Automated optimization of an IPM synchronous machine

In cooperation with Siemens Large Drives, CADFEM has magnetically optimized the rotor of an IPM (Interior Permanent Magnet) synchronous machine. The goal was to find an improved motor design in an automated way and, in parallel, to develop a workflow that can be applied at Siemens for further projects. With regard to the desired design, several aspects were to be fulfilled. Both a low cogging torque and an induced stress that is as sinusoidal as possible in an idle state were to be achieved. At the operating point, a high torque was also desired, while maintaining current limitation. Furthermore, a low torque ripple and a high power factor (cosφ) were to be achieved.

With the jointly developed workflow, an optimized design variant with improved operating parameters was found for Siemens Large Drives, all within the same installation space. The total harmonic distortion (THD), a measure of the harmonic content of a harmonic signal, was reduced by 50 percent, the power factor was improved from 0.84 to 0.88, and the output torque - with the same current supply - was increased from 2540 to 2770 newton meters. With the behavioral model that applies to the entire design space, Siemens can very quickly examine and evaluate a wide variety of optimization goals, weightings and constraints for this machine. In addition, the jointly developed workflow can be transferred to further use cases in order to then also analyze various operating scenarios.

A parametric model was created in ANSYS Maxwell based on the initial design. The parameter limits were chosen in such a way to ensure a producible geometry of the sheet metal cut. Using a time-transient FE model, several working points were defined in order to calculate the result variables of interest. The coupling of ANSYS Maxwell with ANSYS optiSLang made it possible to optimally represent the possible design space using statistical variation and also to perform a sensitivity study. The study provided a clear indication of which parameters have which influence on the result variables. The data from the sensitivity study formed the basis for creating a behavioral model over the entire design space. An optimization was used to identify a desired design and recalculated in Maxwell for verification purposes.