Develop and validate an embedded thermal digital twin for high-voltage electric machines. The virtual thermal sensor employs both physics-based and data-driven models to predict real-time temperatures in critical motor components such as stator windings and rotor magnets, achieving high accuracy (+/- 3.25°C).

Technology Features:

Virtual Sensors: Enable real-time condition monitoring and spatial temperature predictions.

Model Accuracy: Accurate representation of local hotspots with error typically below 5%.

Efficient Calibration: Completed in two days using bench tests.

Embedded Execution: Average compute time of 256.4 µs on an Infineon Aurix Tricore KIT TC3x7 V2.0.

Managing complex designs (e.g., hairpin coils, cooling channels).

Avoiding oversimplification and ensuring clarity in model physics.

Integrating 3D effects and fluid dynamics into the model.

Consistently accurate temperature prediction across varied operating conditions.

Enhances the reliability and safety of high-voltage electric machines.

Optimizes thermal management without over-relying on physical sensors.

Establishes a scalable methodology for future product development.

SEG Website