An Inference-based Knowledge Reuse Framework for ...

Available online at www.sciencedirect.com

ScienceDirect Procedia CIRP 41 (2016) 472 – 477

48th CIRP Conference on MANUFACTURING SYSTEMS - CIRP CMS 2015

An inference-based knowledge reuse framework for historical product and production information retrieval Dimitris Mourtzisa*, Michael Doukasa, Christos Giannoulisa a

Laboratory for Manufacturing Systems and Automation, Dept of Mechanical Engineering and Aeronautics, University of Patras, 26500, Rio Patras, Greece

* Corresponding author. Tel.: +30-261-0 99-7262; fax: +30-261-0 99-7744. E-mail address: [email protected]

Abstract The reuse of past knowledge constitutes a key factor for improving manufacturing performance, during design, planning, and operational phases. However, valuable knowledge generated and associated to products and processes on a daily basis, remains implicit and its reusability is confined to a specific machine operator or within legacy IT databases. The work proposed in this paper introduces a framework to enable reusability of manufacturing knowledge through inference rules applied on manufacturing ontologies. Initially, a mechanism is used to access a Relational Database Management System (RDBMS) and through a rule-based approach, it exports the analogous ontological schema. The ontology constitutes the knowledge base and is stored in a knowledge repository. An inference engine is then used to query this repository and derive additional assertions, which are entailed from the base schema, on product, process, and resource. The method is developed in a web app, which offers to the users the capability to create, activate, and share rules and results on engineering topics. The effectiveness of the framework is validated using a real industrial case from a high-precision mould-making SME. © 2015 2015 The © The Authors. Authors. Published Publishedby byElsevier 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: Knowledge reuse; Inference; Manufacturing

1. Introduction In modern manufacturing, the reuse of expert human knowledge constitutes a key factor for improving manufacturing performance, during the design, planning, and operational phases [1, 2]. Indeed, human operators still play the most important role in most product and production lifecycle stages. Yet, their expertise and knowledge lies in tacit form with them and is not sharable. Valuable knowledge generated and associated to products and processes on a daily basis, remains implicit and its reusability is confined to a specific machine operator and domain [3]. In this research work, a framework for capturing, storing, and retrieving knowledge from already existing engineering projects is proposed. The developed framework is capable to parameterize and create new knowledge through user-defined customizable semantic rules and queries. The framework consists of a mechanism for the conversion of a relational database into an ontological schema, a knowledge repository, and an inference engine. It offers to the users the capability to

create and apply rules and queries on the repository and retrieve relevant knowledge. Rules created are stored and remain available for anyone to access, edit, and use. For instance, during the design phase of a new injection mould, an engineer will be able to infer knowledge from past cases that meet the new mould’s requirements and reuse it after parameterization based on the custom new needs. Consequently, the design phase is facilitated and can be completed faster and more accurately. 2. State of the Art 2.1. Product Design Knowledge repositories and reasoning mechanisms are employed to assist knowledge reuse about product design and development at various stages of the product lifecycle [4, 5, 6]. In [7], a methodology for knowledge management to support concurrent engineering is developed. A knowledge classification scheme is created, and a framework is

