ZF has introduced an AI-driven approach to temperature estimation deep within an electric vehicle’s (EV) motor, where direct measurement is typically impractical due to space, heat, or cost constraints. TempAI, uses self-learning models to estimate internal motor temperatures, without additional physical sensors. It enables precise control of thermal operating limits in EV drivetrains.
How it works
TempAI applies a data-driven model that estimates temperature in the motor core by analyzing sensor signals already available within the powertrain system.
According to ZF, the approach improves temperature forecast accuracy by more than 15% over traditional estimation techniques. This allows electric motors to operate closer to thermal limits without exceeding them, supporting optimized use of the motor’s capabilities.
The system uses an AI model that’s automatically generated from test data and designed for rapid deployment. Because the model is lightweight computationally, it can run on existing ECU hardware, which simplifies its integration into existing EV control architectures.
Accurate internal temperature estimates allow for higher peak power output and efficiency gains in standard EV drive cycles. For example, ZF reports a 6% increase in peak power and measurable efficiency improvements during WLTP cycle testing. In high-load driving scenarios, such as track-style acceleration or hill climbs, estimated energy savings range from 6% to 18%, depending on operating conditions.
Additionally, improved thermal control enables more targeted use of materials, potentially reducing the need for heavy rare earth elements in rotor designs. The optimized temperature management also supports accelerated development, with thermal models that once took months to build and validate now generated in days using test bench and vehicle data.
Model training and data sources
TempAI is trained using large datasets collected during system-level test cycles. These datasets include accessible temperature signals (such as oil pan measurements) and motor operating conditions (e.g., rotor speed, torque demands, ambient temperature).
Using this data, the model identifies correlations and behaviors that influence rotor and stator temperature changes, particularly under dynamic load conditions common in EV applications.
While traditional sensors are limited in their ability to capture internal rotor temperature during operation, data-driven models like TempAI offer an alternative for real-time control and validation without added hardware complexity. This approach reflects a broader trend toward predictive, software-defined control strategies in EV powertrain systems.
Filed Under: Electric Motor, Technology News