FEA based fast topology optimization method for switched reluctance machines

Tekgun D., TEKGÜN B., ALAN İ.

Electrical Engineering, vol.104, no.4, pp.1985-1995, 2022 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 104 Issue: 4
  • Publication Date: 2022
  • Doi Number: 10.1007/s00202-021-01453-9
  • Journal Name: Electrical Engineering
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Compendex, INSPEC, DIALNET
  • Page Numbers: pp.1985-1995
  • Keywords: Fast optimization, Switched reluctance machine, Multi-objective differential evolution algorithm, Design of experiment, Response surface analysis, Finite element analysis, MULTIOBJECTIVE OPTIMIZATION, DIFFERENTIAL EVOLUTION, MOTOR DESIGN
  • Kayseri University Affiliated: No


© 2021, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.In this paper, a finite element analysis (FEA) based fast optimization method to optimize a lightweight in-wheel switched reluctance machine is presented. This method speeds up the switched reluctance machine optimization procedure by running the FEA simulations with single-phase constant current excitations for half electrical cycle and estimating the machine performance metrics using the gathered FEA data. Hence, the machine`s dynamic performance estimation process takes shorter for each design candidate. The optimization algorithm employs designs of experiments (DOE), response surface (RS) analysis method, and differential evolution algorithm (DE). Here, the DOE method is used to reduce the search space by narrowing down the upper and lower boundaries of each design variable based on the RS analysis. Although this process does not guarantee getting the Pareto front, it places the search space close to the actual one. Hence, the multi-objective DE optimization finds the Pareto optimal solution set without requiring a large number of iterations as well as a large number of candidate designs for each iteration. The method is applied to a 24/16 SRM that is intended to be used in a lightweight race car as a hub motor. Six dimensionless geometric variables are optimized to satisfy three objective functions, namely torque ripple, motor mass, and copper loss. While the conventional DE takes at least 3000 candidate designs, the proposed method considers only 559 designs to reach a similar Pareto front. It is observed that the proposed method takes about 6 h 30 min compared to the conventional method that takes 32 h 50 min using the same computer. Therefore, the computation time is reduced almost five times with the proposed method.