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.2015.12.026

Dimitris Mourtzis et al. / Procedia CIRP 41 (2016) 472 – 477

developed that utilizes agents to model interaction among key persons and knowledge repositories. A set of methods for knowledge-based design of machine tools is described in [8]. Knowledge is related to product data, which consists of the data generated in various applications as well as metadata that contain control information about the product. In [9], a methodology is developed that addresses engineering changes during the product design phase using a knowledge-based method to resolve design conflicts. The method studies changes in the physical, structural, and functional domains as well as the interdependence between them. A matrix-based method is proposed to synthesize the relationships between the three domains and provides an effective way to capture design changes. In [10], a knowledge-based methodology is developed as a link between existing product design capabilities ('what we have') and requirements ('what they want'). Furthermore, in [11], a comparison of a case study of practices for reusing manufacturing knowledge is performed and the improvement of existing knowledge reuse practices is presented based on three different business areas. An integrated knowledge management and reuse framework for Product-Service Systems business in construction machinery industry is presented and its contribution to Product-Service Systems design is highlighted in [12]. In [13], a novel approach for capturing the service damage knowledge about each component is presented. Another industrial case study is presented in [14], creating new knowledge on the relationships between degradation and component design to manufacture. Based on service experience and event data, the identification of service risk drivers is achieved. In [15], a framework to integrate requirements management and design knowledge reuse is presented and demonstrated on a case from vacuum pump design, and enables the application of requirements management as a dynamic process. It takes into account the changes in requirements that may occur between the stages of product development. Another study presented an effective reuse and retrieval system that eases the registration of parts [16]. The system also generates optimized parts and consists of three modules. The first offers reuse and modification capabilities to the designers, the second offers the automatic generation of parts and the last module enables the retrieval and classification of standard parts. 2.2. Factory planning and operation Knowledge repositories and reasoning mechanisms have been also employed to assist factory planning and operation stages as well. In [17], a knowledge system for lean supply chain management (KSLSCM) is developed using artificial intelligence system shells. The combination of the AI shells provides a development platform with a logic programming function and visual presentation. In [18], a model is proposed to discover knowledge embedded in the relationship between product features and the corresponding manufacturing capabilities. The model aims at capturing designers’ knowledge in order to encapsulate the relationship between product features, managing and preserving acquired systems design knowledge and automatically generating the required systems to produce new products. In [19], a knowledge

management framework is described for supporting a factory throughout its lifecycle, from design to decommissioning. The purpose is to model existing data with the use of Semantic Web technology in order to extract useful knowledge. The knowledge framework was based on three parts, the ontology, the knowledge repository, and the knowledge association engine. An advantage of this platform is the versioning system for the data, which provides a set of methods for versioning management and structured data retrieval. In [20], a system for distributed knowledge sharing and reuse in the PFMEA domain is developed using an ontology based on description logics allowing knowledge inference and retrieval in manufacturing environments. 2.3. Benefits of the proposed framework The proposed knowledge management framework can promote the reuse of the product and production related data regardless of domain. It can support the easy discovery of complex data structures and it is web-based, which means that every user can have access to its services. Moreover, unlike previous work on the field, it provides a generator with the capability of automatic transformation of any relational data model to an ontology schema, enabling that way the operability of this specific framework with any current of future case study instantiation. This effectively diminishes any required ontology modelling expertise and can work with any legacy database system. Finally, the ability for an engineer to form queries and knowledge-generating rules without any programming skills is considered as a necessary function towards enabling and simplifying accessibility to knowledge. Thus, a user friendly editor is developed to facilitate the composition of complex queries and rules that can be executed by the engineer. 3. Methodology and Framework Description The proposed framework includes a set of standalone components, each being responsible for a set of functions. The modular architecture allows a decoupling of the components and individual operability. The components are: the ontology generator, the ontological schema, and the vocabulary that are used during knowledge capturing, the rule and query builders that are used by the reasoning engine, and the graphical user interfaces that expose the services to the end-users (Fig. 1). Multiple instances of industrial cases can be accommodated by the framework and be accessed by unique user credentials. Moreover, third-party software apps can connect to the framework through the exposed REST interfaces [5, 6]. 3.1. Relational Schema to Ontology Generator In 2014, 79 percent of data were stored in relational databases on the web [21] (down from 89.2 percent in 2004). Thus, the ability to automatically translate every relational schema into an ontology brings many benefits for nonexperts. The RDBMs-2-Ontology (R2O) generator is designed to access a relational database and export the equivalent ontological model and its stored instances.

473

474

Dimitris Mourtzis et al. / Procedia CIRP 41 (2016) 472 – 477

Ontology Generator Ontology

Knowledge Framework

R1, R9 Class

Vocabulary

Knowledge Capturing

Rules

Inference ce e Engine

Relations between classes

Attribute

Queries Relational Schema

Properties R2, R3, R9

Mapping

Industrial Case Instance

App1

Knowledge g Reuse e

Concept Classes

App… App2

Fig. 1. The proposed Knowledge Management Framework.

This development alleviates the need for expert knowledge on ontology modelling and reduces the overhead effort for the IT department to develop and test an accurate ontology based on their legacy data structure. As referred to in [22], there are certain specific and well-defined rules to generate an ontology out of a database. These rules, adapted for our purposes, are presented shortly hereafter: R1. Every table from the database is transformed into an ontology class.

