Research Of Parts And Fitting Faults Diagnosis For Internal Combustion Engine

The internal combustion engine is the key equipment for national production anddefense which is widely used in industry, agriculture, energy, transportation and otherfields. The operation condition of the engine may affect the reliability and safety of thewhole dynamic system directly. Therefore, in order to ensure the system operatingrightly and safely, and improve the maintenance quality and efficiency of the equipment,the condition monitoring and fault diagnosis of the internal combustion engine isparticularly important.The condition monitoring and fault diagnosis methods based on vibration signalcan identify the running status of the engine rapidly and effectively. However, due tothe changing speed working process of the engine, the time sampled vibration signalmay appear the condition of energy leakage and frequency fuzzy. At the same time,because of complex structure of the engine, numerous vibration excitation sources, andmultiple interface transfer paths, the vibration signals from the surface present typicalcharacteristic of polymorphism impact, transmission and the time-varying characteristicof gauss. It’s difficult to identify the internal combustion engine fault accurately basedon the traditional time signal analysis and feature extraction technology.Therefore, this paper proposed characteristic parameters extraction methods basedon diagonal cumulant and wavelet feature bands of angular domain signals, and a faultdiagnosis method of the engine based on support vector machine (SVM). Simulationand experimental results show that the new characteristic parameters based oncumulant-wavelet bands can extract engine fault information from complex signalsaccurately and efficiently, the SVM classification algorithm based on new featureparameters as input samples has better more ability for fault diagnosis. This articlemainly content as follows:①The feature parameters extraction method based on diagonal cumulant: For theextracting fault features from the nonlinear vibration signal effectively, diagonalcumulant is defined based on the third order cumulant and its properties is analysised,the statistical parameters (diagonal cumulant energy, diagonal cumulant standarddeviation) based on the diagonal cumulant are presented as feature extractionparameters. The performance of the fearture parameters are verified by using simulationsignals. ②The feature parameters extraction method based on wavelet feature band: Forextracting the fault feature from the nonstationary vibration signals effectively, first ofall, the feature frequency bands of wavelet decomposition selection algorithm based onthe biggest correlation coefficients is proposed. Further, the statistical parameters(energy rate, standard deviation rate, spectrum energy rate and spectrum mean rate)based on the feature frequency bands are presented as feature extraction parameters.③The internal combustion engine fault diagnosis method based on support vectormachine (SVM): By using the cumulant-frequency feature parameters as data sampleset, the multiple fault diagnosis method of internal combustion engines was established.The performance of the multiple fault diagnosis method is verified by using simulationsignals, and compare with the classification algorithm based on the traditional inputparameters.④The experimental verification of engine multiple fault diagnosis method: Theconnecting rod bearing faults of engine are recognized by using the fault diagnosismethod based on diagonal cumulant-feature frequency band parameters. The diagnosisaccuracy of the classification method is far higher than method based on traditionalparameters. The efficiency and accuracy of the multiple fault diagnosis algorithm beverified.