Automobile Fault Diagnosis System Based On Neural Network And Its Application

As the core components of cars and other vehicles, the engine’s structure became morecomplex. The engine could run securely and stably that was related to the safe driving of thecar. For the status quo of automotive engine fault diagnosis research, the typical fault ofelectronically controlled gasoline engine has been analyzed comprehensively and deeply. Onthe base the BP neural network was used to fault diagnosis for the electronically controlledgasoline engine.Large number of literatures on automobile engine fault diagnosis, and neural networkwere summarized. On this base, the development history and trends of fault diagnosistechnology, the research status of domestic and foreign automobile fault diagnosis technologywere discussed. According to the research in the automotive repair shop and vehicle inspectionstation as well as detection and analysis in the automotive repair shop for186vehicles a faultyengine， the starting system fault, ignition system fault, fuel supply system fault, lubricationsystem fault, cooling fault, engine knock fault and electronic system fault of electronicallycontrolled gasoline engines were summarized and discussed.BP neural network was taken as an example which basic idea, structural characteristics,learning algorithm and the algorithm flow were analyzed. The segmented adaptive method forartificial fish’s view and step was designed to improve the BAFSA. Simulation results showedthat the convergence efficiency and global optimal solution accuracy of segmented adaptiveAFSA (AAFSA) were improved, and the performance of the algorithm was improved.For thedeficiency of BP neural network the improved AFSA was used to optimize BP neural network,which improved the performance of BP neural network. Simulation results showed that thetraining number of neural network optimized by the improved AFSA decreased.According to the characteristics of the car engine failure, the engine failure dataexpressed by the two-valued logic, which avoid the impact of pathological samples of neuralnetwork effectively. On the base of the above-mentioned typical faults analysis, start-updifficulties, stall, acceleration tempering etc.11kinds representative automobile engine faultand idle fault, ignition coils, ignition timing wrong etc.11kinds fault reason were taken astraining samples. On the MATLAB platform, BP neural network was used to research the fault diagnosis for automotive electronic control gasoline engine. The simulation results showedthat this method could diagnose the fault of automobile engine accurately.