Ontology

Data Properties

R4-R9

And / Or

Object Properties

Fig. 2. Mapping between Relation Ontological Schemas using rules R1-R9.

3.2. Ontological Schema The derived ontology, in the manufacturing domain, defines information about product, process, and resource entities. It accommodates all structural relationships and physical properties needed to explicitly describe essential entities in the examined domain and comprises the knowledge repository.

R2. Every attribute that is a primary key (PK) of the table, is transformed into a functional property.

3.3. Inference Engine

R3. Every attribute that is neither a Primary Key (PK) nor a Foreign Key (FK) of each table in the database system is transformed in a data-type property with the same name as the attribute.

The Inference Engine (INE) enhances the developed framework with the semantic technology. The INE includes a reasoner that allows the manipulation of the knowledge that is stored inside the ontology and the knowledge repository. The reasoner supports SWRL rules, which stands for Semantic Web Rule Language. SWRL combines OWL and RuleML technologies. The reasoner provides also the ability to directly load a file that contains SWRL rules and rules will be parsed and processed. Finally, SWRL rules can be mixed with OWL axioms in an ontology. INE is developed as a web app with server / client sides. The client is accessible from the Graphical User Interface (GUI) of the application (Fig. 4 and Fig. 6). The server side is the part where all the services of the application are implemented. The user is able to select one of the available repositories that have been created from ontology generator component. Depending on the account privileges and access rights, the user will be able to access the repositories available. An administrator has access to all available repositories.

R4. Every attribute that is a foreign key (FK) of the table, is translated to two object properties, inverse with each other, between the tables that share the attribute. R5. If an attribute that is not a PK is set Not Null, the property’s cardinality created from the above rules (either 2 or 3) is set to 1. R6. If an attribute that is not a PK is set as Unique, the property’s max cardinality created from the above rules (either 2 or 3) is set to 1. R7. If an attribute is a FK and is set Not Null, the property’s min cardinality created from the above rule (4) is set to 1. R8. If an attribute is a FK and is set as Nullable, the property’s max cardinality created from the above rule (4) is set to 1. R9. All data and their values are transformed as individuals that receive the same values. The mapping of the analogous elements between a relational and an ontological schema is achieved using these nine rules (R1-R9) as Fig. 2 schematically shows.

3.4. Query and Rule Builder The query and rule builder is a user-friendly interface, where a user can create complex queries without requiring expertise on database management. Prefixes of the query are already pre-filled to avoid manual writing of standard strings.

Dimitris Mourtzis et al. / Procedia CIRP 41 (2016) 472 – 477

User

Knowledge Repository

select Repository

Reasoning Engine

select OWL file

create Repository return Repository ID create/ activate Rule

create Rule send/active Rule

return Rule status

Fig. 4. Screenshot of the Graphical User Interface for the Creation of tuples.

5. Case Study from the Mould Making Industry return status

Fig. 3. Sequence diagram for the rule creation / activation on a repository.

The syntax of rules and queries is demonstrated through examples in section 5 below. The sequence diagram for creating a rule on a selected repository is shown in Fig. 3. Query triples are created in a similar way (Fig. 4). 4. Software Tool Development Among the several available tools for each component, certain criteria were used for selecting the tools in this work: 1 Cost of each software tool: Open-source software was preferred so that the framework is accessible by SMEs that cannot afford commercial licensed solutions and high investment costs in IT. 2 Platform-independency: The tools were selected based also on their capacity to run on several operating systems. 3 Popularity and online community of each tool: The most widely used tools, which also provide online support and active communities were preferred. Based on these criteria, the selected software tools were: MySQL as a high-performing, free to use, secure, and userfriendly data server that supports many interfaces [23]. The OWL-API was selected to assist in generating the entities for the ontology schema created by R2O, which supports the cardinality restrictions feature of OWL2. Data storing is made using Semantic Web technology. Ontologies are developed using the Web Ontology Language (OWL). The OWL language offers benefits against other ontology languages. Also every .owl file is supported from OWL 2, the newest and richest version for ontologies. Jena Framework was selected to enable the manipulation of the instantiations of the repositories created by applying rules and queries to them. Rules are compiled using Jena Rule Language. The Java framework is used for the implementation of the web application along with Java Script open source libraries. As so, every browser can support the framework. The application is hosted on an Apache Tomcat web-server, which is free, secure, and offers many features.

