Multi-objective Topology Optimization of Electromagnetic Negative Stiffness Mechanism using a Deep Neural Network Based Parametric Level Set Method

To improve the magnitude of negative stiffness and reduce the non-linearity of the nested electromagnetic negative stiffness mechanism, a multi-objective topology optimization framework based on a deep neural network level set and NSGA-II is proposed. Firstly, a multi-objective topology optimization model of the electromagnetic negative stiffness mechanism is established. Secondly, an implicit level set function based on the deep neural network is constructed. Finally, a multi-objective genetic algorithm (NSGA-II) is used to solve the problem, and the corresponding topology design scheme is obtained. The simulation results show that the magnitude of negative stiffness and the linearity of the optimized electromagnetic negative stiffness mechanism is greatly improved. Specifically, the negative stiffness index has increased by 114%.