Informative singular value decomposition and its application in fault detection of planetary gearbox

Abstract

The faults of rotating machineries are generally characterized by the periodic singular impulses. However, the fault features of planetary gearbox are often complex modulated and submerged by other components, for its vibration signal has the characteristics of multi-source and multi transmission path. An informative singular value decomposition (ISVD) filter based on the modulation performance of singular value component signal (SVCS) is developed to enhance the fault characteristics of planetary gearbox and weaken the interference in this paper. The fault features can be detected by the envelope spectrum analysis (ESA). In this method, a fast SVD is firstly performed to decompose the vibration signal into a series of SVCSs. Secondly, the information detector of negentropy combined with cyclic autocorrelation (CA) is applied to estimate the informativeness of each SVCS. Then the ISVD filtered signal can be obtained by the reconstruction of amplified informative SVCSs. Finally, the fault features can be extracted by ESA from the ISVD filtered signal. The performance of the proposed method is verified by simulation study and experimental data measured from a planetary gearbox. Results show that the proposed ISVD-ESA strategy can enhance and extract the weak features of multi-modulation, and it can accurately extract the faults of tooth tip pitting and misalignment of sun gear of the planetary gearbox.

Linear and Nonlinear Dimensionality Reduction from Fluid Mechanics to Machine Learning

“…Focusing on the literature in experimental fluid dynamics, the POD has also been extensively used as a filter, for example to remove outliers in PIV measurement (Higham et al, 2016;Raiola et al, 2015) or to pre-process velocity fields for pressure integration (Charonko et al, 2010), to pre-process images (Mendez et al, 2017), to fill 'gaps' in experimental data (Saini et al, 2016) to construct efficient regressors and interpolators (Bouhoubeiny and Druault, 2009;Casa and Krueger, 2013;Karri et al, 2009), to validate numerical simulations (Kriegseis et al, 2010) or to build estimators of quasi-periodic flows (Bourgeois et al, 2013;Loiseau et al, 2018). Moreover, the POD has been used to enhance adaptive least square problems (see Yao et al (2017)), fault diagnostics (Shen et al, 2022) or optimal pressure placement (Castillo and Messina, 2020).…”

Investigating the vibration response and modulation mechanism for health monitoring of wind turbine planetary gearboxes using a tribodynamics-based analytical model

“…For this reason, the inherent characteristics of modulation sidebands can result in the invalidation of traditional fault diagnosis methodologies, which have been proven to be effective applied for fixed axial gearboxes [2,3]. In view of this, numerous studies have been implemented in which traditional signal processing techniques have been modified and adapted to effectively diagnose WT planetary gearboxes [4][5][6]. Furthermore, plenty of data-driven machine learning algorithms have been applied to diagnose uneven running conditions of planetary gearboxes, while treating the diagnostic procedure as a black box [7][8][9].…”