Multi-scale Analysis For Condition Monitoring Of Rotating Machinery

The origin and development of multi-scale analysis is on the basis of the people’s deepening understanding of the world. Multi-scale thought is much more abutted the natures of things and much more in line with the habit of people exploring new things than the view of single scale. On the basis of multi-scale analysis, rotating machinery condition monitoring and fault diagnosis is deeply studied on four kinds of multi-scale methods, and finally these multi-scale analysis methords are introduced to the high-speed train bearing acoustics diagnostic systems.According to the characteristic frequency and high-frequency resonance frequency that typical rotating machinery state signals contained, the wavelet coefficient variance feature is proposed with wavelet coefficient variance’s energy property. The wavelet-based multi-scale slope features which can identify the different working conditions of rotating machinery are estimated from the slope of logarithmic variances after the DWT analysis of the signals. The effectiveness of the proposed feature was verified by two experiments on bearing defect identification and gear wear diagnosis. Experimental results show that the wavelet-based multiscale slope features have the merits of high accuracy and stability in classifying different conditions of both bearings and gearbox, and thus are valuable for machinery fault diagnosis.The status mutations phenomenon often encountered in the equipment condition monitoring, and the multi-scale wavelet entropy analysis is studied to solve this problem. This method combines information entropy theory and wavelet transform, and eliminate the noise in low SNR signals by selecting the characteristic scale. Experimental results show that the wavelet time entropy, the wavelet singular entropy and the wavelet scale entropy are sensitive to the early mutation of gearbox vibration signal and can make warning of equipment failure.During the gearbox experiences wear process, there is a structure resonance phenomenon when the failure point of the gear periodic meshing. This resonance signal has a strong scale behavior and there is a strong correlation between adjacent signal points. Based on detrended fluctuation analysis, local scaling exponent method is proposed var refinement the concept of subspace to each sampling point, and adjusting the weighting function to calculate the moving average and second-order center distance in different scales. The new local scaling exponent method pays more attention to the fine local structure of the signal than the detrended fluctuation analysis. A gearbox wear test verified the local scaling exponent analysis can filter out other frequency components or noise, and effectively monitor the weak fault component.The lack of local singularity characterization make the single fractal only can reflect the overall analysis of the signals, when multi fractal analysis can characterize the signal local-scale behavior of rotating machinery more effectively. The detrended trend multifractal method is introduced to the condition monitoring of rotating machinery, and the morphological characteristics of the multifractal spectrum are proposed for quantitative extraction of the state parameters of the signal for fault classification. A gearbox wear test verifies the multi fractal analysis has a good effect in condition classification.In the context of the rapid development of high-speed railway, the Trackside Acoustic Diagnostic System (TADS) receives much more attention. A test platform is designed and manufactured for high-speed train bearing acoustic diagnosis and take advantage of the proposed multi-scale analysis method by this paper for high speed train bearing acoustic diagnosis. The analysis results further confirm the validity of multi-scale method and suggest a feasible research direction for acoustic diagnostics of high-speed train bearings.