As shown in Fig. 1, the framework can indistinctly accommodate industrial case instances. Here, a case study is presented from a high-precision mould making SME. The discussed functionalities of the framework shown concern its ability to enhance app with semantic technology and its ability to create new knowledge to support the entire lifecycle of manufacturing engineering projects. The mould-making sector is a crucial enabler for the realisation of mass production and customization, and moulds are the tools with the highest productivity at the disposal of a producer. The SME of the case study produced engineer-toorder moulds. These moulds are one-of-a-kind extremely complex products that are designed and manufactured in order to produce one or more plastic components with a single stroke of the press and eject it without any imperfections. Yet, as reported by the SME, new mould-making cases share a number of commonalities with previous projects on a component, processes, and specification levels. Thus, the reuse of existing knowledge on mould making projects is a target for the SME. The first step of the case study is to convert the legacy data schema of the SME into an ontology, using the R2O generator, to allow the execution of semantic rules on the data. A snapshot of the generated ontology classes and their interconnections is shown in Fig. 5. The entity Component includes the name, part / component type (top plate, core cups, etc.) and dimensional data information. The Components belong to a Mould, which is the core entity of the model. The Mould entity includes order related data (e.g. data of order creation), name and description of the mould, the customer id, the number of cavities, the cooling structure and any specific surface quality attributes. Moreover, information related to the wall thickness of the injected product, earpins, tamper evident, polishing, way and side of injection, and related Bill of Materials ID, are included. The Bill of Material ID is a reference to an external Product Data Management software suite. The Product datatype refers to the plastic product or mouldling, which is produced by the mould with each stroke. The entity includes information on surface quality, polishing, and wall thickness of the mouldling.

475

476

Dimitris Mourtzis et al. / Procedia CIRP 41 (2016) 472 – 477 range type Integer

domain

Functional Property Id Class Mould

DataTypeProperty domain earPins DataTypeProperty tamperEvident domain

domain

DataTypeProperty domain weight has DataTypeProperty domain polishing

range

type String

ObjectProperty isProductOf

domain

DataTypeProperty ejectorRings

has

ObjectProperty rootComponentId

has inverseOf

ObjectProperty ProductId type Boolean

type Double type Boolean

DataTypeProperty cavities

DataTypeProperty domain numberOfCircuits DataTypeProperty domain slides

DataTypeProperty domain wallThickness

range

domain domain

Class Component DataTypeProperty domain typeOfCircuits

Class Product

ObjectProperty isComponentOf

type Integer inverseOf has

Fig. 6. Screenshot of the Graphical User Interface of the stored rules in the "Rule Editor" component.

Fig. 5. Ontological schema of the entities Product, Mould, and Component.

After the formation of the ontology and its storage at the knowledge repository, rules and queries can be formed by the domain expert using the forms shown in Fig. 6. The purpose of these rules is the creation of new knowledge. Rules can reveal patterns in stored data, group entities according to custom characteristics that were not originally modeled in the ontology, and aid in retrieving similar stored engineering cases. The general structure of a rule respects the following template: [rule name: antecedent -> consequent]. If the individuals (instances) of the ontology satisfy the statements that constitute the antecedent part of the rule, then the statements of the consequent are added to the ontology as inferred knowledge. The first rule checks if the melted plastic is injected from the core side. The rule can reveal a pattern in the type of received and executed orders by the SME: If the rule result is true, then the mould complexity is set to “medium”: [InjectionComplexity: (?Mould1 rdf:type case:mould) (?Mould1 case:mould_wayOfInjection ?WayofInjection1) equal(?WayofInjection1, "0"^^xsd:string) -> (?Mould1 case:mould_complexity "medium"^^xsd:string)] Such a rule can be used to group mould orders according to their complexity. Upon executing the rule, the results show that 30 out of the 37 order cases are characterised by this complexity factor, revealing to the engineer that the majority of the projects that the company has undertaken during the past 2 years were complex. A second rule can help characterise the manufactured moulds based on their size. The characterisation “large mould” can be inferred by forming the appropriate rule. This attribute is not originally contained in the ontology, thus, it comprises directly inferred new knowledge.

