A Compositive Method of Neural Networks and Control Charts for Monitoring Process Disturbance Based on Integrated SPC/EPC

To solve the problems of "window of opportunity" and autocorrelation under the integrated scheme of statistical process control (SPC) and engineering process control (EPC), a compositive method of neural networks and conventional SPC techniques was presented in this paper. Neural networks technique was used to monitor the process output, while EWMA chart and Shewart chart were adopted to detect the process input. In order to validate the advantage of the compositive method, a large number of simulation experiments were done and results showed: the neural networks technique made a significant improvement to recognize the relatively large disturbance, such as step 3 or more. It can detect the output deviation at the beginning of the disturbance. However, the neural networks technique can not successfully monitor the small disturbance, especial small drift. Monitoring process input is more effective than monitoring output when the conventional SPC method are used, while small disturbance can be detected and false alarm can be avoided using the compositive method.