Djordje VUKELIC Branko TADIC Janko HODOLIC Nenad SIMEUNOVIC
Abstract: The problem of fixture design, as one of the pivotal elements which significantly influence the quality of process planning in manufacturing engineering, remains topical, especially in a highly automated, flexible manufacturing environments. Computer-aided fixture design has, for a number of years now been in the focus of a large number of researchers. The results of this research have allowed a systematized knowledge of the possibilities offered by computer-aided fixture design. Emphasized in this paper is the importance of automation in fixture design. Characteristic approaches to fixture design automation are reviewed and analyzed. Basic features, advantages and disadvantages of particular approaches are examined. The paper concludes with future directions of research.
Key words: fixture, design, automation, CAFD
1. INTRODUCTION
During the last couple of decades the computers have been increasingly used to assist design activities. The beginnings of their application date back to the sixties of the last century, when they were first successfully used to control machine tools. This was followed by an expansion of their application in various domains of manufacturing engineering. The concepts of flexible manufacturing system (FMS) and computer aided manufacturing (CIM) have evolved since. Today, the emphasis is placed on the intelligent manufacturing systems which are able to solve problems without the use of an explicit and detailed algorithm or a mathematical interpretation of the problem. Various CAx (Computer Aided Everything) systems are used to assist product design and manufacture in order to shorten the time and related costs. Numerous systems have been developed which allow automation at particular stages of design and manufacture. Computer technologies have revolutionized modern manufacturing. From the standalone CAD/CAM applications, to PDM/ERP (Product Data Management/Enterprise Resource Planning) systems, the computer technologies have fulfilled the dreams of every manufacturer shorter product development time, higher quality, and lower costs. Computer-aided fixture design (CAFD) has become reality as part of this revolution by integrating the fixture design know-how and CAD platforms. With the aim to reduce fixturing costs, over the years, various CAFD systems have been developed to assist fixture designers. Although numerous research efforts have been aimed at development of fixture design systems, there is still a need for development of a method which would assist the designer at fixture elements level, where the basic task would be to identify the required structure of fixture i.e. the optimal number and arrangement of elements according to predefined criteria. Though the primary function of a fixture is allow precise location and clamping of workpiece, there are a lot of other criteria to be satisfied, which most often pertain to ergonomic issues. Finally, one of the most important aspects of fixtures is their impact on manufacturing costs through assembly time, materials, costs of fixture manufacture, etc. Another interesting feature of fixture design is the fact that the various requirements to be met are usually opposed to one another. Fixture design and manufacture costs can account for up to 10-20% of the total manufacturing costs [30]. These costs not only pertain to fixture material, manufacture and assembly, but also encompass the costs of fixture design. Lowering of fixture design costs induces significant economic effects. There are two approaches to this goal. One is focused on the development of flexible fixtures, while the other relies on the simplification of fixture design through computer application.
2. GENERAL CHARACTERISTICS OF SYSTEMS FOR AUTOMATED FIXTURE DESIGN
Computer-aided fixture design is a more recent way of design (the first attempts at fixture design automation date back to the eighties of the last century). This approach evolved as an attempt to overcome the negative aspects of the conventional design paradigm. CAFD uses computer to automate the steps in fixture design. The goal is to generate an adequate fixture solution within an acceptable time frame, while at the same time minimizing the subjective influence and effort of the designer. The most important assumptions for efficient application of computers in the process of fixture design are the 'translation' of designer's knowledge and experience into a computer-usable form, the development of a logic used in selection and decision-making, etc [27]. In essence, the systems for automated design are based on information technologies (IT), while their structure and principle of functioning make them specialized IT systems. Regardless of methodology applied in the development of a system for automated fixture design, the constituent parts of every system are: module for defining of input information, module for generation of output information, and a data base [28]. 8 Considering the systems for automated fixture design, the input information can be broken down into three larger categories: workpiece information (shape, geometry, dimensions, etc.), technological information (machine tools, cutting tools, cutting regimes, etc.), production organization and management information (time of design, time of manufacture, etc.). Output information contain all the required data for successful assembly and exploitation of fixtures. They are always adjusted to end-users individual needs. The necessary output information include: assembly drawing of fixture and bill of materials (BoM). Very often, part drawings are also required (regardless of whether the parts are standard or not), while among the other information are: weights of particular fixture elements, total fixture weight, position in storage, etc. The data base is fundamental to software support. Its basic task is to provide system efficiency through quality and effective functioning. Its primary functions are data acquisition, storage, keeping, searching, and updating. The data base must contain all the data necessary for a successful functioning of the software support. It should contain the required information (alpha-numerical and graphical) pertaining to existing fixture design solutions, fixture elements, and workpieces for which the fixtures were designed.
