Profile Monitoring Using Artificial Neural Network

• Activity Number: 354
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Quality and Productivity Section
• Author(s): Yi-Hua Wang*+ and Jen Tang
• Companies: and Purdue University
• Keywords: Profile Monitoring ; Artificial Neural Network ; EWMA Control Charts

Abstract

In many applications the quality of a process or product is best characterized and summarized by a functional relationship between a response variable and one or more explanatory variables. At each sampling stage one observes a collection of data points that can be represented by a curve (or profile). Profile monitoring is used to understand and to check the stability of this relationship or curve over time. The profile can be represented adequately by a simple linear regression model or more complicated models. One difficulty is how to estimate the functional relationship when we are given a profile without assuming the process is in-control. The purpose of this paper is to estimate the functional relationship of the profile using a three-layer ANN approach with radial basis functions as the combination function for the hidden neurons. We will derive the approximate likelihood ratio test statistics for the stability of the functional relationship and estimate the variance of the random error. Then, we will use the EWMA control chart approach to monitor the stability of the functional relationship and the variance of the random error.