Abstract
The aim of this study is to introduce a topology optimization approach to improve the driving force of magnetic actuators along with their thermal conductivity considering the nonlinearity of composite materials. The anisotropic magnetic composite is composed of two parts, taking into account differences in magnetic saturation effect and thermal conductivity. The first part has low magnetic reluctivity and high conductivity, while the other part has high reluctivity and low conductivity. The representative volume element (RVE) method and deep neural network (DNN) were used to obtain a dataset of effective composite material properties and generate a machine learning (ML) module for material property determination used in the optimization process. To optimize and verify both performances, a multi-objective function was established. By employing gradually changing preferences with an initial and utopia points-based adaptive weighting method, design processes were performed to obtain Pareto-optimal solution sets evenly distributed in the objective space. Numerical examples are presented for both symmetric and asymmetric magnetic actuator models, aiming to validate the effectiveness of the proposed design process. To investigate the effects of nonlinearity in magnetic material properties, design results are compared when subjected to high and low currents.