3. CHARACTERISTIC APPROACHES TO FIXTURE DESIGN
During the past two decades a number of researchers have put in effort to solving the problem of rationalization, i.e. automation of fixture design. There are numerous examples of developed systems to confirm that claim. Based on available references and consultations with professionals and scientists from that area, it can be concluded that the so far fixture design automation has been developing through three main approaches: expert systems, Case Based Reasoning, and Artificial Neural Networks.
3.1. Application of expert systems in fixture design
Operation of the majority of expert systems for fixture design is based on symbolical representation and processing of built-in knowledge. The knowledge is represented using formal symbols and other appropriate data structures presented in a formal language, while the problems are solved by deriving conclusions through manipulation of these symbols and data structures. Expert systems whose knowledge is formalized through rules are often termed rule-based systems. The rules can be understood as the elements of knowledge, i.e. the quantum of knowledge pertaining to a particular domain of fixture design [15]. A rule represents a logic relation, and can be expressed as: If X Then Y
meaning: "if assumption X holds true, then Y is inferred" or "If situation is X Then action Y is in order". For instance: if "the base surface is a hole" then "the locating element is a long pin" [23]. In other words, a rule is a logical If-Then loop meaning: if a premise holds (or a set thereof), then an inference can be made (or a set thereof) or an action can be undertaken (or a sequence of actions). Such way of formalizing knowledge is very natural and is suitable for modular representation. It also allows easy modifications of the knowledge, for the new rules can be added independently from the existing ones, while the system is transparent, i.e. its inference in easy to follow. Due to their form, production rules are suitable for representing the logic- based knowledge [21,22]. The inferences are made through comparison of a set of rules with the set of facts pertaining to a given situation. If the "If" part of the rule holds true, then the action defined by "Then" is executed. When this happens it is usually said that the rule has fired. For instance: If (Premise) Machining operation =drilling (and), Batch size =single (and), Machine tool =vertical milling machine (and), Machine tool type =conventional (and), No. of machining surfaces =one (and), No. of cutting tools =one (and), Drilling diameter =three (and) Then (Inference) Guiding element =fixed bush
Such or similar production rules have been developed mostly for locating and/or clamping elements [2,6,7,9,12]. The elements are selected by re-iterating rules until satisfactory elements have been found in the data base. Should a larger number of eligible candidates be found, the designer decides upon the one, based on his/hers experience. Elements are selected based on their function. Since there are several elements with identical functions, they are grouped together. When applying the rules, the function of each element is checked. The beginnings of application of expert systems in fixture design is attributed to the works of Markus et al. [20] and Gandhi and Thompson [13]. Larger part of these researches was theoretical while pointing out at some possible ways of practical application. They marked the onset of intensive development of expert systems in the domain of fixture design. Darvishi and Gill [10], and Ferreira, et al. [11] proposed a general model for knowledge representation within an expert system. They placed emphasis on the representation of declarative knowledge. However, the rules which describe mutual relationships and functions of elements were not developed. Besides, the geometry of workpiece, its dimensions and tolerances were also disregarded. Procedural knowledge is vital to expert system operation, and for its representation production rules are most often used. Decision trees and rules networks were used to define production rules. Decision trees were first introduced by Bugtai and Young [4], and Dai, et al. [8] and they are clear and comprehensible for the user. As their main feature, they require a hierarchy of knowledge in a particular domain. When representing knowledge using rule networks, strict hierarchy is not required but, on the other hand, they are less comprehensible. Fixture design is, by nature, much closer to this way of representing knowledge. 9
Fig. 1. Fixture designed using an expert system [31]
The reason for application of expert systems is to place the available knowledge on fixture design at disposal through various application software. Expert systems cannot entirely replace fixture design experts, especially when it comes to creativity and general knowledge. However, their advantage over humans lies in the permanence of their knowledge, i.e. it does not depend on repeated use. Expert systems use existing rules for fixture design and can function relatively well as initial solutions. Nevertheless, they lack flexibility, are difficult to adjust and self-improve. In addition, there is no formalism to allow detection of possible conflicts among production rules when updating the database with new rules. For these reasons, expert systems are recommended to production systems with a well established and permanent production program.
