Bearing performance degradation assessment using locality preserving projections

Research highlights? Locality preserving projections (LPP) can extract effective information from the given dataset. ? LPP improves the performance of recognizers in severity classification of bearing failure. ? LPP-based statistical control is effective for degradation assessment of bearing performance. ? A guidance for machine prognostics method is provided by applying local information extraction. The sensitivity of various original features that are characteristics of bearing performance may vary significantly under different working conditions. Thus it is critical to devise a systematic approach that provides a useful and automatic guidance on extracting the most effective information from the original features generated from vibration signals for bearing performance degradation assessment without human intervention. This paper proposed a locality preserving projections (LPP)-based dimension reduction and feature extraction (FE) approach. Different from principal component analysis (PCA) that aims to discover the global structure of the Euclidean space, LPP is capable to discover local structure of the data manifold. This may enable LPP to find more meaningful low-dimensional information hidden in the high-dimensional feature set compared with PCA. By using the extracted information by LPP, a multivariate statistical process control (MSPC)-based bearing performance quantification index is proposed, where an Exponential Weighted Moving Average (EWMA) statistic is developed by combining two effective statistic T2 and squared prediction error (SPE) statistics. LPP-EWMA does not need too much prior knowledge to improve its utility in real-world applications. The effectiveness of LPP-EWMA is evaluated experimentally on bearing test-beds. The experimental results indicate its potential applications as an effective and simple tool for bearing performance degradation assessment.