Injection Product Defects Diagnostic System Based On Fuzzy Neural Network Research And Development

With the research on the injection molding process and the mould technology, the symptoms, reasons and solutions of the faults are studied and summarized on the basis of the experts' knowledge and experience. With fuzzy theory, neural network technology and expert system technology, a fuzzy neural network (FNN) based fault diagnosis expert system for injection-molded products is studied and developed.A FNN model is established with the study on the fuzziness and the uncertainty in the process of finding the reasons for the faults' symptoms. In the realization of the neural network, the traditional structure is divided into 6 sub-modules according to the categories of the reasons. The experts' knowledge and experience of solving the problems is converted into the training samples for FNN. After training, the knowledge implicated in the samples can be saved as the weights and threshold of the neural network. By this means, the knowledge base of FNN is formed.The knowledge of the solutions to the faults is expressed into several rules and all the rules are saved in rule base. In the practical fault diagnostic process, the system obtains the reasons for the fault' s symptoms by FNN parallel processing, and the solutions to the reasons can be obtained by rule-based reasoning.The knowledge bases of the system are created by SQL Server2000 technology. With ADO database accessing technology, the maintenance and the management of the knowledge bases are realized by programming and accessing the data files under the circumstance of Delphi.The method and the process of using the system are presented by an example. The diagnostic result of the instance shows that the system has the characteristics of high performance and high accuracy.