Weak Fault Detection and Health Degradation Monitoring Using Customized Standard Multiwavelets
Jing Yuan,Yu Wang,Yizhen Peng,Chenjun Wei
DOI: https://doi.org/10.1016/j.ymssp.2017.03.005
IF: 8.4
2017-01-01
Mechanical Systems and Signal Processing
Abstract:Due to the nonobvious symptoms contaminated by a large amount of background noise, it is challenging to beforehand detect and predictively monitor the weak faults for machinery security assurance. Multiwavelets can act as adaptive non-stationary signal processing tools, potentially viable for weak fault diagnosis. However, the signal-based multiwavelets suffer from such problems as the imperfect properties missing the crucial orthogonality, the decomposition distortion impossibly reflecting the relationships between the faults and signatures, the single objective optimization and independence for fault prognostic. Thus, customized standard multiwavelets are proposed for weak fault detection and health degradation monitoring, especially the weak fault signature quantitative identification. First, the flexible standard multiwavelets are designed using the construction method derived from scalar wavelets, seizing the desired properties for accurate detection of weak faults and avoiding the distortion issue for feature quantitative identification. Second, the multi objective optimization combined three dimensionless indicators of the normalized energy entropy, normalized singular entropy and kurtosis index is introduced to the evaluation criterions, and benefits for selecting the potential best basis functions for weak faults without the influence of the variable working condition. Third, an ensemble health indicator fused by the kurtosis index, impulse index and clearance index of the original signal along with the normalized energy entropy and normalized singular entropy by the customized standard multiwavelets is achieved using Mahalanobis distance to continuously monitor the health condition and track the performance degradation. Finally, three experimental case studies are implemented to demonstrate the feasibility and effectiveness of the proposed method. The results show that the proposed method can quantitatively identify the fault signature of a slight rub on the inner race of a locomotive bearing, effectively detect and locate the potential failure from a complicated epicyclic gear train and successfully reveal the fault development and performance degradation of a test bearing in the lifetime. (C) 2017 Elsevier Ltd. All rights reserved.