A Review of Singular Spectral Analysis to Extract Components from Gearbox Data

Abstract

Condition monitoring for rotating machines under time-varying environmental and operating conditions remains an important research problem for several industries, including wind turbines within the renewable energy sector; ship, train and freight transport within the supply-chain sector; crushing and grinding comminution within the mining sector. Proposed methods to solve this problem include synchronous statistics, the squared envelope spectrum, the order-frequency spectral coherence and the integrated squared spectral coherence. Singular Spectral Analysis (SSA) offers a non-parametric alternative to automatically identify potential components of interest. The components of interest are obtained from the resultant matrices, which are computed from grouping elementary matrices. Their contribution to signal reconstruction is achieved through diagonal averaging. SSA is fundamentally a reconstruction-focused linear latent variable model aiming at efficiently explaining the variance in the signal. Firstly, we expect SSA to be more informative when the component of interest manifests strongly in the signal’s variance. Secondly, we only expect SSA to isolate the component of interest from the other components in the signal, if variance can isolate its contribution. However, the latter is rather unlikely. Although Singular Spectral Analysis (SSA) is a well-established technique, it has not been critically analysed for its ability to separate components for a damaged gearbox under time-varying operating conditions. This work shows that SSA can separate a damaged gearbox under time-varying operating conditions into various components. Furthermore, SSA is shown to decompose vibration signals from a damaged gearbox into various signal components that could aid fault diagnosis.