3.2. Case-Based Reasoning in fixture design
Case Based Reasoning (CBR) is one of the approaches to intelligent database search. Every new search is based on previous similar cases, thus utilizing some sort of experience. This approach has been widely used in fixture design. CBR is a concept which comes up with new fixture solutions based on the existing ones, applying modifications where necessary. The process of reasoning can be described as follows. There is a new problem to be solved, i.e. new workpiece for which a fixture must be designed. The designer searches the knowledge base in order to find similar design solutions. The next step is to adapt the existing solution to new requirements and to generate a new working solution. Should the same problem be solvable using two solutions (all the parameters are identical or close enough), the solution that better solves the problem is stored in the knowledge base. Knowledge representation in such fixture design systems is performed based on cases. Every case contains criteria which are classified as input and output. The input criteria store information required for proper fixture design, while the output criteria store information with complete fixture design solutions. A subset of criteria which allow the knowledge base to be searched are the indices. Each case consists of the so called indexed and non-indexed criteria. The indexed criteria allow the search of the knowledge base. Kumar and Nee [16] performed indexing using numbers as attributes which describe the workpiece for which a fixture is designed. Liqing and Kumar [19] developed their cases from indexed criteria in the Internet environment based on XML technology. In this way they allowed exchange of information with other CAD/CAM systems. Non-indexed criteria do not allow comparison of cases (they are not normalized), but are the carriers of semantic information which are not suitable for normalization. The indices, i.e. search criteria, are essential to proper system functioning. Sun and Chen [25] developed the cases from both indexed and non-indexed criteria. To simplify the problem, he also developed a system for concept fixture design. The inference process was based on cases which define required locating elements for prismatic workpieces using the 3-2-1 locating method. The criteria were represented in a tree form. After a case is structured, it is necessary to reduce all data to a form which allows efficient manipulation and management. Normalization requires all the cases to be organized in a way which allows efficient comparison of all values. Most often used for this purpose are the well known metrics-based normalizations. Normalization is to be viewed upon as the process in which a particular set of data is allotted a common attribute in order to facilitate its management. The so far research has efficiently tackled the problem of normalization based on classifiers. Vukelic et al. [29] approached normalization by adding a system for unification of elements within complete fixture design solutions. Once normalization is completed, it is possible to perform aggregation (synthesis) of data. With the case-based fixture design systems a concept of similarity was used to calculate the difference between the existing and required fixture design solutions. Aggregation is required to generate a single value from several indices (results column) which is subsequently used to select - and, in the case of several similar solutions - sort the existing fixture design solutions. Sun and Chen [25] applied the nearest neighbour method which determines the similarity between the existing and the novel fixture solution based on a weighted sum of indices. Vukelic et al. [29] were using the induction method. The induction algorithm was used to form a decision tree, which was subsequently used to measure similarity based on If-Then rules. Boyle et al. [3] applied the knowledge guided induction method which is based on an expert system. It involves a series of questions set by the system, which have to be answered by the designer in order to narrow the domain of feasible solutions. However, the indexing remains an issue here. There are no formal methods which could be used to define those indices, thus the use of indices still relies on designer's experience. Poor understanding of design requests is causes inadequate indexing. If the design requests can be adequately formalized, then the chances are greater that those requests shall be used to generate indices, either directly or as general guidelines. It is worth noting that there exist some techniques for automated indexing, such as the criterial learning method, inductive learning method, method based on differences, method based on similarities, etc. In majority of cases, aggregation of data is solved by the nearest neighbour method. The largest problem with this design method is determination of significance and slow 10 search of the required solution. Fastness of this