Studying Of Quality Abnormal Pattern Methods For Dynamic Process Based On Wavelet Transformation

Recently, with many industries becoming larger, more continuous and automatic，huge amounts of data is generated by the dynamic process. In order to ensure thesecurity and stability of dynamic production process, more and more attention is paidto the dynamic process monitoring and fault diagnosis. And the compound faultdiagnosis is becoming the intensive research subject. Designing effective extractionalgorithm and combining it with the classifier is becoming a mainstream trend in thefield of fault diagnosis.Based on the problem of complex dynamic process pattern recognition, the mainworks has been done in this paper are listed as follows.1. Taking the auto-processing of tobacco as the background, we studied the dataprocessing, feature extraction, classification algorithm related to the patternrecognition process through the theoretical analysis on abnormal patternrecognition method of dynamic process.2. According to the offline measurement data and quality abnormal information ofdynamic process, abnormal patterns dynamic process were defined.3. A singularity detection algorithm is presented to the dynamic dataflow withstrong noise and many abnormal patterns, which improved the detectionspeed, accuracy and anti-noise significantly.4. The adaptive feature extraction to abnormal patterns of dynamic dataflow isimplemented on the basis of the last job and recognized the abnormal patternsthrough the combination with classifier.