| As a typical tool of process quality diagnosis,quality control chart is widely used in modern industrial production.Quality control chart is used to test the stability of the production process.Fluctuations of data points in the quality control chart reflect the quality fluctuation in the production process.Fast and effective quality control chart pattern recognition is helpful for the timely detection of abnormal factors in the production process.First,this paper uses Monte Carlo method to generate simulation data,the data is used to train and test the model.Then,two kinds of characteristics,primitive features and features extracted based on technology of Principal Component Analysis,were evaluated by the principle of separability criterion based on distance,and the characteristics with good separability were selected.Next,the distribution of quality control chart patterns is visualized using technology of Principal Component Analysis,and the distribution of quality control chart patterns in feature space is analyzed,this will pave the way for the establishment of the model.Then,based on OCSVM algorithm and SVM algorithm,the PCA_OCSVM_SVM quality control chart pattern recognition model is constructed,and the moving window method is used to simulate the experiment.The recognition speed and accuracy of the model in the experimental results were used as evaluation indexes,having studied the effect of different OCSVM training samples and different quantities of principal components and different classifier parameters on model performance.And the performance of the model based on three different feature vectors,primitive feature vectors and PCA feature vectors and statistical feature vectors,is studied.The superiority of the model is illustrated by comparing it with other quality control chart pattern recognition model.Simulation results show that the proposed model has better performance,to achieve high recognition accuracy while maintaining a fast recognition speed. |
PDF Full Text Request