A Research About Multivariate Quality Diagnosis Using LS-SVM Residual Control Chart

Quality control and diagnosis is an important part of quality management. Quality control chart is an important tool of statistical process control. It can be used to monitor the process and recognize whether the process is in control. The identification of an out-of-control variable or variables after a multivariate control chart signals has been an interesting topic for many researchers over the last few years. As many methods for detecting which of the different variables are out-of-control are presented, they are not wholly satisfactory.The paper introduces the existing methods of quality diagnosis and the characteristic of the support vector machine, and then puts forward the LS-SVM residual chart which can be used for the quality diagnosis. The method uses support vector machines to fit the multivariate relationship, and constructs the residual control chart. The application of the method in actual cases and simulation data shows that its performance is better than univariate control charts, PCA method and the residual control charts based on multiple linear regression in respect of diagnostic accuracy and it is more convenient than the MYT decomposition method. From the ARL of the LS-SVM residual chart, we can know the control chart can recognize the out of control signal more quickly and detect the variables which is the reason of the signal more accurately. In addition the method can diagnosis accurately based on the small sample.