Research On Multivariate Quality Diagnosis And Pattern Recognition Of Control Chart Based On SVM For Manufacturing Process

In the current situation of increasing complexity and diversification of products,how to ensure the quality of products will become increasingly important.For the products produced by the company,the quality requirements are getting higher and higher,and the quality control factors are also increasing.The use of multiple quality control charts and multi-quality diagnostic techniques can help companies improve product quality and product competitiveness.Under the shortcomings of the traditional static control chart,which can not meet the real-time and dynamic quality monitoring of the enterprise,this paper first proposes a multi-T2 control chart of the variable sampling strategy,and introduces the relevant theory and establishment criteria of the control chart.The Markov chain method is used to give the calculation formula for the average alarm time ATS of the VSSI-T~2 control chart and the comparative MEWMA control chart.The optimal sampling strategy of the two under different parameters is given,and it is proved that the average alarm time of the improved dynamic control chart is shorter when the runaway occurs.Secondly,when the control chart is monitored to the out-of-control signal,the control chart mode needs to be identified.The abnormal pattern of the control chart can find the potential quality problems in the production process.Based on the theoretical basis of SVM and its algorithm model,this paper is based on effective The classification of statistical features and shape features establishes a classifier of MSVM model,and optimizes the parameter combination of the classifier based on the network search method.The BP neural network algorithm is used to verify the six kinds of control of the multivariate classifier.The identification of graph patterns is feasible and has a high accuracy rate.Finally,taking the step offset in the control chart mode as an example,a multivariate process mean migration diagnosis model with positive correlation is established.When an alarm occurs in the multi-control chart,the model can accurately locate the specific variable where the mean shift occurs,and then take corresponding measures to restore the relevant variable to a stable and controlled state as soon as possible.