co-evolution. The model is then used as a base for future manufacturing planning. 2.2. Changeable MS Design Azab et al [13] propose a change framework and control loop that enables companies to systematically assess the need for reconfiguration in light of market supply and demand for their products. This approach allows them to determine the extent, timing, economic viability and feasibility of contemplated changes. In [14], Taha and Rostam use a hybrid approach of fuzzy analytic hierarchy process and preference ranking organization method to select the most suitable machine from several alternatives during flexible manufacturing cell design. Similarly Abdi [15] proposes a fuzzy multi-criteria decision model for evaluating reconfigurable machines. This research investigates reconfigurable machining system characteristics in order to identify the crucial factors influencing the machine selection and the machine reconfiguration. 2.3. Evaluation of Changeability Mourtzis et al [16] make use of the Penalty of Change method to evaluate the flexibility in the design of manufacturing systems. This method is used to evaluate the capability of alternative system designs to expand and increase their capacity and production range in order to meet the demand. Zaeh [17] similarly proposes an approach for evaluating capacity flexibilities in manufacturing systems. This approach considers the implication of uncertainty in market conditions and also proposes a demand forecasting method. These methods are utilized to optimize the capacity planning of a manufacturing system. In [18], Xie et al propose a method for analysis of reconfigurability in manufacturing systems using artificial intelligence theory based on Rough Sets Theory (RST). RST gives the benefit of data analysis and knowledge discovery from imprecise and incomplete information. In Guan et al [19], and Chuu [20], a similar approach is adopted to analyze the flexibility of manufacturing systems but instead using Fuzzy Set Theory (FST) and fuzzy linguistic terms. 2.4. Research Gap The means reviewed show that there are several approaches that can be employed to consider product evolution, provide support during changeable MS design and also evaluate and analyze the MS designs developed. That said, none of these approaches provide proactive support in determining the changeability level of a MS based on the uncertainty of product evolution and the change strategy adopted by a company.

Emmanuel Francalanza et al. / Procedia CIRP 41 (2016) 3 – 8

Fig. 2. Changeability level domain Fig. 1. Changeable manufacturing system synthesis

3. A Prescriptive Approach for Changeable MS Design To develop a tool to support Changeable MS design, it was first deemed critical to gain a deeper understanding of the design activities which are involved in this process. This section describes the activities which make up the design process of a Changeable MS as illustrated in Fig. 1. This approach is based on work carried out by [7] and is founded on the basic design cycle proposed by Roozenburg et al [21]. Roozenburg’s approach describes the transformation from product requirements into the detailed description of the desired product. Since we base our arguments on Westkämper’s [1] paradigm that factories are complex and long life products, we can apply Roozenburg’s product design approach to MS design. The proposed approach emphasizes the changeable MS synthesis design activity, as this is where commitments are made on the changeability level and enablers to be implemented. 3.1. Analysis of Requirements In the first activity of the changeable MS design the designer must review the parts to be produced. The designer must analyze the range of features of the product family. This includes those features which are the same across the range and those features which are different, for example the product size. 3.2. MS Solution Synthesis The next activity in the MS design cycle is Synthesis. As explained by Andreasen [22], synthesis can be defined as the combination of components or elements to form a connected whole. It is in this activity of the design cycle that the MS designer develops solutions for the MS problems. It is also important to note that designers make commitment based on the requirements defined in the analysis activity. During synthesis of changeable manufacturing systems, consciously or not, designers make commitments in the following three domains: the changeability domain, the enabler domain and in the design element domain. We argue that during changeable MS synthesis, designers make provisional commitments in these three domains.

In order to make these decisions, the factory planner has to know both the dynamic of the product evolution and the change strategy adopted by the company. This commitment should also be based on knowledge of changeable MS design, such as knowledge of changeability enablers and changeability strategies. The result of the synthesis from the different domains of changeable MS design is a provisional design solution, i.e. a Provisional Factory Model. 3.2.1. Changeability level domain Within this domain, MS designers commit to the level of changeability to implement. As illustrated in Fig. 2, MS designers have to commit to either implementing transformability on a factory level, to the generalized flexibility paradigm of FMS or to the customized flexibility of RMS. We argue, based on [6], that in order to make these commitments the MS designers must take into consideration the change strategy that has been adopted by the company and the dynamic of the products evolution. If the MS is to be designed to manufacture the current product requirements, than it only needs to implement the necessary changeability [23]. On the other hand if the company wants to adopt an investment strategy with the aim of gaining a superior competitive advantage, then the required changeability must encompass both current and future product requirements. 3.2.2. Changeability enablers domain The range of change enabler options available to the designer is dependent on the previously chosen changeability level. If the designer commits to developing a RMS then the designer has the option to commit to the six types of RMS enablers. 3.2.3. MS elements domain For the purpose of deriving clearly defined factory elements that are elements of changeability, Nyhuis [24] uses the classification of means, organization and space. Schuh et al. [25] use the classification for factory elements of resources, processes and organization. Since this research is strictly focusing on the technical and not on social aspects of changeability, MS elements are classified as layout, machine, material handling, and services. The range of options to be considered by the designer may also

