Research On Fault Diagnosis Method For Power Machinery Based On Local Mean Decomposition And Support Vector Machine

Fault diagnosis for power machinery is of great significance for people’s life andproperty. Signal acquisition signal processing and pattern recognition are the threemain parts of power machinery fault diagnosis process. Among them, signalprocessing and pattern recognition is the key to the fault diagnosis. In this paper, thelatest time-frequency analysis of local mean decomposition (LMD) and advancedpattern recognition methods support vector machine(SVM)were combined for thediagnosis of power machinery typical fault.By using LMD method, a complex multi-component AM-FM signal signal canbe self-adaptively decomposed into a set of product functions, each of which is theproduct of an envelope signal and a frequency modulated signal. Envelope signal isthe instantaneous amplitude of the PF, while the PF component of the instantaneousfrequency can be obtained from the frequency modulated signal. Through all theinstantaneous amplitude and instantaneous frequency of each PF we can get thetime-frequency distribution of the signal.Compared with the neural network, SVM does not have the problem ofover-fitting and Under-fitting.what’s more it has characteristics of small sample,nonlinear and high dimensional pattern recognition. In this paper, SVM was used forthe diagnosis of typical aircraft engine fault, the results show that SVM is an efficientfault diagnosis method.LMD and SVM were combined for the diagnosis of dieselengine valve fault. The results show that the combination of LMD and SVM methodsused in fault diagnosis is an effective way.This paper mainly carried out the following work:1. Experimental data collection: use LMS data acquisition system for collectingthe Vibration signal on cylinder head of WP7diesel engine, which is under normaland fault conditions2. Introduce the principle of LMD method, analyze the shortcomings of themethod, then propose improvement strategies. Through the simulation signal analysis,we found the improved LMD algorithm are better than the original algorithm both onthe precision of decomposition and the computing speed, then the improved algorithmwere used on the fault diagnosis of bearing inner and outer ring fault, which furtherproves that the improved LMD algorithm is valid.3. Introduce the principle of SVM, make a deep study on the parameter selection, then propose GAPSO algorithm and improved particle swarm algorithm. Iris data isused for algorithm validation, the results show that with the proposed algorithm forSVM parameters optimization, the SVM classification accuracy had an obviouslyincrease, and the speed also got some degree of ascension.4. Use aircraft engine fault influence matrix to generate typical aircraft enginefault samples, then make a comparison of GAPSO-SVM algorithm and other typicalfault diagnosis algorithm on aircraft engine typical fault diagnosis. The results showthat GAPSO-SVM method is better than the BP neural network, self-organizationneural network, SVM algorithm and GA-SVM algorithm.5. The improved LMD algorithm was used for feature extraction of the dieselengine valve fault signal, then using the improved PSO-SVM as a classifier for faultdiagnosis of diesel engine valve, the experimental results show that the method is verysuitable to be used in diesel engine valve fault diagnosis6. Use matlab GUI to make a fault diagnosis software, which is based on SVMand neural network.