Control Limits for Changepoint Method Using Neural Networks

In this paper we first present the Change Point Problems in Statistical Process Control. Then we focus on the special case of abrupt shift in mean. A very appropriate method for this problem is the generalized likelihood ratio test statistics introduced by Hawkins, Qui and Kang named as Changepoint Method. As one can see the flaw of this method is that the control limits cannot be gained through statistical theoretical calculations. We calculate these limits using neural networks. After analyzing different structures, we use a one-layer neural network using a sigmoid hyperbolic tangent function with delta learning rule. At the end, the performance of calculated control limits is analyzed using Monte-Carlo simulation.

Keywords:

Statistical Quality Control , Change Point Problem , Changepoint Method , Shift in Mean , Control Limits , Neural Network