Currently, the PMSM is simulated using Look-up-tables (LUTs) with linear interpolation in order to model the depency of the magnetic flux linkage on the currents in dq-coordinates.
It is to be investigated...
- how the automatic differentiation is generated for this part?
- how this affects the overall computation time of gradients that go through this component?
- what alternatives could be employed (Other types of function approximation, e.g. small NNs)?
- how the alternatives differ in terms of computational complexity?
Currently, the PMSM is simulated using Look-up-tables (LUTs) with linear interpolation in order to model the depency of the magnetic flux linkage on the currents in dq-coordinates.
It is to be investigated...