Skip to main content
APEC 2026 Main banner

Other Niche and Emerging Applications

D31.4 - Data-Efficient Motor Condition Monitoring with Time Series Foundation Models

Thursday, March 26, 2026
12:00 PM - 1:45 PM CT
Location: Hemisfair Ballroom
This digest presents a data-efficient framework for motor condition monitoring using time series foundation models (TSFMs). By leveraging pre-trained models such as MOMENT and Mantis, inter-turn and inter-coil short-circuit faults in permanent magnet synchronous motors (PMSMs) are accurately diagnosed with minimal fine-tuning. Mantis achieves over 90% accuracy with only 1% labeled data, surpassing conventional methods by 22%, while MOMENT validates scaling laws, confirming performance gains with larger datasets and model sizes. The framework addresses two critical challenges in power electronics—data scarcity and label imbalance—by combining self-supervised pre-training, model assessment, and efficient fine-tuning. Results highlight TSFMs as promising tools for scalable, generalizable, and real-time predictive maintenance in electric drives.