Detection of structural changes is critically important in operational monitoring of a rotating machine. This paper presents a novel framework for this purpose, where a graph model for data modeling is adopted to represent/capture statistical dynamics in machine operations. Meanwhile we develop a numerical method for computing temporal anomalies in the constructed graphs. The martingale-test method is employed for the change detection when making decisions on possible structural changes, where excellent performance is demonstrated outperforming exciting results such as the autoregressive-integrated-moving average (ARIMA) model. Comprehensive experimental results indicate good potentials of the proposed algorithm in various engineering applications. This work is an extension of a recent result (Lu et al., 2017).

Additional Metadata
Keywords Graph model, Machine monitoring, Martingale test, Structural change detection
Persistent URL
Journal Mechanical Systems and Signal Processing
Lu, G. (Guoliang), Liu, J, & Yan, P. (Peng). (2018). Graph-based structural change detection for rotating machinery monitoring. Mechanical Systems and Signal Processing, 99, 73–82. doi:10.1016/j.ymssp.2017.06.003