5

6

Emmanuel Francalanza et al. / Procedia CIRP 41 (2016) 3 – 8

be constrained depending on the changeability enablers previously committed. 3.3. MS Solution Simulation The next activity of the design cycle involves the Simulation of the provisional design solution. Simulation involves the generation of an artificial history of a system and the observation of that history to draw inferences concerning the operating characteristics of the real system. The result of this study is the expected properties of the provisional MS design solution. 3.4. MS Evaluation During Evaluation of the provisional MS design solution the expected properties are compared to the design criteria established during the analysis stage. A value is then given to that design solution to quantify how well the provisional solution meets the product and business requirements. 3.5. Final Decision and Implementation The MS designer will then decide whether to continue developing the solution. The design cycle may be repeated to elaborate further the provisional design or to try a different type of solution to generate a better design proposal. Once the manufacturing system designer is satisfied that the provisional design meets the criteria, then the status will be upgraded to that of final design and the project can move on to implementation planning.

sets on the universe X that can accommodate “degrees of Equation 2 represents the “degree of membership”. membership”ߤ஺෨ ሺ‫ݔ‬ሻ, of element ‫ݔ‬in fuzzy set ‫ܣ‬ሚ. ߤ஺෨ ሺ‫ݔ‬ሻ ‫ א‬ሾͲǡͳሿ

(2)

This approach to define the “degree of membership” of an object is particularly useful to the problem of selecting MS changeability level. The reasoning employed is based on expert knowledge which can be translated into rules. These rules consist of terms such as necessary, sufficient and competitive, that lead to linguistic uncertainty. Fuzzy Logic hence allows us to handle the fuzziness associated with the linguistic uncertainty in describing product evolution, change strategy and changeability. 4.2. Fuzzy Sets Fuzzification is the process by which we convert crisp data into fuzzy sets. Triangular and Trapezoidal membership functions were used in order to describe the fuzzy sets used by the FLS. In the case of our MS design approach the fuzzy terms used to describe product evolution are None, Maybe and Certainly as illustrated in Fig. 3. The fuzzy terms used to describe change strategy are Necessary, Sufficient and Competitive and are illustrated inFig. 4. The fuzzy terms used to describe Changeability Level are Outsource, Flexible and Reconfigurable and are illustrated in Fig. 5

4. Intelligent Support for Changeable MS Design To address the identified gap, in this paper we contribute an approach based on intelligent computing with the aim of proactively supporting the MS designer in the selection of MS changeability level. Such an approach will allow designers to proactively foresee [26] and explore consequences of their selection decisions. In order to meet this aim, we have developed a solution based on a Fuzzy Logic System (FLS). Fuzzy Logic provides a means to deal with vagueness resulting from the utilization of natural language.

Fig. 3. Product evolution fuzzy set

4.1. Membership Functions When proposing the idea of Fuzzy Logic, Zadeh [27] proposed that set membership is the key to decision making when faced with uncertainty. Classical sets contain objects that satisfy precise properties of membership. As illustrated in equation 1, for crisp sets, an element x in the universe X is either a member of some crisp set A or not. ͳǡ‫ܣ א ݔ‬ ܺ஺ ሺ‫ݔ‬ሻ ՜ ቄ (1) Ͳǡ‫ܣ ב ݔ‬ On the other hand fuzzy sets contain objects that satisfy imprecise properties of membership. Zadeh extends the notion of binary membership to accommodate various “degrees of membership” on the real continuous interval [0,1], were 0 and 1 represent no and full membership accordingly. Fuzzy Sets are

Fig. 4. Change strategy fuzzy set

Fig. 5. Changeability level fuzzy set

7

Emmanuel Francalanza et al. / Procedia CIRP 41 (2016) 3 – 8

Manufacturing System (DMS). In the case of a DMS when future products that fall outside of the capability of the MS, then it makes strategic sense to outsource these products in the remote case that they are implemented into the product range. 4.5. Defuzzification

Fig. 6. Fuzzy inference system (Adapted from Mendel [28])

