Research On Information Fusion Technology For Engine Fault Diagnosis

The engine is a typical multi-level system and the mutual influence of its components is complex.The working state of engine is affected by environment,operation conditions and other uncertain factors,and then the engine faults have the feature of diversity and fuzzy.Therefore,traditional methods for fault diagnosis are no longer suitable due to its low detection rate and the requirement of shutdown inspection,which induces substantial economic losses.A fast and accurate approach for fault diagnosis is essential for practical application when no disintegrated of the engine is preferred.Based on the difficulty of engine fault diagnosis and the application of information fusion in the engine fault diagnosis.This paper proposes a decision level information fusion model combined with the artificial neural network and D-S evidence theory.The model mainly consists of the fault signal feature extraction,local fault diagnosis and fault signal fusion on decision level.For the fault signal feature extraction,the autocorrelation method for analyzing the wavelet detail coefficients is proposed.Based on Mallat wavelet,the noise signal is separated from the original signal,and then the useful signal is reserved maximally,which is helpful for the information fusion stage.To fully realize the fault diagnosis,the paper introduces the signal characteristic of instantaneous engine speed and the cylinder pressure,which provides three different characteristic vector for subsequent diagnosis.In the stage of fault diagnosis,the limitations and disadvantages of classic BP network method are analyzed.Based on the optimized method,the superiority of improved BP neural network in fault diagnosis is verified by experiments.To further improve the fault diagnosis accuracy,D-S evidence theory is proposed to implement the decision level fusion for local diagnosis results.Then,a weighted optimization method is used to deal with the fusion problem of conflict evidence.Finally,the data fusion method based on BP network and weighted D-S evidence theory is applied to achieve the engine fault diagnosis.The experimental results show that the proposed method can diagnose the failure of fuel oil pump in engine fuel system,the damage of the injector and the damage of fuel pressure regulator.Finally,the results of fault diagnosis is consistent with the actual fault.Then,the feasibility of the proposed method in fault diagnosis of engine is verified.