DESIGN HEURISTICS EXTRACTION: Acquiring engineering knowledge from previous designs

摘要

During the design process, it is necessary to evaluate mechanical designs to decide if they are worth pursuing. Early evaluation of such designs can reduce the cost due to repeated iterations and corrections. This paper introduces a preliminary approach for using principal component analysis and self-organizing neural networks to evaluate designs at the conceptual stage in the design life-cycle. The paper focuses on a method for extracting the evaluation rules that are implicit to the neural network with the aim of generating design heuristics. A case study is described, and design heuristics are generated based on analysis of previous designs using the proposed approach.