A Multi-Objective Optimal Design Method for Differential Evolution Magnet Motors Based on Deep Stochastic Configuration Network

This paper focuses on the multi-objective optimal design of differential evolution magnet motors (DEMM) applied to high-precision laser engraving machines. A new and highly efficient integrated multi-objective design optimization method for DEMM is proposed to achieve the optimal performance provided by high average thrust, low minimum stiffness value and low stiffness fluctuation value under different motor design parameters. First, based on finite element analysis (FEA) data, a neural network learning algorithm called deep stochastic configuration network (DSCN) is innovatively introduced to computationally solve the nonlinear modeling problem of the DEMM by plotting the nonlinear and complex relationship between the input motor structural parameters and the output motor performance. The accuracy and superiority of the DSCN is confirmed by back-validation using a conventional FEA method. Then, a fast non-dominated multi-objective evolutionary algorithm (MOEA) with an elite retention strategy, the NSGA-II algorithm, is used to search for the Pareto-optimal frontiers of the structural parameters under the best optimization performance by iterative optimization calculations of the multi-objective function, and then the optimal solution is selected by the hierarchical analytical decision-making method. Finally, finite element analysis simulation experiments demonstrate the effectiveness of the proposed method.