Fault detection in propulsion motors in the presence of concept drift

TL;DR

The paper addresses the challenge of detecting overheating in marine propulsion motor stator windings under concept drift. It proposes residual-based fault detection with drift adaptation, using an additive drift term $\hat{b}_t$ and two adaptation strategies: on-demand and continuous. Temperature is predicted with a HistGradientBoostingRegressor, and residuals are monitored by a drift-aware anomaly detector (CUSUM/EWMA-based) to deliver early, reliable alarms without full model retraining. Evaluations on real and simulated drifts show significant improvements in time-to-detection and false-alarm control, highlighting the practical value of drift-adaptive monitoring for long-lived, safety-critical propulsion systems.

Abstract

Machine learning and statistical methods can improve conventional motor protection systems, providing early warning and detection of emerging failures. Data-driven methods rely on historical data to learn how the system is expected to behave under normal circumstances. An unexpected change in the underlying system may cause a change in the statistical properties of the data, and by this alter the performance of the fault detection algorithm in terms of time to detection and false alarms. This kind of change, called \textit{concept drift}, requires adaptations to maintain constant performance. In this article, we present a machine learning approach for detecting overheating in the stator windings of marine electrical propulsion motors. Using simulated overheating faults injected into operational data, the methods are shown to provide early detection compared to conventional methods based on temperature readings and fixed limits. The proposed monitors are designed to operate for a type of concept drift observed in operational data collected from a specific class of motors in a fleet of ships. Using a mix of real and simulated concept drifts, it is shown that the proposed monitors are able to provide early detections during and after concept drifts, without the need for full model retraining.

Figures (8)

• Figure 1: A rise in baseline temperature leads to a rise in the residuals which again triggers a false alarm.
• Figure 2: The baseline temperature decreases, leading to a negative bias in the residuals, masking an overheating event.
• Figure 3: System overview of an electrical propulsion motor with a Variable Frequency Drive.
• Figure 4: A real case of concept drift observed in one of the residual temperature series. After the drift, a persistent offset of around $-7$ appears in the residuals.
• Figure 5: Drift adjustments $\hat{b}_i$ (top) and drift adjusted residuals $\tilde{e}_t = e_t - \hat{b}_t$ (bottom) for both the CUSUM and EWMA drift adaptors. Ten days of data is shown around the time of a true drift, indicated by the dashed red line. The hyperparameters of the CUSUM and the EWMA are set to the values used in the simulation experiments, described in Appendix B.