Research, Quality Control Based On Genetic Algorithms And Neural Networks

Control chart developed by W.A.Shewhart is an important tool of statistical process control widely used in quality control field. As control charts, of which kinds of patterns are not easy to be identified by just a few rules, must be based on some definite statistical suppose, it seems not perfect just using control charts in manufacturing process.In recent years, with the development of artificial neural network, it has been widely applied in the modeling and differentiation of non-linearity system. As a kind of local-approaching network, probabilistic neural network (PNN) is suitable for identification and classification of patterns, owe to its simple structure and arithmetic. To overcome the large amount of work brought by the unsupervised learning arithmetic of PNN, a new supervised learning algorithm based on the genetic algorithm (GA) for the PNN is developed, that is, Bring adaptive genetic algorithm which trains the parameters into the PNN, thus form a new supervised PNN based on GA. This new model does pattern recognition of control charts in quality process, to make sure that the manufacturing process is in control.In simulation, sampling data of visible characters are established to train and test the model. Then the paper compares its results with those of unsupervised PNN and BP network. From the comparisons, it is found that the model has the highest recognition accuracy. Finally, to contact with actual quality control in manufacturing process, the model is brought into LH Corporation to do actual quality control. We hope that with help of the model, quality control problems in LH Corporation can be solved.