4.3. A Fuzzy Inference System As shown in Fig. 6, a rule based Fuzzy Logic System (FLS) consists of four main components for mapping crisp inputs into crisp outputs [28]. These components are the input fuzzifier, inference system, fuzzy rules and defuzzifier. The main component of the FLS is the inference system, or as sometimes referred to the inference engine. The inference system is the component that executes the reasoning in the FLS. In a FLS the inference system reasoning is accomplished by carrying out fuzzy set operations, such as fuzzy unions, fuzzy intersections and fuzzy complements. 4.4. MS Design Fuzzy Rules Core to the inference activity are the fuzzy rules. There are various ways to represent knowledge in an artificial intelligence system. One of the most common approaches it to use rules that are expressed as a collection of IF-THEN statements. These rules are expressions of natural language and are formed by two parts. The IF (antecedent or premise) part and then THEN (conclusion or consequent) part. Rules may be extracted from numerical data or formulated by experts. The following are the fuzzy rules developed during this research and used by the inference engine: x RULE 1: IF Product Evolution IS certainly OR Change Strategy IS competitive THEN changeability IS reconfigurability; x RULE 2: IF Product Evolution IS maybe OR Change Strategy IS sufficient THEN changeability IS flexibility; x RULE 3: IF Product Evolution IS none OR Change Strategy IS necessary THEN changeability IS outsource; The reasoning behind these rules is derived from the changeability and flexibility strategies proposed by [6], [23]. If the planners are sure, i.e. product evolution uncertainty is low, that that the products are going to evolve and that the change strategy is competitive, then a Reconfigurable Manufacturing System (RMS) should be designed. The MS designers would then implement reconfigurability enablers during the MS design process so that when product evolution occurs the MS can be reconfigured accordingly. On the other hand, if the planners are highly uncertain if the product mix or volume over will evolve over time, and the change strategy is to produce only the necessary product range, then the MS designer should design a Dedicated

Defuzzification is the process by which the FLS converts the output of the inference system into a crisp variable. The center of gravity (CoG) method, shown in equation 5, was employed by the prototype system as it provides an accurate result based on the weighted values of several output membership functions. ೉

‫ ܩ݋ܥ‬௑ ൌ

‫׬‬೉ ೘ೌೣ ఓಲ ෩ ሺ௫ሻ‫כ‬௫ௗ௫ ೘೔೙ ೉

‫׬‬೉ ೘ೌೣ ఓಲ ෩ ሺ௫ሻௗ௫

(3)

೘೔೙

5. Prototype Support Tool Implementation The intelligent approach described in Section 4 was implemented into a prototype digital tool for testing and evaluation purposes. In order to implement the FLS, jFuzzyLogic [29] an open-source java library was used. The user interacts with the system through an internet browser and the Graphical User Interface (GUI) illustrated in Fig. 7, web services were used in order to communicate between the GUI and the backend which was running the FLS. The user interacts with the system through the commands in the GUI. The GUI allows the user to edit the fuzzy sets, to customize the membership functions for the Product Evolution, Change Strategy and Changeability. In order to receive support, the user enters the crisp values for both product evolution and Change Strategy. The FLS then returns the defuzzified changeability extent value. 6. Evaluation The graph illustrated in Fig. 8, represents the response of the FLS to varying Product Evolution and Change Strategy inputs. At the extreme cases the graph shows that when Product Evolution and Change Strategy are low, then the Changeability Extent will be low. Likewise when Product Evolution and Change Strategy are high then the Changeability Extent is high.

Fig. 7. GUI for prototype support tool implementation

8

Emmanuel Francalanza et al. / Procedia CIRP 41 (2016) 3 – 8

[7]

[8]

[9]

[10]

[11] Fig. 8. Fuzzy response surface

The graph clearly shows that as the certainty of product evolution increases then the recommended change extent increases.

[12]

[13]

7. Conclusion [14]

This research has shown that whilst means exist to consider product evolution, provide support during changeable MS design and also evaluate and analyze the designs developed, none of these support the MS designer in determining the changeability level. This research therefore contributes an approach based on intelligent computing to guide the MS designer in selecting the changeability level of a MS based on the uncertainty of product evolution and the company’s change strategy. A digital prototype tool was also implemented to evaluate if this approach effectively supports MS designers. Future work will involve more exhaustive evaluations of this digital tool using industrial case-studies. This research will also further develop the approach to not only provide guidance on the changeability level to be employed, but to also proactively provide MS designers with the changeability enablers that can be used in order to implement the required changeability.

[20]

Acknowledgements

[21]

The first author would like to thank the University of Malta for the financial assistance provided for carrying out Ph.D. research visits to the IAO Fraunhofer, Stuttgart and acknowledge the financial support through the Research Grant “Digital Planning and Simulation for the Factory of the Future” (Vote No. IMERP 05-04).

