基于小波重构与SVM-BPNN的动态过程在线智能监控

为了提高动态过程运行状态在线监控效率,提出了基于小波重构与支持向量(support vector machine,SVM)-反向传播神经网络(back propagation neural network,BPNN)相结合的在线智能监控方法.首先,运用离散小波变换对动态过程实测数据流进行重构,并提取其形状特征.其次,利用训练好的小波重构特征的SVM、均值特征的BPNN及重构后形状特征的SVM,对"监控窗口"内实测数据流进行异常模式识别.最后,应用该方法对某精密轴加工过程进行在线智能监控.结果表明:所提模型识别精度高、训练耗时少,其整体性能明显优于小波重构的BPNN模型与基于统计和形状特征的多分类支持向量机(multi-class support vector machine,MSVM)模型,是一种更为有效的动态过程在线智能监控方法.

Online intelligent monitoring for dynamic process based on wavelet reconstruction and SVM-BPNN

In order to improve the online monitoring efficiency, a novel monitoring method for dynamic process operating state is proposed by combining support vector machine (SVM), back propagation neural network (BPNN) and wavelet reconstruction. Firstly, discrete wavelet transform is used to reconstruct the measured data flow of dynamic process and extract shape features from reconstructed data series. Then, SVM based on wave reconstruction feature, BPNN based on mean feature and SVM based on shape feature from reconstructed data series are used to recognize the abnormal patterns of data flow in the "monitoring window". Finally, the proposed monitoring method is used to monitor the operating state of an axis machining process. Results indicate that the proposed monitoring method has a higher monitoring efficiency as compared to BPNN based on wavelet reconstruction and multi-class support vector machine (MSVM) based on statistical feature and shape feature, which has demonstrated its effectiveness.