This rule can be valuable to the warehousing department, since a mould can weight several hundreds of kg and occupy expensive storage area. Similarly, meaningful rules that were created by the engineers, using the rule and query editor, included a characterisation of moulds as “typical” (if the mouldling is a thin-walled container), and as “floating” (if the description of the order contains the word “stacks”, which translates to a mould that consists out of more than seven components). Moreover, a quantitative assessment is performed to compare the benefits brought with the application of the Knowledge Framework against the traditional empirical approach. The results are demonstrated in Fig. 7. Three daily activities of the mould-maker are examined and the comparison of the two approaches is made using time and cost indicators. The values of the indicators for the empirical approach are obtained through an interview with expert designers, management staff, and machine operators, while the same personnel tested live the software framework to obtain the values of the indicators for the new approach. The first comparison concerns the retrieval of CAD files and related metadata (tolerances and quality characteristics) needed during the design of a mould. It is typical to reuse parametric CAD drawings of mould components, after adjusting them to the requirements of a new mould. It was shown that the time for the designer to identify a similar case and retrieve its data was reduced by a relative difference of 107.7% by using the knowledge framework. Considering the cost of the person-hour of an expert designer, this is translated in cost savings of 136.1%. This is in accordance with recent studies, which depict that a designer spends 20% of his time searching and absorbing design information [24]. In a similar case the number of mould orders need to be characterised based on their size and weight characteristics (to satisfy limited storage area constraints). The benefit, against the traditional approach, was 136.1% for time and 177% for cost.

Dimitris Mourtzis et al. / Procedia CIRP 41 (2016) 472 – 477 Empirical approach vs. Knowledge Framework (KF)

1000 100

10 10 1

Cost (€)

Time (hr)

100

1 Retrieve CAD data for similar cases

Mould categorisation

Time (hr) KF Cost (€) KF

Compare moulds / similarities Time (hr) Empirical Cost (€) Empirical

Fig. 7. Cost and time benefits of the proposed Knowledge Framework (KF) against the traditional empirical approach.

Finally, when trying to identify similar past cases, the overhead for a designer, for searching handwritten documents and trying to identify similar features on mould orders cannot be avoided in the empirical approach. On the other hand, this task is automatically performed by the proposed framework after appropriate rules and queries are formed, leading to 163.6% in time savings and 188.6% in cost reductions. 6. Conclusions and Future Work This work aims to allow capturing, inferring, and reusing engineering knowledge. The developed framework provides the ability to convert any legacy database into an ontology. The ontology forms the knowledge repository, on which custom rules / queries are applied to create new knowledge. The repository and the instances are stored in a file enabling easier manipulation and portability. Moreover, the described inference engine is responsible for the actual manipulation of the knowledge stored. The benefits of implementing the framework to a mould-making SME are demonstrated, through the creation of new knowledge and by comparing time and cost indicators during searching, retrieving and reusing engineering information from past similar projects. Future work will focus on the design and implementation of a linguistics component. A vocabulary will be created that maps engineering terms in different languages to allow the collaboration between designer teams of different origin. Acknowledgements The work reported in this paper is partially supported by the EU funded research project “Applications for Advanced Manufacturing Engineering – Apps4aME” (GA No. 314156). References [1] Chryssolouris G, Mavrikios D, Papakostas N, Mourtzis D, Michalos G, Georgoulias K. Digital Manufacturing: History, Perspectives and Outlook, Special Issue Paper 2008; 222/5:451-462 [2] Karmarkar US. Manufacturing Lead Times, Order Release and Capacity Loading. Handbooks in Operations Research and Management Science 1993; 4: 287-329. [3] Efthymiou K, Sipsas K, Mourtzis D, Chryssolouris G. On an Integrated Knowledge-based Framework for Manufacturing Systems Early Design Phase. Procedia CIRP, 2013; 9: 121-126.