[22]

References [1] [2]

[3] [4]

[5]

[6]

Westkämper E. Digital Manufacturing In The Global Era. Digital Enterprise Technology, 2007, p. 3–14. Schenk M, Wirth S. Fabrikplanung und Fabrikbetrieb: Methoden für die wandlungsfähige und vernetzte Fabrik [Factory Planning and Factory Management: Methods for Adaptable and Networked Factory]. Springer; 2004. Chong YT, Chen C-H. Customer needs as moving targets of product development: a review. Int J Adv Manuf Technol 2009;48:395–406. Benjamin J. C. Yuan CYL. An application of a foresight-based new product planning model: A case study of a large household appliance manufacturer in China. Chinese Management Studies 2012;6:444–61. Davenport TH. How strategists use “big data” to support internal business decisions, discovery and production. Strategy and Leadership 20140701;42:45–50. ElMaraghy HA, Wiendahl H-P. Changeability – An Introduction. In: ElMaraghy HA, editor. Changeable and Reconfigurable

[15]

[16]

[17]

[18]

[19]

[23] [24]

[25]

[26]

[27] [28] [29]

Manufacturing Systems, Springer London; 2009, p. 3–24. Francalanza E, Borg J, Constantinescu C. Deriving a Systematic Approach to Changeable Manufacturing System Design. Procedia CIRP 2014;17:166–71. Francalanza E, Borg JC, Constantinescu C. A Case for Assisting “Product Family” Manufacturing System Designers. Procedia CIRP 2012;3:376–81. Constantinescu C, Lucke D, Westkämper E. Smart Factory - A Step towards the Next Generation of Manufacturing. Manufacturing Systems and Technologies for the New Frontier, 2008, p. 115–8. Kivi A, Smura T, Töyli J. Technology product evolution and the diffusion of new product features. Technological Forecasting and Social Change 2012;79:107–26. Ahuett H, Aca J, Molina A. A Directed Evolution Modularity Framework for Design of Reconfigurable Machine Tools. In: Luo Y, editor. Cooperative Design, Visualization, and Engineering, Springer Berlin Heidelberg; 2005, p. 243–52. ElMaraghy HA, AlGeddawy T. Co-evolution of products and manufacturing capabilities and application in auto-parts assembly. Flexible Services and Manufacturing Journal 2012:1–29. Azab A, ElMaraghy H, Nyhuis P, Pachow-Frauenhofer J, Schmidt M. Mechanics of change: A framework to reconfigure manufacturing systems. CIRP Journal of Manufacturing Science and Technology 2013;6:110–9. Taha Z, Rostam S. A hybrid fuzzy AHP-PROMETHEE decision support system for machine tool selection in flexible manufacturing cell. J Intell Manuf 2011;23:2137–49. Abdi MR. Fuzzy multi-criteria decision model for evaluating reconfigurable machines. International Journal of Production Economics 2009;117:1–15. Mourtzis D, Alexopoulos K, Chryssolouris G. Flexibility Consideration in the Design of Manufacturing Systems: an industrial case study. CIRP Journal of Manufacturing Science and Technology 2012;5:276–83. Zaeh MF, Mueller N. A modeling approach for evaluating capacity flexibilities in uncertain markets. International Journal of Flexible Manufacturing Systems 2007;19:151–72. Xiaowen X, Wei X, Beirong Z. Reconfigurability analysis of manufacturing system based on rough sets. 2nd IEEE International Conference on Computer Science and Information Technology, 2009. ICCSIT 2009, 2009, p. 513–7. Jun G. Measurement of manufacturing system flexibility with fuzzy set theory. , International Conference on Management Science and Engineering, 2008. ICMSE 2008. 15th Annual Conference Proceedings, 2008, p. 713–8. Chuu S-J. Evaluating the flexibility in a manufacturing system using fuzzy multi-attribute group decision-making with multi-granularity linguistic information. Int J Adv Manuf Technol 2006;32:409–21. Roozenburg NFM, Eekels J. Product Design: Fundamentals and Methods. John Wiley & Sons; 1995. Andreasen MM. Machine Design Methods Based on a Systemic Approach, - a contribution to a design theory. Lund Institute of Technology, 1980. Gerwin D. Manufacturing Flexibility: A Strategic Perspective. Manage Sci 1993;39:395–410. Nyhuis P, Heinen T, Brieke M. Adequate and economic factory transformability and the effects on logistical performance. International Journal of Flexible Manufacturing Systems 2007;19:286–307. Schuh, G, Brussel, H Van, Boër, C, Valckenaers, P, Sacco, M, Bergholz, M, et al. A Model Based Approach to Design Modular Plant Architectures, Saarbruecken, Germany: 2003. Yan XT, Rehman F, Borg JC. FORESEEing design solution consequences using design context information. Proceedings of the Fifth IFP Workshop in Knowledge-Intensive Computer–Aided Design, 2002, p. 18–33. Zadeh LA. Fuzzy sets. Information and Control 1965;8:338–53. Mendel JM. Fuzzy logic systems for engineering: a tutorial. Proceedings of the IEEE 1995;83:345–77. Cingolani P, Alcalá-Fdez J. jFuzzyLogic: a Java Library to Design Fuzzy Logic Controllers According to the Standard for Fuzzy Control Programming. International Journal of Computational Intelligence Systems 2013;6:61–75.