method is directly linearly proportional to the number of cases stored in the knowledge base. Thus, this method is more useful in the case of a relatively small knowledge base, which is rather unlikely in fixture design where quality fixture solutions demand large number of cases. Analysis of selected solutions is also at the hands of designer and his experience. There is no possibility to revise previous stages or improve the selected solution without repeating the whole fixture design procedure from scratch. Li et al. [17] proposed a solution for this problem introducing an arbitration of conflicts between dimensions and functions of fixture elements which could be reviewed in the form of intermediate results at any moment throughout the design process. Nevertheless, this solution proved inadequate.
Fig. 2. An example of output results from the case-based fixture design system [19]
Case-based reasoning can be successfully applied at the conceptual stage of fixture design, when locating and clamping surfaces are defined for predefined schemes. The essential problem in the so far investigations has been clear and systematic definition of all parameters which are necessary for implementation of adequate modifications, whether at the conceptual- or detailed fixture design stage. The set of currently available design rules is still incomplete. In addition, it is impossible to define all design rules, since their number grows with experience.
3.3. Application of Artificial Neural Networks in fixture design
Artificial Neural Networks (ANN) represent a transition from symbolical to sub-symbolical level of data processing where the rules used to transform input information into output result - fixture design, are not known. This approach does not require user to explicitly formulate knowledge. ANN are trained iteratively, over a period of time. It is important to define proper structure and interrelations. ANN consists of an input and output layer, and one or several hidden layers, each of them having certain number of neurons. Unless user instructs the system otherwise, the initial values of weight coefficients are set on a random basis. During training, weight coefficients are updated so that the next iteration produces output values which are closer to the input. Once the desired accuracy is achieved, or the predefined maximum number of iterations is reached, the training stops and the network is ready for the testing phase. In their feasibility study on application of ANNs in fixture design, Balasubramanian and Herrmann [1] give a mainly theoretical contribution which lacks a practical case study. The authors suggested application of ANNs to define measures of similarity between the existing and novel fixture design solutions as an alternative to case-based reasoning. Their goal was to establish prerequisites for fixture design based on existing design solutions. Lin and Huang [18] used ANNs to generate concept design of modular fixture for cutting-off operations based on the principles of Group Technology (GT). GT concept was used to classify workpieces of various shapes and dimensions. Once a workpiece is classified using ANN and an appropriate heuristic algorithm, concept solution is generated. The system was developed for a particular group of workpieces and in that respect lacks generality (cannot be used for a broader class of workpieces). Yuperl and Cus [32] developed a system for optimization of clamping forces which is based on artificial neural networks for prismatic parts, using the 3-2-1 clamping method, and assuming that both fixture and workpiece are rigid bodies. The 3-2-1 method requires three locating elements on the first locating surface, two locating elements on the second locating surface, and one on the third. The system can define various clamping schemes, positions of clamping elements, and magnitude of clamping force. However, it does not allow selection of required elements, but instead solves the problem of clamping on a conceptual level. Hamedi [14] also used ANNs to define positions of clamping elements, using maximum stress and elastic deformations of workpiece as the goal function. To provide input data for ANN training, he used the ANSYS CAE system. Hamedis system generated satisfactory concept solutions of clamping schemes for prismatic parts of simple geometry. Subramaniam et al. [26] used ANNs to generate concept fixture design solutions. ANN was trained to approximate fixture productivity, and the ease with which a workpiece is placed into and taken out of the fixture. For lack of real-life, industrial data, to enhance the process of training they used a performance matrix which was composed based on their own data, which renders this study unrealistic. Similar contribution came from Cai et al [5]. They developed a highly specialized system which allowed generation of concept fixture design solutions exclusively for locating elements, restricted to a predefined group type of workpieces.