[4] Mourtzis D, Doukas M, Fragou K, Efthymiou K, Matzorou V. Knowledge-based Estimation of Manufacturing Lead Time for Complex Engineered-to-order Products. Open-Access Journal Procedia CIRP, 47th CIRP Conference on Manufacturing Systems 2014; 17: 499-504. DOI: http://dx.doi.org/10.1016/j.procir.2014.01.087 [5] Mourtzis D, Doukas M, Vlachou K, Fragou K, Vandera C. Knowledge Enriched Short-term Scheduling for Engineer-to-order Products. Procedia CIRP, 2nd Robust Manufacturing Conference (RoMaC 2014); 19: 160167. DOI: http://dx.doi.org/10.1016/j.procir.2014.05.012 [6] Mourtzis D, Doukas M. Knowledge Capturing and Reuse to Support Mass Customisation: A Case Study from the Mould Making Industry. 24th CIRP Design Conference 2014; 21: 123-128 [7] Wang CB, Chen YM, Chen YJ, Ho C.T. Methodology and system framework for knowledge management in allied concurrent engineering. International J of Computer Integrated Manufacturing 2005; 18/1: 53-72. [8] Naca J, Bueno R, Alzaga A, Kovacs GL. Knowledge Management Support for Machine Tool Designers Using Expert enablers. International Journal of Computer Integrated Manufacturing 2007; 18/7: 561-571 [9] Fei G, Gao J, Owodunni O, Tang X. A Method for Engineering Design Change Analysis Using System Modelling and Knowledge Management Techniques. International Journal of Computer Integrated Manufacturing 2011; 24/6: 535-551 [10] Liang JS. Development of a knowledge-based assisted product design system. International Journal of Computer Integrated Manufacturing 2012. DOI: 10.1080/0951192X.2013.857125 [11] Krogstie L, Andersson P. A Case Study on Reuse of Manufacturing Knowledge - Comparing Defense Practices with Automotive & Aerospace Practices. Procedia CIRP 2012; 3:430-435 [12] Zhang D, Hu D, Xu Y, Zhang H. A Framework for Design Knowledge Management and Reuse for Product-Service Systems in Construction Machinery Industry. Computers in Industry 2012; 63/4: 328-337. [13] Roy R, Mehnen J, Addepalli S, Redding L, Tinsley L, Okoh C. Service Knowledge Capture and Reuse. Proceedings of the 6th CIRP Conference on Industrial Product-Service Systems 2014; 16:9-14. [14] Morar N, Roya R, Mehnena J, Redding LE, Harrison A. Data Mining and Knowledge Reuse for the Initial Systems Design and Manufacturing: Aero-engine Service Risk Drivers. 2nd International Through-life Engineering Services Conference 2013; 11:130-134. [15] Baxter D, Gao J, Case K, Harding J, Younc B, Cochrane S, Dani S. A Framework to Integrate Design Knowledge Reuse and Requirements Management in Engineering Design. Robotics and Computer-Integrated Manufacturing 2008; 24/4:585-593. [16] Moka H, Kim CH, Kim CB. Automation of Mold Designs with the Reuse of Standard Parts. Expert Systems with Applications 2011; 38. [17] Liu S, Leat M, Moizer J, Megicks P, Kasturiratne D. A Decision-focused Knowledge Management Framework to Support Collaborative Decision Making for Lean Supply Chain Management. Interntional Journal of Production Research 2012; 51/7: 2123-2137. [18] AlGeddawy T, ElMaraghy H. Manufacturing System Synthesis Using Knowledge Discovery. CIRP Annals 2012; 60/1:437-440. [19] Efthymiou K, Alexopoulos K, Sipsas P, Mourtzis D, Chryssolouris G. Knowledge Management Framework Supporting Manufacturing System Design. 7th International Conference on Digital Enterprise Technology, Athens, Greece (DET 2011), pp. 577-585. ISBN 978-960-88104-2-6 [20] Mikos WL, Ferreira JCE, Botura PEA, Freitas LS. A System for Distributed Sharing and Reuse of Design and Manufacturing Knowledge in the PFMEA Domain Using a Description Logics-based Ontology. Journal of Manufacturing Systems 2011; 30: 133-143. [21] Infoworld. Not so fast SQL, SQL still reigns 2014. URL: http://www.infoworld.com/article/2607910/database/not-so-fast--nosql---sql-still-reigns.html [22] Zhou S, Ling H, Han M, Zhang Hu. Ontology Generator from Relational Database based on Jena. Computer and Information Science 2010; 3/2. [23] Trujillo G. MySQL versus Oracle Features/Functionality 2008. URL: https://blogs.oracle.com/GeorgeTrujillo/entry/mysql_versus_oracle_featur es_functionality [24] Lowe A, McMahon C, Culley S. Information Access, Storage and Use by Engineering Designers - Part 1. Engineering Designer. March 2004; 30-2

477