Research On Small Sample Multivariate Process Quality Monitoring And Diagnosis Based On SPC

The formation of product quality often runs through the entire product lifecycle and is the lifeblood of enterprise development,while the process quality in the processing is the cornerstone of the final product quality.With the continuous improvement of quality requirements from the market,customers,and enterprises,traditional quality testing and process parameter optimization can no longer meet the current quality needs.The emergence of process quality control in the processing provides a solution to this problem.With the complexity of the structure of processed products and the improvement of precision requirements,new problems have been raised for quality control in the processing,and the technology of multivariate process quality control has been developed to a certain extent.Therefore,this article focuses on the characteristics of limited sample data and correlation in the multivariate process quality in small batch processing,studies the monitoring and diagnosis method for small sample and multivariate process quality based on SPC.The main research content includes:Firstly,in view of the fact that the traditional multivariate process quality monitoring method is vulnerable to outlier in the case of small samples,a small sample multivariate process quality monitoring method based on0)7)7)4)9)2)~2 is designed based on Bayesian theory.By using Bayesian theory,historical sample information is taken as prior information,and then combined with existing sample information to solve control chart parameters,the accuracy and rationality of quality monitoring are improved.Secondly,to the issue of difficulty in providing accurate diagnostic results for existing multivariate process quality correlation diagnosis methods,a correlation diagnosis method based on the maximum correlation component pair set is proposed.Represent the multivariate process quality correlation relationship with the binary correlation relationship of the maximum correlation component pair set,then analyze the abnormal correlation relationship through multiple binary~2 statistics to reduce the spatial complexity of the calculation and the problem of redundant or missing diagnostic result information.Next,a multivariate process quality component diagnosis method based on improved BP neural network is proposed to address the problem of quality component fluctuation diagnosis.Divide the quality component fluctuations into six basic types,transform the diagnostic problem into a pattern recognition problem,and use an improved BP neural network classifier designed to identify the actual fluctuation types,thereby diagnosing the abnormal types and quality points of quality component fluctuations.Finally,for the issue of process abnormality adjustment,a process quality adjustment plan based on the knowledge base of abnormal factors is proposed.Associate abnormal factors that affect process quality with abnormal fluctuation types or points through association rules and store them in the knowledge base to reduce the response speed of process quality adjustment processes.