A quality abnormality diagnosis method for dynamic process based on pattern recognition

摘要

In order to detect and recognize each anomaly pattern from the dynamic dataflow, which contains a variety of abnormal patterns, this paper present a quality abnormality diagnosis method based on stationary wavelet transform and neural networks. First, stationary wavelet transform was chosen to decompose the dynamic dataflow. Second, we propose the improved modulus maxima method and find the position of singular point from the dynamic dataflow. Finally, the characteristic data are extracted according to the found location and inputted to the neural networks for classification. Where our idea is to choose the stationary wavelet transform and find the most suitable wavelet for decomposing the dynamic dataflow. Moreover, modulus maxima are extracted directly from the decomposition level corresponding to the large scale. Experiment results show that the proposed method not only can recognize quality anomaly pattern fleetly and accurately from the dynamic dataflow, but also has a good anti noise property.