Fault Diagnosis Research Of Diesel Engine Typical Fault Based On WPT-GA-BP Approach

Diesel engine as the power source of the engineering machinery is widely used in ships,vehicles,generators,and other fields.Due to the complexity of operation condition and poor working conditions,it is very easy to fall down.In the past,preventive maintenance was widely used on diesel engine valve fault handling maintenance and maintenance method,but this method is lack of accident foresee ability,high maintenance cost,low efficiency,so the practical intelligent fault diagnosis methods is necessary.In addition,due to the simple measuring of cylinder head vibration signal,it is always a hot topic to detect diesel engine fault by using diesel engine cylinder head vibration signal without disintegration.There are a lot of excitation sources in diesel engine cylinder head vibration signal,which belong to the non-stationary vibration signals,so,the appropriate non-stationary vibration signal processing method to deal with the cylinder head vibration signal is necessary.The main research contents were as follows: Establish experimental platform for diesel engine cylinder head vibration signal.Acquire diesel engine cylinder head vibration signal under different condition of normal and fault samples.The vibration signal of diesel engine is decomposed by wavelet packet analysis technology into three layers about the exhaust valve clearance fault,fault injection advance Angle and fuel delivery failure.In this paper,wavelet sub-bands of the different diesel engine fault vibration are analyzed by power spectral density analysis technology,and determined the scope of the diesel engine fault characteristics of different frequency band.The largest singular value distribution,root mean square value distribution,distribution of skewness,kurtosis distribution of different fault characteristic frequency band are analyzed.These parameters are as the input vector of BP neural network,GA-BP,SVM and GA-SVM.The different input vector's pattern classification accuracy is analyzed in the diesel engine in single fault and multiple faults.The results show: When the single fault is classified: For an single fault: BP neural network as classifier,the classification accuracy is above 90%;GA-BP as classifier,the classification accuracy is above 95%;SVM as classifier,the classification accuracy is above 92%;GA-SVM as classifier,the classification accuracy is above 96%.For three faults: BP neural network as classifier,the classification accuracy is above 90%;GA-BP as classifier,the classification accuracy is above 95%;SVM as classifier,the classification accuracy is above 87%;GA-SVM as classifier,the classification accuracy is above 95.8%.The WPT-GA-SVM for three kind typical faults of diesel engine,the classification accuracy is above 95%.