Joint suppression of normal gear mesh component and background noise for early local fault detection based on dynamic evolutionary digital filter. (30th September 2022)

• Record Type: Journal Article
• Authors: Wang, Liming; He, Jiafu; Xiao, Huifang; Zeng, Qiang; Ding, Xiaoxi; Shao, Yimin
• Is Part Of: Measurement. Volume 201(2022)
• Publication Date: 2022-09-30
• Subjects: Adaptive noise cancellation -- Evolutionary digital filter -- Fault diagnosis -- Gearbox; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2022.111711
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: A new dynamic evolutionary digital filter method is proposed to realize SNR improvement of early fault. Cloning and mating evolutionary strategies in Dyna-EDF method is dynamically adjusted according to the iterative process. The proposed method achieves better denoising performance than traditional EDF methods and the classic denoising methods. Abstract: Signal denoising is one of the most essential steps in mechanical fault diagnosis, which can help improve the signal-to-noise ratio (SNR) and find faults as earlier as possible. Many denoising techniques have been developed to suppress both Gaussian white and non-Gaussian background noise. However, most of them achieved poor performance in terms of assessing local gear fault in the early stage. The key problem is that only one or several gear tooth's stiffness or displacement excitation are aroused in the beginning of local faults and most of the gear teeth or gear pairs are healthy, especially for the multi-stage gearbox. And both the normal gear meshing signals and background noise lead to low SNR of fault characteristics. To overcome this problem, this paper proposes a dynamic evolutionary digital filter (Dyna-EDF) method to realize the optimal suppression of background noise and normal gear meshing component in the acquired signal aiming to enhance the SNR of early fault characteristics. The proposed method belongs to an adaptive noise cancellation (ANC) method, which can find and suppress the relevant component between the acquired signal and reference signal based on an iterative optimization of a digital filter. In the proposed method, a filter parameter population and ANC structure is initialized and established firstly. Then the filter parameters of the population are dynamically adjusted to find the optimal filter based on cloning and mating evolutionary strategies. Furthermore, "survival of the fittest" idea is employed for the optimal filter searching and applied during each iteration in Dyna-EDF method to prevent the premature problem that found in traditional EDF. Finally, the derived optimal filter is applied to suppress the relevant component between the acquired signal and reference signal using an ANC process. Results show that the proposed Dyna-EDF method successfully increase the SNR of fault characteristics, both normal gear mesh component and background noise are suppressed effectively. The proposed Dyna-EDF method seems achieve better denoising performance than traditional EDF method, its improved versions and the traditional denoising methods.