4.2.2. Model Selection Approach

Based on the decision parameters presented, the output variables or effects to be investigated are defined on the basis of those variables of the machine model which influence the decision parameters. This is the case for the diverse loss types of the IM, since they determine the average losses over the driving cycle. For this reason, all loss types are included as output variables to be considered. However, other output variables and effects are not defined in this use case, since none of the other variables has an influence on either of the decision parameters.

In Table 3, the output variables to be investigated are shown with the required levels of detail. Here, identical precision is required for all stages of the optimization environment, since the resulting higher computation times are acceptable in the context of this work. The illustrated levels of detail must be achieved for all operating points of the operating point matrix, so that the precision of the map calculation is ensured. The required level of detail of the ohmic losses is obtained assuming a measurement deviation of 5%, a deviation of the electromagnetic reference model of 5%, and a scatter of the optimum identified by the optimization environment of 5%. A deviation of 25% is allowed for the iron losses since they are not dominant compared to the ohmic losses.

From the model selection approach, the TH-FEM follows as the model that can represent the required levels of detail at all operating points while minimizing the computational effort. Consequently, in the context of the optimization problem considered in this work, the TH-FEM simulation is used in all stages of the optimization environment to describe the behavior of the IM.


**Table 3.** Output variables of the machine model to be investigated.
