控制图异常模式, 小波包分解, 支持向量机, 模式识别

Control chart pattern recognition based on wavelet analysis and SVM was presented to improve the accuracy of control chart pattern recognition, in particular the accuracy of hybrid control pattern recognition. Process quality characteristics data were decomposed in wavelet packet to extract low-frequency approximation signals and the band energy information, which were taken as the inputs of SVM classifier recognizing trends, step and periodic signals respectively, and these signals figure the final model together. The simulation experiment used for testing recognition effect shows that the method based on wavelet analysis and SVM has an advantage over traditional methods in identification accuracy.