Research On Quality Control Based On Probability Neural Networks

In recent years, with the development of artificial neural network, it has been widely applied in the modeling and differentiation of non-linearity system. Since it requires sampling data to be of high density, comprehensiveness and compatibility, the traditional random sampling method could not satisfy this requirement. In this paper, the orthogonal analysis method has been chosen to select sampling data, because it is of high regularity, which can satisfy the requirements of the artificial neural network. Meanwhile, when the traditional back-propagation neural network is applied in pattern recognition, it has low generalization ability, which could lead to problems in the training process like local minimum and inability to ascertain the number of nerve cells on the hidden layer. So in this paper, the probability neural network has been applied into pattern recognition process, which can provide a better solution to the problem.In case study, the paper takes the "tooth form poor" and "rot booth" of the jointing nut in the fork vehicle hand brake as the research subjects. By collecting sampling data with the orthogonal analysis method, the paper gives the prediction to product quality by use of PNN, and then compares its results with that of BP. From the comparison, it is found that though the same orthogonal analysis method has been applied, PNN has higher prediction accuracy than BP, which reveals the advantages of the former in pattern recognition of quality control.