With its AI-based TempAI solution, ZF is integrating a new method that takes temperature management in electric drives to a new level. By using a self-learning temperature model, TempAI improves forecast accuracy by over 15%, enabling significantly more precise thermal utilization of the electric machine.

This accurate data allows significantly more power to be extracted from an electric motor, and it works entirely without additional hardware – only through AI.

TempAI is based on a platform that automatically generates physically based models from measurement data and makes them operational in a very short time. Existing control units are sufficient, as the AI models used require low computing resources. This leads to very cost-efficient implementation in series production.

This technology enables us to further increase the efficiency and reliability of our drives. At the same time, TempAI demonstrates how data-driven development can be not only faster, but also more sustainable and more powerful.

—Dr. Stefan Sicklinger, Head of AI, Digital Engineering, and Validation in R&D, ZF

The AI-based technology is ready for series production and available for the new generation of ZF electric motors.

More precise temperature prediction enables more targeted control right up to the thermal operating limit. The result: up to six percent more peak power and a verifiable increase in efficiency in the WLTP cycle. During dynamic driving – as for example, on the Nürburgring Nordschleife – energy consumption is reduced by 6 to 18 percent, depending on the load point.

In addition to performance, TempAI also offers ecological and economic advantages: the optimized thermal design allows significant quantities of heavy rare earths to be saved. At the same time, the development time per project is significantly reduced: from several months to just a few days.

During the development of electric drives, AI helps to understand and record processes inside the electric motor for which there is no physically reliable model for cost or time reasons. The challenge: the temperature inside a rotor can only be measured directly during operation at high cost.

However, there is a wealth of measurement data that is systematically recorded during extensive functional tests on the test bench and later in the test vehicles. This includes temperature values from the environment, such as from the oil pan, and the rotor speeds. The various possible operating points and their temporal progression result in millions of data points. These depend on whether and when drivers call up full power or coast along at walking pace. AI algorithms are “trained” to filter out precisely those dependencies that are particularly significant for temperature changes in the rotor and stator.