Automotive Electronic Fuel Injection Engine Fault Diagnosis Technology Research

Engines are a source of automotive power. With their job performance continues to improve, the degree of electron continuous improvement, its structure has become more and more complex, although the probability of engine failure is not too high, but the event of failure will be difficult to diagnose. With the rapid development of artificial intelligence techniques, neural network has been applied to more and more and more complex system fault diagnosis. Based on EFI engine idle speed instability failure as an example, using multi-sensor information fusion method of neural network of several major cause of the malfunction at the MATLAB environment simulation.The paper first analyzes the domestic and foreign automotive fault diagnosis technology status and development situation, summed up a number of fault diagnosis of the main theories and methods, on the basic principle of neural network knowledge, radial basis function neural network gave a more detailed introduction; Secondly, electric control system of the engine's basic knowledge in the introduction and analysis of several main sensors and actuators of the waveform that the waveform characteristics and the relationship between fault; then to study the multi-sensor data fusion neural network engine fault diagnosis method in order to EFI engine idle speed instability as an example, an analysis of relevant sensors, actuators waveform characteristic parameters for the original feature vector and studied the application of principal component analysis method of feature extraction and information fusion center design; Finally in MATLAB environment designed diagnosis of unstable idle speed multi-sensor information fusion RBF neural networks, and carried out a test.Last, In MATLAB environment, the fault diagnosis approach on the RBF neural network performs the fault pattern recognition and the diagnosis. Put multi-sensor information fusion method of neural network and single-sensor information of the neural network method for comparison to prove that the multi-sensor information fusion method of neural network fault diagnosis accuracy rate higher.