4.2.3. Parameter Selection Approach

Based on the defined problem-specific output variables and the machine model of the IM resulting from the model selection methodology, the parameter selection approach is carried out. With this methodology, seven optimization parameters, which have no geometric correlation among each other, are determined. Possible optimization parameters include all geometry parameters of the IM.

Using the approach described in [27], the optimization parameters presented in Table 4 with descending elasticity are derived. Consideration of the specified lower and upper bounds on the variables reduces the size of the solution space and thus the computational effort required. A large part of the bounds results from experience. However, the upper bounds on the rotor diameter and active length can be estimated by considering the installation space limitations in combination with the diameter or length increase of the housing and, in the case of the active length, the winding head length. The lower limit of the outer diameter of the rotor follows from the shaft diameter. In the context of successive optimization, the top four optimization parameters shown are chosen as significant parameters due to their higher elasticity, and the remaining three variables are chosen as less significant parameters.


**Table 4.** Optimization parameters with associated lower and upper bounds.

4.2.4. Convergence Parameters of the Optimization Environment

For the different stages of the optimization environment, numerous adjustable convergence parameters result, such as variances, tolerances, population sizes, or the maximum number of iterations. These parameters can be used to adapt the convergence behavior of the individual stages to the problem-specific specifications of precision and solution effort. In addition, the behavior of the ANN can be influenced by settings for the network construction and the database. The definition of these convergence parameters is thereby experience-based.
