Research Of FTA-SVM-based Fault Recognition Method For Vehicle Engine

The continuous development and progress of automobile industry makes the automation and performance of engine continuously improve. At the same time, its structure has become more and more complex, coupled with the poor working environment makes its fault has the characteristics of uncertainty, frequent and destructive etc. The fault recognition is more and more difficult. Therefore, To explore a kind of reliable and accurate fault identification method of engine has important meaning for vehicle safety and environmental pollution.Support vector machine(SVM) have good ability to recognize on high-dimensional nonlinear problem and in relatively small amounts of samples can be both good generalization ability. It Is one of the highlights in agro-scientific research in the pattern recognition. The fault tree analysis method is a kind of clear thinking,logical, for system reliability analysis methods of qualitative and quantitative analysis.It Is one of the main analysis method of safety system engineering. In this paper,combined with the characteristics of SVM and FTA proposed a fault diagnosis method based on FTA-SVM. In order to improve the reliability of process to study the reliability of SVM classification data driven SVM, the use of FTA to improve the reliability of the drive of the classifier, to explore the fault recognition engine problems through the integrated application of FTA and SVM analysis method.In this paper, using FTA-SVM method of engine misfire fault and abnormal sound engine for fault recognition. Firstly, To analysis and improve the reliability of SVM mapping model by using the fault tree, the fire engine and engine abnormal sound fault tree analysis, The failure mechanism of engine misfire and knock out the fault tree model is established. To establish a nonlinear mapping model of fault data and the bottom event of fault tree. The use of a multi classification method to construct multi classifier, and using genetic algorithm to optimize the SVM parameters, and establish the SVM model.Finally, based on the MATLAB environment, we design a friendly man-machine interaction of SVM data processing system for SVM data processing part. Through the test data processing, and to verify the reliability and applicability of the method.