Research On Some New Methods In Fault Diagnosis Of Gearbox With Compound Faults

Gearbox is one of the most important power transmission components, Itshealth condition will directly affect the performance of mechanical equipment.Ifthe fault in the gearbox can be predicted timely and precisely, we can effectivelyavoid enormous economic losses. So the study of advanced fault diagnosistechnology plays an important role in contributing to the normal operationalstate of the gear box. Vibration signals collected by the acceleration sensors areusually non-stationary signals, especially in the working place where the faultfeatures of weak signals may be interfered or even submerged in variousbackground noises. In addition, once the fault in the gearbox occurs, and thefault are always a compound faults which take place in different locations, formsand degree. What’s worse, these fault signals can be interfered, affected andcoupled mutually. So it is a big challenge to diagnose the compound faults underthe strong background noise nowadays. In order to solve above problems,supported by National Natural Science Foundation of China (Project number:50775157) and Basic Research Project of Shanxi Province(Projectnumber:2012011012-1) and Shanxi Province Foundation for Returness ofHigher Education(Project number:2011-12), taking gearbox as the researchobject, the new noise reduction method in recent years as the research tool, andcompound fault diagnosis as the research target, the paper has carried on aprofound and systematic research mainly on the fault features separation and the fault features extraction from the vibration signal of wind power gearbox withcompound faults.The main research methods in this paper are as follows:(1)In view of the problems such as EMD modal aliasing phenomenon and EEMD decomposition precision which affected by the singularity of amplitude ofthe added white noise, in this paper, an improved EEMD with CMF is proposed.Firstly, the combined mode function (CMF) are used as the pre-filter to improveEEMD decomposition results. CMF combines the neighboring IMFs which areobtained by EMD to get two new IMFschandcL.chcontains high frequencycomponents andcLlow frequency components. Then we determine the properadded noise amplitude according to the vibration characteristics to decomposechandcLwith EEMD, and the purpose is that EEMD is further improved toincrease the accuracy and effectiveness of its decomposition results. Finally,what extracts weak fault frequency more effectively is cyclic autocorrelationfunction analysis for every characteristic IMF. The proposed method is appliedto analyze the multi-fault of a wind power growth gearbox setup, and the resultsconfirm the advantage of the proposed method over EEMD with cyclicautocorrelation function.(2) Under the Complex environment，The rolling bearing’s fault feature understrong background noise is very weak and usually overwhelmed by noise.Ensemble empirical mode decomposition (EEMD) has been used in the faultfeature extraction of rolling bearing，but its performance is very poor when thebackground noise is very strong. The minimum entropy deconvolution (MED)and EEMD are combined for rolling bearing’s weak fault diagnosis. Firstly，thestrong background noise of rolling bearing is decreased by the MED method.Secondly, the above given signal is processed by the EEMD. At last the sensitiveintrinsic mode function (IMFs) are analyzed by cyclic autocorrelation functiondemodulation and the result is good. In the end, the simulation and the measured signals verify the feasibility of this method.(3)Cyclostationary signal has non-stationary characteristics, hence the smoothcirculation features to study the circulation statistics is very necessary. Second-order spectrum demodulation is applied to periodic vibration signal, but discrete time domain did not lead to cyclic autocorrelation function in a loop within cycling by simulation of the signal in strong background noise. In addition, when multi-carrier frequencies coexistence or relatively close, the aliasing phenomenon occurs at high frequencies inevitablely.(4) Studing on the noise reduction features of the maximum correlationkurtosis deconvolution (MCKD), while its parameters (displacement number,cycle and the number of iterations) are discussed and analyzed.The cross-term interference caused by multi-modulator and multi-carrier makesthe use of the cyclic autocorrelation demodulation limited. A new method calledthe MCKD (Maximum correlated Kurtosis deconvolution)-cyclic domaindemodulation is proposed. First the strong background noise of the originalsignal is decreased by the MCKD method in order to extract period of interest T,the above given signal is processed by the cyclic autocorrelation functiondemodulation.The new method can decrease the cross-term interference causedby multi-modulator and multi-carrier and improve the reliability of the analysis.As the application of the MCKD-cyclic domain demodulation theory in faultdiagnosis, the method is used to successfully extract fault feature from vibrationsignals.(5)The composite fault diagnosis of rotating machinery in strong noise background is the current difficulty of mechanical fault diagnosis. This paper takes thegearbox as research object, and the composite fault vibration signal of gear pitting and bearing inner and out punctuate corrosion is analyzed here. The simulationsignal and application examples show that the characteristic frequency of composite fault under strong background noise can be successfully extracted by usingthe integrated approach of EEMD, MED, MCKD,CMF and cyclic domain demo dulation, and achieve breakthrough from single fault feature extraction to multiple single fault feature extraction, which possesses a wide application prospect.