Rotating Machine Fault Diagnosis Based On Parametric Adaptive Variational Mode Decomposition

Rotating machines play a critical role in the modern industry which are widely used to perform various functions in different fields.As a result,condition monitoring of rotating machinery has received more and more attention.An important procedure in rotating machine fault diagnosis is to acquire the related information such as temperature,displacement,acceleration,stress,sound to be used for fault diagnosis.It is well-know when a failure of a moving element or an interrelated component of a rotating machine occurs,it can cause the inbalance of a mechanical movement and to generate vibration.Consequently,a change of vibration signal without a change of the operation condition of a rotating machine can directly reflect the change health condition of the machine.Nonetheless,as the vibration signal acquired from the machine usually contains multi signal components and noise,and is often characterized by non-stationary signals due to change of loads and speeds.This then leads to an increased difficulty in mode decomposition and feature extraction during the signal processing.Taking this into consideration,this thesis intends to solve some of these problems by carrying out research in areas elaborated below:Firstly,rolling element bearing and planetary gear are chosen as the research objects in this study as they are the key moving elements in a rotating machine.The vibration signal generation and signal characteristics due to faults in a machine is investigated and summarized.Then,the commonly employed adaptive time-frequency analysis techniques such as Empirical mode decomposition(EMD),ensemble empirical mode decomposition(EEMD),local mean decomposition(LMD)and variational mode decomposition(VMD)are introduced and their pros and cons in the signal analysis are discussed.Particular attentions are devoted to study the performance and characteristics of the recently proposed VMD technique in the non-stationary signal analysis.The result of a comparison study shows that VMD outperforms other methods in the signal decomposition,and is much more robust in the analysis of noise contaminated signals.A drawback of VMD is that the number of IMFs and the quadratic penalty term used in the decomposition need to be pre-determined which are usually selected manually without considering the signal characteristics,which can lead to inaccuracte results.To overcome this problem,partical swamp optimization(PSO)is proposed in this study to optimize the IMF number and the quadratic penalty term according to the signal characteristics so that these two parameters can be determined adaptively to the signal characteristics.A detailed procedure of this process is elaborated as follows:(1)a weighted kurtosis index composing of kurtosis index and a correlation coefficient is chosen and used as the fitness function in PSO optimization,(2)the maximum weighted kurtosis index is used as the optimized objective of PSO to determine adaptively the IMF number and the value of the quadratic penalty term,(3)the related IMFs from VMD are used in the envelope analysis to obtain the defect signal components for fault diagnosis.The effectively of the proposed technique in extracting the fault characteristic frequency from noise contaminated signals is evaluated using simulated bearing defect signals,and experimental data from a bearing test rig and from a planetary gearbox test rig.The analysis results confirm the effectively of the proposed technique in extracting characteristic signal components in noise contaminated signals for machine fault diagnosis.