Study of the fault diagnosis method based on rbf neural network

Based on the surprising development of information technology, there tends to be more electronic apparatus installed in automobiles which presents a new challenge for vehicle fault diagnosis. Then how to locate the existence and type of the traditional faults that occur in automobile electronic control systems proves to be of great significance. This paper puts forward extraction condition characteristic signal data from car´s running status information, using RBF neural network to build anomalies and normal signal state mapping relationship model. Through the method of decision-making in fault diagnosis to recognize faults, in MATLAB man-machine environment, three common faults (needle valve wear, nozzles carbon, injector spring break) of fuel system in automobile engine are verified, and the result indicates that the diagnosis method is effective and feasible.