基于小波重构的控制图并发异常模式识别研究

对于统计质量控制过程中的复杂过程而言，多种异常的并发现象比较普遍，而常规的基于规则的方法以及人工神经网络（ANNs）技术均针对单一异常模式的识别，难以完成对并发异常模式的识别任务。提出一种混合方法，将小波分析与ANNs相结合，通过小波分解重构将并发异常模式分解为基本的异常模式组合，无须用并发异常样本训练ANNs，实现对并发异常模式的有效识别。

Wavelet-reconstruction based recognition method for control charts concurrent abnormal patterns

For the complex process of Statistical Quality Control(SQC),the concurrent of various abnormity is ordinary.However,the traditional rule-based methods and Artificial Neural Networks(ANNs) technique can only recognize the single pattern.In this paper,a hybrid method is developed through the combination of wavelet analysis and ANNs.By wavelet decomposition and reconstruction,the concurrent abnormal patterns can be decomposed into different basic patterns.Without being trained by concurrent abnormal patterns samples,the ANNs will effectively recognize the concurrent patterns by taking the decomposed patterns as input.