Using Artificial Neural Networks for Experimental Design in Off-Line Quality Control

摘要

This paper explains how optimal design can be achieved by using design of experiments in conjunction with neural networks. It is becoming common practice in industry to find the optimal design through the Taguchi experimental design method. In order to identify the optimal design, Taguchi replaces the need for running a full factorial design of experiments by a fractional factorial design using orthogonal arrays (OA). However, the compromise between the use of fractional factorial design and full factorial design requires some assumptions to be made in identifying the optimal design parameters and consequently leads to some uncertainty in the result. The neural network was trained using the results of a fractional factorial design for an intelligent sensor example. The neural network was then used to predict the response values for the full factorial design. A comparison between the Taguchi method and the neural network approach highlights the superior results produced by the neural network.