基于小波分析和PSO-SVM的控制图混合模式识别

The mixed patterns recognition of control chart based on wavelet analysis and PSO-SVM

由于质量过程的复杂性,质量过程数据常会有多种异常的混合现象,为了提高对控制图混合模式的识别效果,将支持向量机(Support vector machine,SVM)分类器的参数经过粒子群优化算法(Particle swarm optimization,PSO)算法优化,然后与小波分析技术相结合,设计了三层控制图模式识别模型.该模型首先识别模式是否正常,如果发现异常,则对异常的模式进行小波包分解,将分解后的低频部分和高频部分分别输入第二层和第三层PSO-SVM分类器中进行模式的分类识别.通过仿真实验的验证得,该模型的平均识别率为98.33%,对混合模式的识别率也在95%之上,由此证明了该控制图模式识别模型的有效性.最后,对该模型进行了实例验证,该模型可以很好的识别出控制图混合模式,证明了模型的可行性.

Because of the complexity of the process quality, quality process data often have a mixture of abnormal phenomenon. In order to improve the effect of control chart mixed patterns recognition, the three layers of model of control chart patterns recognition based on PSO-SVM and wavelet analysis technology was designed. Control chart patterns was distinguished firstly. Then if pattern was abnormal the quality data was decomposed with wavelet packet, and the LF and HF portions was imputed in the second and third PSO-SVM classifier to recognize the pattern. Through the simulation experiments, the average recognition rate of the model is 98.33% approximately, the recognition rate of mixed mode is about 95%. Finally, the model is validated feasible through examples.