Pattern Recognition for Control Charts Using AR Spectrum and Fuzzy ARTMAP Neural Network

Abstract

The control charts pattern recognition can detect the unnatural fluctuation of the machining process and enhance the automation level of quality management, so it is important for manufacturing enterprise to implement the automatic pattern recognition of control charts. In this paper the quality data generated by Mote-Carlo simulation are preprocessed through AR spectrum analysis and the characteristic quantity are picked up in frequency domain. This approach decreases the complexity of characteristic quantity compared with traditional encoded mode. Then Fuzzy ARTMAP neural network which has incremental learning ability compared with traditional BP neural network is presented to recognize the control chart patterns. The recognition result indicates that the introduced method in this paper has advantage of traditional methods.