Rotor temperature prediction using multiple sensors with R-squared value over 96.9%

Predictive thermal management for PMSMs is critical. Machine learning-based time series forecasting, utilizing edge AI at the motor controller, anticipates rotor temperature and torque changes based on sensor data eliminating cloud latency. This enables proactive thermal control and load optimization, maximizing motor efficiency and preventing failures. Our MSP microcontrollers, featuring hardware acceleration, achieve >96.9% R-squared accuracy in rotor temperature prediction.

The dataset comprises 185 hours of sensor measurements collected at 2 Hz from a permanent magnet synchronous motor (PMSM) deployed on a test bench at Paderborn University, featuring 69 measurement profiles where the motor is excited by hand-designed driving cycles that denote reference speed and torque through random walks in the speed-torque plane to imitate real-world driving conditions.

The dataset includes 12 sensor attributes: currents in d/q-coordinates (i_d, i_q), voltages in d/q-coordinates (u_d, u_q), motor_speed, torque, ambient temperature, coolant temperature, permanent magnet surface temperature (pm), and three stator temperatures (stator_winding, stator_tooth, stator_yoke). This open-source dataset is available on Kaggle.

Accelerate development with advanced optimization and searching algorithms for design, training, and depolying AI models with superior performance. 

CCStudio™ Edge AI Studio gives a start to finish workflow for deploying trained models to embedded targets.For developers seeking deeper customization and control, the MSPM0-SDK offers a comprehensive framework for building and integrating edge AI functionality into your own embeded applications.

The MSP-series MCUs deliver scable performance for executing and accelerating rotor temperature prediction models, along with key SoC features critical to your application.