Combing Artificial Neural Network With Case-Based Reasoning In Fault Diagnosis

This paper mainly investigates the combination of Artificial neural network (ANN) and Case-based reasoning (CBR) used in fault diagnosis. ANN and CBR, which are the typical technology of AI, have been used successfully in fault diagnosis. However, with regard of the limitation of the technology itself and the sophistication of modern faults, there has the limitation, which is unavoidable, to use these technology singly. According to the characters of the two technologies used in fault diagnosis, combining them maybe a good method to solve those limitations. Oriented to this background, this paper has conducted a study on how to combine ANN and CBR in fault diagnosis. In big complex system, the method of fault diagnosis based on ANN is lack of clarity, which can't have a good result, and the one based on CBR run slowly besides of its imperfect diagnosis precision. According to these, this paper has combined them, and raised the project. In this project, ANN is treated as a pre-classifier because of its function in pattern matching, and then cases in CBR are been indexed by the ANN results, which also guide the retrieve of cases. Then the ANN-CBR model has been raised. Moreover, this paper has checked the model by using data acquired from a simulation of the fluid catalytic cracking (FCCU) system in oil refining industry and The result shows that the performance of the combination model is better than the situation of individual evidently, which proved this method to be practical as well as valid.