Research Of Intelligent Fault Diagnosis Algorithm On Circuits

With the rapid development of artificial intelligence, and deep research on theory of statistics, neural networks and support vector machine combined with other algorithms apply to all kinds of fault diagnosis, and achieve good results. Through analysing the advantages and disadvantages of these two models, and combined with other algorithms, it constructs two improved diagnostic models, and realizes comprehensive diagnosis of circuits fault, and proves it’s superiority through the experiments.Firstly, in order to solve the weakness of hard to converge when the input data is high dimension and in large quantity for the traditional BP network, a new integrated BP neural network based on constraint-based clustering is proposed to fasten the convergence of training process, with momentum and adaptive learning rate. Using clustering algorithm, the training samples are firstly divided into several sample-sub-sets with similar size and trained respectively. Moreover, the corresponding outputs as well as the diagnostic data’s membership factors relative to each sub-set are both taken into account during diagnosing. The feasibility was verified through data gathering, feature extracting and BP diagnosing on a demo circuit board.Secondly, it’s hard to raise the correct rate in fault diagnose, because of inseparable and mis-partitionable regions and low generalized performance of the multiclass support vector machine, a fault diagnosis algorithm based on fuzzy support vector machine integrated is proposed. According to the decision function value of each sub-classifier, the next iterative training set is constructed through replacing the corresponding categories of samples in the current training set with the inseparable ones in the validation set, thereby it expands the degree of individual differences among the basic classifiers. Then integrating the optimal basic classifiers selected by gamele meothd, it implements fault diagnosis based on fuzzy fusion and the corresponding fault location rules. The experiments show that it has overcome the issues of existing inseparable areas in multi-class support vector machine and its low generalized performance.Finally, in colligate diagnosis system for simulative circuit failure, it constructs an integrated knowledge model sets which covering neural networks and support vector machine, and then it achieves fusion diagnosis mechanism and the quick update for knowledge model sets through the feedbacks of diagnosis and relearning function.