Fig. 3. An example of output results from an ANN-based fixture design system [26] 11 Advantages of ANNs lie in their robustness to noisy input data as well as their ability to learn. ANNs should therefore be used for solving particular problem areas in a highly complex matter of fixture design which is virtually impossible to express algorithmically. All the so far results of application of ANNs in fixture design only allow generation of partial concept solutions. Completion of the process from conceptual to detailed fixture design is often time-consuming. The basic problem of creating an adequate ANN pertains to collecting sufficient quantity of input training data based on systematized examples from the industry.
4. CONCLUDING REMARKS AND FUTURE RESEARCH
Fixture design is a complex process which requires extensive knowledge. For this reason, fixture design automation is a very demanding undertaking and represents one of the bottlenecks in flexible manufacturing systems. None of the reviewed approaches to automated fixture design have managed to completely satisfy demands. One of the general traits of these systems is their ability to produce partial solutions, i.e. the locating and clamping elements for simple prismatic workpieces. Although this is not the only way to perform locating, the so far research has relied on the 3-2-1 locating method, as well as on a complete restraint of the workpiece, in spite of the fact that this increases both costs of fixturing and the number of constituent fixture elements. The designer is very often left on his/her own devices when it comes to decision- making, which many authors corroborate in their papers claiming that 'The quality of fixture solution depends on the designer's experience'. The influence of locating error is completely disregarded, though it greatly contributes to the total machining error, i.e. machining accuracy. Although the intelligent systems have been undergoing constant development, it is still early to talk about a general solution for intelligent fixture design, since none of the so far proposed systems have seen successful industrial application. This is due to the fact that the processes are very complex, and depend on numerous factors which must be taken into consideration (but are, in some cases, impossible to define). Obviously, the artificial neural networks are an attempt to simulate the work of the human brain. However, they are still far from sophistication and complexity required to completely replace humans in fixture design. Despite the obvious advantages of automated fixture design, there is still the problem of integration of Computer Aided Design (CAD) and Computer Integrated Manufacturing (CAM). In order to increase the efficiency of the manufacturing system as a whole, there is a logical requirement to integrate the information systems which support design decisions, and other systems which support enterprise management functions. The process of fixture design is not entirely synchronized with other design processes. In spite of numerous research efforts to automate the methodology, there was little advancement in the development of methods to support the design within Concurrent Engineering (CE) environment based on team collaboration. Future approaches to fixture design methodology must be focused not only on CAD/CAM integration, but more importantly, on providing support for multidisciplinarity through concurrent engineering paradigm. Efficient application of the concept of virtual fixture design would enable the development of fixture solutions through co-operation of multidisciplinary teams which would be in a position to understand, review, evaluate, edit and improve a virtual fixture solution before it has assumed any physical form. All this supports the claims that there is still a lot of space for potential research in the domain of automated fixture design. It is logical to expect that the novel approaches to automated fixture design and manufacturing automation in general, shall be based on non-algorithmic programming and methods of artificial intelligence. Special efforts should be aimed towards creating systems which are integral, universal, consistent, flexible, simple, and intuitive.
