Study On Valve Clearance Fault Diagnosis For Internal Combustion Engine

The changes of valve clearance directly affect engine fuel economy and emission performance in internal combustion engine.The location of valves are affected by high temperature and high pressure environment for a long time,as well as the mechanical shocks brought by the reciprocating motion of piston,which makes that the valve-train is prone to wearied.In order to identify the internal combustion engine valve clearance fault accurately,the precision and rapidity of valve clearance fault diagnosis using fault signal processing and recognition algorithms are mainly discussed in this study.The main works are as follows:1)Spectrum determination of the fault signal Considering that the directly exciting force is not taken into account in the traditional valve clearance fault diagnosis analysis,a valve-train quality model is established through AVL EXCITE Timing Drive platform for fault simulation.The cam bearing force and valve seat force are determined as the main exciting force,then the exciting force is analyzed in time domain and frequency domain with different valve clearance fault.The variation range of amplitudes about exciting force is determined in frequency domain from 2 kHz to 5.5 kHz.Therefore,fault feature extraction has a clear frequency range in the subsequent study,and the vector dimensions and computational complexities are reduced.2)Time domain analysis and frequency domain analysis about cylinder head vibration acceleration signal of internal combustion engine The results of time domain analysis show that the cylinder head vibration acceleration signal amplitude increases with the fault severity augment.And the results of frequency domain analysis show that the variation of signal amplitude is proportional to the valve clearance fault severity in the frequency range from 2 kHz to 5.5 kHz.The overall trend of valve fault information can be acquired from cylinder head vibration signals analysis by time domain analysis method.The amplitude which is component of cylinder head vibration signal in the frequency domain can be clearly observed by frequency domain analysis method,but the collected fault information is inadequate.The analysis result of short time Fourier transform(STFT)shows that fault feature extraction of cylinder head vibration signal is more effective by the time-frequency analysis method.3)Valve clearance fault diagnosis based on the wavelet packet decomposition and k-Nearest NeighborConsidering that wavelet analysis has a low resolution on high frequency and there is an overlap in the fault feature of valve clearance fault samples in different categories,the method of valve clearance fault diagnosis using wavelet packet and k neighboring algorithm is proposed in this study.Firstly,the exhaust fault signal can be decomposed into three layers by wavelet packet decomposition.Secondly,the fault frequency band that related to the fault band determined in the step(1)is selected,and the energy value of fault frequency band is further calculated and normalized.Thirdly,the eigenvector is built based on disposed fault frequency band and input to the k neighboring algorithm.Finally,the proposed method using wavelet packet and k neighboring algorithm is applied to diagnose valve clearance fault of the internal combustion engine.Compared with BP neural network(BPNN),support vector machine(SVM)and least squares support vector machine(LSSVM),the results of fault diagnosis based on k neighboring algorithm can reach an identification precision of 95% in a short time.4)Valve clearance fault diagnosis based on empirical mode decomposition(EMD)and BPNNConsidering that EMD does not need to pre-determine basis function and the characteristic of fault is more practical,the method of valve clearance fault diagnosis using EMD and BPNN is proposed in this study.Firstly,the acquired signal is decomposed by EMD.Secondly,the intrinsic mode function(IMF)that related to the fault band determined in the step(1)is selected,and the energy value of IMF is further calculated and normalized.Thirdly,the eigenvector is built based on disposed IMF and input to the BPNN.Finally,the proposed method using EMD and BPNN is applied to diagnose valve clearance fault of the internal combustion engine.The results show that the method is more robustness and the diagnostic accuracy is higher compared with the fault diagnosis method based on wavelet packet and k-Nearest Neighbor.The diagnostic accuracy can reach 97.5%.Aiming to the problem of modal aliasing in the process of fault signal decomposition using EMD,Ensemble Empirical Mode Decomposition(EEMD)which has stronger ability of resistance to modal aliasing is used to decompose fault signal in this study.Considering that BPNN has the problem of slow convergence and easily falling into local optimum,the BPNN is optimized by Particle Swarm Optimization(PSO)and further used to identify the valve clearance fault diagnosis.Compared with the former methods,the results show that the method can obtain the optimal solution faster,and the diagnosis accuracy can achieve 98.33%.