Research On The Method Of EMD And Wavelet Threshold Of Particle Swarm Optimization In The Fault Diagnosis Of Gearbox

Gearbox is one of the important machinery transmission parts, machinery equipment ofapproximately80%of failures are derived from the gearbox, with a high practicalsignificance for healthy diagnosis gearbox in modern production. In this paper, combiningEMD with PSO wavelet threshold techniques, can diagnose effectively gearbox health. Due tonatural causes and human causes, and other factors, the acquisition process of vibration signalmay contain a lot of noise, which can affect healthy diagnosis. This article describes theapplication of the principles of particle swarm optimization algorithm, and describes thewavelet transform and Wavelet threshold noise principle, the use of particle swarmoptimization (PSO) algorithm seek the optimal solution for each sub-band wavelet threshold,as a global threshold to reduce noise for wavelet threshold of vibration signal. This articledescribes several methods of wavelet threshold de-noising, calculating their SNR and MSEand PSNR, the results show that PSO algorithm based on wavelet threshold de-noising isbetter.Empirical Mode Decomposition (EMD) have a good effect in dealing with nonlinear,non-stationary signals, and has a good localization characteristics, it can be non-linear signalinto a plurality of linear component signals. According to the actual application problems of EMD,trying mirror extension methods analyzes to solve the end point effect, simulation resultsdemonstrate the effectiveness of this method.Finally, the type of JZQ250gearbox in the laboratory was made as test object, theimproved EMD method will be applied to the gearbox health diagnostics. using the EMDanalyze respectively gearbox in normal signal within the signal and to diagnose the fault of snaggletooth, bearing cup and bearing cone, and draw out the decomposition of the powerspectrum of the levels of intrinsic mode function, obtaining better effect on the gearboxdiagnosis by analyzing the power spectrum. Five order of IMF components are removed fromsignal decomposition under the condition of normal signal and the fault of snaggletoothwithin the signal, and then separately for each IMF component under conditionsapproximating entropy calculation. Which proving faulty signal complexity and nonlinearityis stronger than under normal conditions, especially the first component. Experiments showthat the empirical mode decomposition (EMD) is feasible in the gearbox healthy diagnosis.