BIW Welding Automatic Production Line Fault Diagnosis Expert System’s Design And Implementationon

Nowadays, the flow production line is used popularly in most enterprises, especially for the automatic welding production line. The upstream and downstream production line will be badly influenced if the failure cannot be diagnosed and repaired in a short time. Although there are many sensors in the imported equipment, the data from sensors couldn’t be exported because the equipment cannot provide the interface about the data. But most of the diagnosis methods require those data. After years of developing, the automatic welding production line becomes more mature, the fault doesn’t happen very often, so the fault data is not enough considering that there are lots of fault possibilities. Parts of the fault do not happen independently, so the fault tree diagnosis method meets its shortcomings.To solve the above problem, a fusion fault model and an inference model were built by combing the advantages of fault tree model and the Bayesian network model, in which the advantages of both models could be used. An expert fault diagnosis system for automatic welding production line was built based on the hybrid model.Firstly, based on the structure and working principle of production line, the cause and the type of the fault in welding production line were studied, the basic characteristics of the fault happened in welding production line was analyzed., the relationship between the fault was figured out and the fault tree model of welding production line was established.Secondly, the fault sample table was gotten through the table about fault records and the Bayesian network was built by using the structure expectation maximization learning algorithm. As a result of the restrictions exist in the fault sample, the Bayesian network is incomplete. While the fault tree model is relatively complete, it is a lack of the intrinsic relationship between implicit nodes.Thirdly, in order to avoid the shortcomings of the two models on the fault diagnosis for the target system, this thesis created a hybrid fault model which came from the fault tree and the Bayesian networks.Finally, the hybrid fault model was used as the inference engine of diagnosis expert system. To build the expert system, the storage structure of the fault information, fault reason, repair suggestion data in the knowledge base was designed. The representation, the interface and function modules of expert system were designed at the same time, then the fault diagnosis expert system for welding system was developed.Compared with the traditional Bayesian network model, the hybrid model can solve problems that traditional method meet when it is applied in the situation of being lack of fault sample and improve the accuracy of fault diagnosis in welding automatic production line.