The Research Of Control Chart Pattern Recognition Based On Wavelet Analysis

The control chart is an effective way of modern quality management, real-time quality monitoring and diagnostic tools for the production process, In particular, the control chart in the production process monitoring significantly increased the level of quality assurance of the manufacturing process. Although the control chart principles and methods play a huge role in the field of quality control, however, small probability events principle as the theoretical basis for the abnormal pattern of control chart monitoring guidelines, difficult to identify abnormal pattern of the spike, step, trends, cycles in the production process, which greatly weakened the monitoring effect of the control chart. At present, the method of combining wavelet analysis and neural network technology has become a hot spot of the control chart pattern recognition. How to use wavelet analysis method to extract the characteristics of abnormal control chart patterns, and select the corresponding BP neural network (ANN) control chart pattern recognition is the key to Control Chart Pattern Recognition Based on Wavelet Analysis.In this paper, wavelet analysis and neural network as theoretical basis for the systematic study of the control chart pattern recognition method using wavelet analysis. First of all, this paper based on the domestic and foreign researches about control chart pattern recognition, wavelet analysis and neural network to define the abnormal patterns of control charts. Secondly, Based on the control chart pattern recognition research, One-dimensional discrete wavelet decomposition (DWT) to extract features of the Control Chart Pattern data, Data extracted features as input of BP neural network to identify the control chart patterns, establish a control chart pattern recognition model based on wavelet analysis, that is DWNN control chart pattern recognition model. Furthermore, analysis on the choice of the wavelet function and the parameters of the BP neural network identifier, finally, the use of the Monte Carlo method to generate the data of control chart patterns using MATLAB tools for simulation analysis, empirical analysis of DWNN control chart pattern recognition model. The results show that:Control chart pattern recognition model based on wavelet analysis a higher correct rate compared with a separate neural network recognition model. In the wavelet analysis feature extraction, using Coif4 wavelet three layers decomposition model recognition accuracy was high,characteristics and innovations of this study:Use the one-dimensional discrete wavelet transform and BP neural network to extract abnormal control chart pattern characteristics, and the identification of control chart patterns to establish a control chart pattern recognition model based on wavelet analysis; Study Wavelet Selection and parameter design of the BP neural network in DWNN control chart pattern recognition model; Empirical analysis DWNN control chart pattern recognition model.The results show that:Coif4 wavelet three layers decomposition can get a higher total correct rate of recognition model, this study not only for the wavelet analysis and BP neural network-based control chart pattern recognition provides theoretical models and empirical results and provides a theoretical basis and methods of analysis for other control chart pattern recognition.