Research On Diesel Engine Fault Diagnosis Method Based On VMD And KFCM

As one of the most widely used conventional power sources,diesel engine plays an important role in industry,agriculture,transportation and national defense.Diesel engine is prone to failure because of the complex structure,many parts and components,relatively high explosive pressure in cylinder,and often operating in harsh environment.Once a diesel engine breaks down,it may cause serious consequences and even lead to accidents.Therefore,the research of diesel engine fault diagnosis methods to improve the diagnostic accuracy is of great theoretical and engineering value for improving the theory system of non-disintegration fault diagnosis of complex mechanism,and improving the reliability of diesel engine.The vibration signal of diesel engine contains abundant working information.The vibration testing method is simple and mature.Therefore,the fault diagnosis research based on vibration signal analysis has been favored by many scholars.Two key problems need to solve in this kind of research.One is to extract fault feature information from complex vibration signals,and the other is to use fault feature information for pattern recognition.The diesel engine test bench is built in this paper.The typical diesel engine fault states,such as abnormal valve clearance,insufficient fuel supply,abnormal rail pressure and abnormal injection advance angle,are set up to verify the diagnostic accuracy of the proposed methods in this paper.Vibration signals of diesel engine are complex broadband signals.It is very difficult to extract the fault source signals and their feature parameters directly from them,so it is necessary to study the appropriate algorithm for signal analysis.By comparing the existing algorithms,this paper chooses variational mode decomposition(VMD)as the pre-decomposition algorithm of vibration signals.Compared with the traditional recursive decomposition algorithms,VMD has better robustness.However,some parameters of the algorithm need to select artificially and there is no accepted selection criterion.Therefore,this paper optimizes the mode number K and the quadratic penalty factor of VMD algorithm.Firstly,an adaptive K-value selection method is proposed by using correlation coefficient as the condition of screening effective components.Through the validation of simulation signals and experimental signals,the method can decompose different signals adaptively and obtain narrow-band components with high correlation.Then,according to the parameter variation law of observing the decomposition results of a large number of signals,the appropriate range of value selection is given.The key to improve the accuracy of fault diagnosis is to select suitable fault feature parameters and appropriate fault pattern recognition methods so that different types of faults can be distinguished significantly.In this paper,the kernel fuzzy c-means clustering(KFCM)with high computational efficiency and accuracy is used as the pattern recognition algorithm for diesel engine faults.Then feature parameters are extracted from the decomposition results of VMD and input into KFCM to diagnose.By using different feature parameters to diagnose faults,it found that the diagnostic accuracy is the highest when using maximum singular value as the parameter.The diagnostic results show that for abnormal valve clearance,insufficient fuel supply,abnormal common rail pressure and abnormal injection advance angle,the combined VMD-KFCM method can achieve 98.7%,98.6%,96% and 93.3% diagnostic accuracy respectively,and 94% for multiple faults at the same time.In terms of fault diagnosis accuracy,the optimized VMD-KFCM is obviously superior to EMD-KFCM and the original VMD-KFCM.The research results prove that the optimized VMD-KFCM proposed in this paper has certain guiding significance for the development of typical fault diagnosis technology for diesel engines.