Structured neural network models to improve robust design solutions

Abstract The primary objective of robust design (RD) is to achieve desirable combinations of the process mean and variance while improving the level of product quality. Well estimated models for process mean and process variance (or standard deviation) in RD have traditionally involved several error assumptions. However, the input–output functional relationships can be well estimated without these error assumptions by using a neural network (NN). In this paper, we therefore propose an NN-based estimation method as a RD modeling approach. The modeling method based on the feedback NN structures is first integrated in the RD response functions estimation stage. Two new feedback NN structures are then proposed. Next, the typical recurrent NNs, such as Elman-type and Jordan-type NNs, as well as the proposed feedback NN structures are proposed as an alternative RD modeling method. Simulation studies were conducted to verify the potential of the proposed NN-based estimation method. A case study was used to investigate the effectiveness of the proposed modeling method. Comparative studies between the proposed NN structure-based methods and existing methods (i.e., conventional NN structures, statistical LSM, IP, and RSM) were conducted. The results of these particular simulation and case studies results represents that the proposed approaches may provide slightly better solutions.