Computational Cost Reduction in FEM-Based Optimization of a Magnetorheological Actuator Using Memory Assistance

Magnetorheological (MR) fluids are smart materials widely used in applications requiring controllable braking torque and force generation. The optimization of such devices often relies on the finite element method (FEM) combined with a numerical optimization algorithm, incurring high computational cost due to the large number of required evaluations. This study investigates the geometric optimization of a spherical MR actuator using Differential Evolution (DE), Artificial Bee Colony (ABC), and Teaching–Learning-Based Optimization (TLBO). To reduce computational effort, Short-Term Memory Assistance (STMA) and Long-Term Memory Assistance (LTMA) were integrated into DE and TLBO. The performance of all approaches was evaluated based on optimization time, the number of fitness evaluations, and objective function values. The results show that all algorithms used are capable of finding comparable objective function values; however, substantial differences were observed in computational times, with both DE and TLBO significantly outperforming ABC. The incorporation of memory assistance reduced computational time without degrading solution quality. Unexpectedly, STMA achieved greater time reduction than LTMA. The proposed approach represents a simple and effective strategy for reducing computational cost in the FEM-based optimization of MR actuators. The proposed approach shows potential for application to other computationally intensive electromagnetic design problems, which are common in engineering.