Variable Condition And Fault Diagnosis Of Gearbox Vibration Signal Processing Method

The operation of gearbox is under a varied working condition, with the change ofspeed and load. The vibration signal of non-linear and multi-component could be obtainedinevitably after combining a variety of vibration factors. Therefore, the hotspot anddifficulty of the research is how to diagnose fault signals of variable condition gearboxmore effectively currently. The methods to analyze the non-stationary signals were studied,including wavelet transform, the envelope spectrum, the empirical mode decompositionand wigner-ville distribution signal processing etc. The advantages and disadvantages ofthese methods for analyzing non-stationary signals were compared，and the methods towavelet threshold denoising, empirical mode decomposition and fast independentcomponent analysis were improved. The fault signals of variable condition gearbox weresimulated on the test-bed, also the features of the fault signals were extracted and thefeasibility of some methods to analyze signals was verified.Firstly, the disadvantages to deal with non-stationary signals with traditional spectrummethods were discussed by using traditional spectrum methods to analyze simulatedsignals. And the simulated signals were analyzed by using time-frequency analysis method.The result showed that there was some limitation to analyze non-stationary signals only bytime-frequency analysis method. Therefore, more time-frequency analysis methods werecombined to use when the non-stationary signals are disposed, fault signals would beanalyzed more effectively.Secondly, the changes of load had a little impact on fault signal’ frequency, but theyhad more impact on amplitude spectrum only when the gear box under varied loadconditions. Therefore, the study of signals’ characteristics and extracting methods werefocused on. Due to the more complex vibration signal in the varied conditions gearbox,fast independent component analysis method was adopted to improve signal-noiseseparation of gear and bearing’s fault signals, so as to extract the signals’ featuresaccurately. Under the non-stationary working condition of gearbox, it was more effectiveto adopt wavelet envelope spectrum analysis methods when the speed had a little change.When the speed changed dramatically, treasures of gearbox’s signals could be extractedaccurately by using empirical mode decomposition and wigner-ville distribution analysismethods. The efficiency of signal analysis could be improved effectively by using thesignal analysis system which was designed by GUI function of Matlab. Finally, the fault signal vibration data of variable condition gearbox were analyzed,effective characteristic parameters were extracted, characteristic parameters of variablecondition gearbox fault diagnosis were established, and BP neural network identificationmode and carry model identification of the fault characteristic parameters of the gearboxwere analyzed.The conclusions were obtained by analyzing with various signal processing methodsand test, which would provide certain reference for the study of variable conditiongearbox and failure analysis of large rotating equipment.