Robotics and Computer Integrated Manufacturing 16 (2000) 383 } 396 Design candidate identi"cation using neural network-based fuzzy reasoning J. Sun, D.K. Kalenchuk, D. Xue, P. Gu* Department of Mechanical and Manufacturing Engineering, The University of Calgary, 2500 University Drive NW, Calgary, Alberta, Canada T2N 1N4 Department of Mechanical Engineering, University of Saskatchewan, Saskatoon, Saskatchewan, Canada Received 17 December 1999; received in revised form 15 March 2000; accepted 3 May 2000 Abstract Conceptual design has profound impact on success of a product design. Identi"cation of the best conceptual design candidate is a crucial step as design information is not complete and design knowledge is minimal at conceptual design stage. This paper presents a method for design candidate evaluation and identi"cation using neural network-based fuzzy reasoning. The method consists of the following steps: (1) acquisition of customer needs and ranking of their importance, (2) establishment of measurable metrics and their relations with customer needs, (3) development of design speci"cations and initial evaluation of design candidates, and (4) evaluation and identi"cation of design candidates based on design speci"cations and customer needs using neural network-based fuzzy reasoning. A case study is given to show the e!ectiveness of the proposed method and associated algorithms.  2000 Elsevier Science Ltd. All rights reserved. Keywords: Conceptual design; Design evaluation; Fuzzy reasoning; Neural network 1. Introduction Development of a product undergoes a sequence of processes including conceptual design, embodiment de- sign, detailed design, production process planning, manufacturing, assembly, and so on. With the advances in computer technologies, many of these design and manufacturing activities have been computerized to as- sist engineers such as computer-aided design (CAD), computer-aided manufacturing (CAM), computer-aided process planning (CAPP), and so on [1]. These com- puterized systems have signi"cantly improved quality and productivity of design and manufacturing processes. However, most of the CAD systems focus on detailed stage. Very few computer-aided design systems are avail- able to support conceptual design activities [2]. Conceptual design is a process to develop design can- didate based upon design requirements. The design re- quirements are usually de"ned based upon customer needs, benchmarking of products from competitors and other analysis. These requirements are then translated into measurable technical attributes that are easily used * Corresponding author. E-mail address: gu@enme.ucalgary.ca (P. Gu). for evaluating design candidates. Among all feasible can- didates, the best candidate is selected for further develop- ment. Mapping from design requirements to design candidates and evaluation to these candidates, however, are non-trivial tasks, which require substantial research e!orts. The systematic study on conceptual design was started in 1970s [3]. Pahl and Beitz [3] de"ned design functional primitives such as gears and shafts, and stored them in libraries. At conceptual design stage, a design function is usually decomposed into a number of sub-functions. A design solution is accomplished by selecting design primitives to satisfy these sub-functions. Other system- atic approaches for formulating conceptual design include the axiomatic design method [4] and quality function deployment (QFD) method [5]. In axiomatic design, mapping from functional requirements (FRs) to design parameters (DPs) is modeled by a matrix [4]. A design is evaluated in terms of independence of func- tional requirements and information content. The quality function deployment method is an approach to "rst identify customer requirements and their importance measures, and then translate these data into technical attributes and importance measures [5]. Design speci- "cations are developed by comparing existing designs for improving the competitiveness of the new product. 0736-5845/00/$ - see front matter  2000 Elsevier Science Ltd. All rights reserved. PII: S 0 7 3 6 - 5 8 4 5 ( 0 0 ) 0 0 0 1 7 - X