Research On Intelligent Quality Control Technology Based On Quality Information Integration

This research was supported by the project: Integrative Quality Management Systems based on Quality Information Technology Integration, which was sponsored by National Natural Science Foundation of China (No. 70272032).With the advancement of the society and the development of science and technology, the tendency of current quality engineering was intelligent quality control。 This paper researched mainly on the theory and methodology of intelligent quality control and its key techniques. Several new theories, new methods and new models were proposed: methods of quality information acquirement, transmission and disposal; methods of Intelligent Statistical Quality Control; mathematics models of quality prediction; etc.This paper mainly included the following two aspects: quality information acquirement and integration techniques; processing quality control methods and its intelligent techniques.The researches on quality information acquirement and integration techniques included:(1) Intelligent static correction and dynamic compensation of measuring system based on SVR-FLANN were proposed. With the recognition of the construction method of generic functional link artificial neural networks (FLANN), a novel construction method based on Support Vector Machine (SVM) was discussed and applied for Capacitor Pressure Sensor (CPS)'s temperature compensation and correction. It was proved that the SVR, with the proper choice of the parameters, had a form similar to the FLANN and was equivalent to the FLANN estimate. Based on this, an identification method for dynamic inverse system model of sensors using SVR-FLANN model was presented. And the presented one is faster in speed, higher in accuracy, much capability of noise resistance and better for sensors dynamic system.(2) Identification and compensation methods for nonlinear dynamic system of transducer based on LS-SVM were discussed. The original nonlinear dynamic system of transducer was supposed to be expressed by a linear dynamic subunit followed by a nonlinear static subunit——Wiener model. By the function expansion, nonlinear transfer function of compensator was convert to a class of linear one——intermediate model. And the coefficients of the intermediate model were gotten through LS-SVM regression algorithm. The relation of the coefficients of intermediate model and that of transducer's two parts were derived, through which the nonlinear static part and linear dynamic part of compensator were identified simultaneously. The presented method possessed prominent advantages: only once dynamic calibrating experiment need be made; the analytic expressions of nonlinear dynamic compensator were derived; the compensator was more robust in noise resistance due to the good features of LS-SVM.(3) Feature extraction and integration methods for high dimensional quality information based on LS-SVM were presented. The formulation of linear feature extraction was made in the same fashion as that in the LS-SVM linear regression algorithm. Then the data was mapped from the original input space to a high dimensional feature by following the usual SVM...