Research On Fault Diagnosis For Electronic-controlled Engine-based On BP Neural Network

The car is a very complex mechanical system composed of thousands of parts. The engine is equivalent to the heart of the car is its power source. Engine performance directly or indirectly affects the vehicle operating conditions. Without disassembling the engine, determining the engine technical condition accurately and quickly, which not only can guarantee the security and economy of the car running, but it will save the blind maintenance cost of manpower and resources. This subject is precisely the theory of neural network based design from a smart engine fault diagnosis system, the engine’s specific point of failure can be diagnosed according to the selected engine test data.This paper first describes the engine fault diagnosis system relies on neural network theory, which leads to the application of neural network in fault diagnosis. Type of neural network to determine the subject applies, detailed description of the core content of the selected BP neural network and genetic algorithm optimized principle, laid the theoretical foundation for the realization of diagnostic systems and create a Fault Diagnosis Network. Secondly, the use of MATLAB and ACCESS mixed programming to create a diagnostic system login screen, the use of neural network toolbox in MATLAB software to create the BP neural network. Using a genetic algorithm to optimize the BP network, combined with the GUI graphical interface to achieve human-machine interface, the preparation of the engine fault diagnosis system based on neural networks. And then implementation requires the relevant documents package deployment engine fault diagnosis system, which can be used in the non-MATLAB environment. Again, the comprehensive utilization of fault bench, K81code reader and AVL4000exhaust analyzer to build a test platform. In both cases, the engine idling and acceleration, comprehensive vehicle test with the vehicle testing for the fault symptoms needed to build neural network data sets and verify the test data set of network availability. Finally, according to the input and output data set characteristics combined with the comparison test results and related theory to determine the parameters of the network structure and function. After the neural network’s training, simulation, verifying the correctness of the diagnosis with real-time test data, fault diagnosis system with neural network technology has been set up.This engine fault diagnosis system can not only change the maintenance company’s present statues that they did not make the best of the relative detectors, but also save the time it takes to find fault, to reduce the waste of human, material during replace operation of proving fault.