Development of a New Estimation and method using Neural Network for Robust Design Modeling and Optimization

Robust design (RD) is has emerged a key concept to achieve the consistentimprove product/process performance in an early design stage. It has also been recognized as cost effective methodology in market place for more than twenty years. Many RD methods including experimental design, estimation, and optimization, were reported in literature. However, there is room for improvement, products and processes which has significantly meaning in competitive manufacturing nowadays. Traditional The primary objective of this paper is to propose a new estimation method using neural network (NN) in order to provide an alternative aspect of response surface methodology (RSM) based on ordinary least squares method (LSM). Next, a new modeling by using NN approaches is then proposed in order to estimate a functional relationship between input factors and output responses. Finally，a comparative study using simulation is performed as verification purposes. This simulation study demonstrates that the proposed NN-based RD method provides better optimal solutions than RSM.