The data comes from a permanent magnet synchronous motor on a test bench. The PMSM represents an electric car prototype and multiple sensors are present on the bench.

Multiple driving cycles were performed, randomly varying speed and torque to imitate real world driving.

The following info was measured:

• ambient temperature
• coolant temperature
• rotor temperature
• temperature from multiple parts of the stator
• permanent magnet temperature
• voltage (d and q-component)
• current (d and q-component)
• motor speed
• torque (induced by current)

The most interesting features are the torque and the different stator and rotor temperatures. Especially torque that can't be easily or economically measured in a vehicle.

In this use case, we will take the stator tooth temperature, group it in 5 different intervals, and use the other columns as input features. That way, we are able to transform the problem from a regression to a classification task.

With the help of our algorithms, we are able to obtain very high accuracy. Which means, given some driving conditions we can predict what the temperature will be.