REFERENCES
[1] BALASUBRAMANIAN, S, HERRMANN, J . W., Using Neural Networks to Generate Design Similarity Measures, Technical Research Report 99- 38, University of Maryland, Maryland, 1999. [2] BOERMA, J . R., KALS, H. J . J ., Fixture design with FIXES: the automatic selection of positioning, clamping and supporting features for prismatic parts, Annals of CIRP, 1989, Vol. 38, No. 1, pp. 399-402. [3] BOYLE, I. M., RONG, K., BROWN, D. C., CAFixD: A Case-Based Reasoning Fixture Design Method Framework and Indexing Mechanisms, The J ournal of Computing & Information Science in Engineering, 2006, Vol. 6, No. 1, pp. 81-90. [4] BUGTAI N., YOUNG, R. I. M., Information models in an integrated fixture decision support tool, J ournal of Materials Processing Technology, 1998, Vol. 76, No. 1, pp. 29-35. [5] CAI. Q. H, ZHONG, Q. L, XIN, M. L., Concept design of checking fixture for auto-body parts based on neural networks, International J ournal of Advanced Manufacturing Technology, 2006, Vol. 30, No. 5-6, pp.574-577. [6] CECIL, J ., TAMIL - an integrated fixture design system for prismatic parts, International J ournal of Computer Integrated Manufacturing, 2004, Vol. 17, No. 5, pp. 421-434. [7] CHOU, Y. C., SRINIVAS, R. A., SARAF, S., Automatic design of machining fixtures: conceptual design, International Journal of Advanced Manufacturing Technology, 1994, Vol. 9, No. 1, pp. 3-12. [8] DAI, J . R., NEE, A. Y. C., FUH, J. Y. H., KUMAR, S. A, An approach to automating modular fixture design and assembly, Journal of Engineering Manufacture, 1997, Vol. 211, No. 7, pp. 509-521. [9] DARVISHI, A. R., GILL K. F., Expert system rules for fixture design, International J ournal of Production Research, 1990, Vol. 28, No. 10, pp. 1901-1920. 12 [10] DARVISHI, A. R., GILL, K. F., Knowledge representation database for the development of a fixture design expert system, Management and Engineering Manufacture, 1988, Vol. 202, No. 1, pp. 37-49. [11] FERREIRA, P. M., KOCHAR, B., LIU, C. R., CHANDRU, V., AIFIX: An expert system approach to fixture design, Proceedings of the International Conference ASME, 1985, Miami Beach, Florida, pp. 73-82. [12] FUH, J . Y. H., CHANG, C. H., MELKANOFF, M. A., An integrated fixture planning and analysis system for machining processes, Robotics & Computer Integrated Manufacturing, 1993, Vol. 10, No. 5, pp. 339-353. [13] GANDHI, M. V., THOMPSON, B. S., Automated design of modular fixtures for flexible manufacturing system, J ournal of Manufacturing Systems, 1986, Vol. 5, No. 4, pp. 243-252. [14] HAMEDI, M., Intelligent fixture design through a hybrid system of artificial neural network and genetic algorithm, Artificial Intelligence Review Archive, 2005, Vol. 23, No. 3, pp. 295-311. [15] KUMAR, S. A., NEE A. Y. C., PROMBANPONG, S., Expert fixture-design system for an automated manufacturing environment, Computer-aided Design, 1992, Vol. 24, No. 6, pp. 317-326. [16] KUMAR, S.A., NEE, A. Y. C., A framework for a variant fixture design system using case-based reasoning technique, Manufacturing Science and Engineering, 1995, Vol. 2, No. 1, pp. 763-775. [17] LI, W., LIA, P., RONG, Y., Case-based agile fixture design, Journal of Materials Processing Technology, 2002, Vol. 128, No. 1, pp.7-18. [18] LIN, Z. C., HUANG, J . C., The application of neural networks in fixture planning by pattern classification, J ournal of Intelligent Manufacturing, 1997, Vol. 8, No. 4, pp. 307-322. [19] LIQING, F., KUMAR, S. A., XML-based representation in a CBR system for fixture design, Computer-Aided Design & Applications, 2005, Vol. 2, No. 1-4, pp. 339-348. [20] MARKUS A, MARKCUSZ Z, FARKAS J , FILEMAN J ., Fixture design using PROLOG: an expert system, Robotics and Computer Integrated Manufacturing, 1984, Vol. 1, No. 2, pp. 167-172. [21] NEE, A. Y. C. KUMAR, S. A., A framework for an object/rule-based automated fixture design system, Annals of CIRP, 1991, Vol. 40, No. 1, pp. 147-151. [22] NNAJ I, B. O., LYU, P., Rules for an expert fixturing system on a CAD screen using flexible fixture, J ournal of Intelligent Manufacturing, 1990, Vol. 1, No. 1, pp. 31-48. [23] PHAM, D. T., DE SAM LAZARO, A., Autofix an expert CAD system for jigs and fixtures, International J ournal of Machine Tools and Manufacture, Vol. 30, No. 3, pp. 403-411. [25] SUN, S. H., CHEN, J. L., Knowledge Representation and Reasoning Methodology based on CBR Algorithm for Modular Fixture Design, J ournal of the Chinese Society of Mechanical Engineers, 2007, Vol. 28, No. 6, pp. 593-604. [26] SUBRAMANIAM, V., KUMAR, S. A., SEOW, K. C., A multi-agent approach to fixture design, Journal of Intelligent Manufacturing, 2001, Vol. 12, No. 1, pp. 31-42. [27] VUKELIC, DJ., HODOLIC, J .: System for computer aided modular fixtures design, The Journal of Manufacturing Engineering, 2006, Vol. 2, No. 5, pp. 35-42. [28] VUKELIC, DJ ., HODOLIC, J .: Information system for fixture design, J ournal of Acta Mechanica Slovaca, 2008, Vol. 12, No. 4, pp. 103-114, ISSN 1335-2393. [29] VUKELIC, DJ ., TADIC, B., HODOLIC, J ., KRIZAN, P., SIMEUNOVIC, N.: Development of an inteligent system for fixture design using case-based reasoning (CBR) technique, Journal of Manufacturing Engineering, 2009, Vol. 8, No. 4, pp. 8-11. [30] VUKELIC, DJ , ZUPERL, U., HODOLIC, J ., Complex system for fixture selection, modification, and design, The International J ournal of Advanced Manufacturing Technology, 2009, Vol. 45, No. 7-8, pp. 731- 748. [31] VUKELIC, DJ ., HODOLIC, J ., Machining fixture design via expert system, Machine design, 2009, Vol. 1, No. 1., pp. 17-20. [32] ZUPERL, U., CUS, F., A Model for Analysing and Optimazing Fixtures, Strojniski vestnik - Journal of Mechanical Engineering, 2002, Vol. 48, No. 2, pp. 73-86.
CORRESPONDENCE
Djordje VUKELIC, Mgr. MSc. Eng. University of Novi Sad Faculty of Technical Sciences Trg Dositeja Obradovica 6 21000 Novi Sad, Serbia vukelic@uns.ac.rs
Branko TADIC, Prof. D.Sc. Eng. University of Kragujevac Faculty of Mechanical Engineering Sestre J anjic 6 34000 Kragujevac, Serbia btadic@kg.ac.rs
J anko HODOLIC, Prof. D.Sc. Eng. University of Novi Sad Faculty of Technical Sciences Trg Dositeja Obradovica 6 21000 Novi Sad, Serbia hodolic@uns.ac.rs
Nenad SIMEUNOVIC, Mgr. MSc. Eng. University of Novi Sad Faculty of Technical Sciences Trg Dositeja Obradovica 6 21000 Novi Sad, Serbia nsimeun@uns.ac.rs
Paradox Wiebe E. Bijker, Roland Bal, Ruud Hendriks - The Paradox of Scientific Authority - The Role of Scientific Advice in Democracies (Inside Technology) - The MIT Press (2009)