Advanced Methods for Extrapolating Thermal Measurement Signals to Accelerate Electric Drive Testing

Inhalt:
Profound thermal analysis is crucial for the development of durable, highly efficient, and safe electric drives. However, thermal measurements are often time-consuming and can not always be replaced by simulations. To reduce testing time, this paper introduces, compares, and analyzes four different methods for accurate extrapolation of thermal measurement signals to their steady state: two analytical fit-based methods, an amplitude spectrum-based method, and a neural network-based method. The methods are then developed, parameterized, and verified by various test signals of multiple electric drives, in order to investigate their accuracy and potential for time savings. It was found that the amplitude spectrum-based method provides the highest accuracy, achieving potential time savings of up to 66% while maintaining a steady-state temperature approximation error around 1 %.