Magnetorheological dampers optimization based on surrogate model and experimental verification

Abstract

An accurate and efficient optimization method is required to design magnetorheological dampers (MRDs). To solve the contradiction between calculation accuracy and efficiency in the existing method, an FNB optimization method comprising the finite element method (FEM) surrogate model built upon a novel NM neural network architecture and alternating stacked beluga whale optimization (BWO) algorithm was proposed. A high-fidelity approximation of FEM results with lower time costs was achieved. The corresponding dataset was obtained based on the optimization objective simplified by the multi-physics coupling FEM, and the optimized surrogate model was obtained by the first layer BWO. Extensive ablation and contrast experiments indicated the improvement effectiveness of the proposed surrogate model, which was further utilized by the second layer BWO to optimize the MRD structural parameters. The final results were corrected by FEM to further reduce the approximation error of the surrogate model. Moreover, compared to the optimization results obtained by simple scanning FEM and refined magnetic circuit modelling algorithm, the optimization efficiency and effect of FNB were better improved from a theoretical perspective, and extensive experiments further indicated that the overall relative improvements amounted to 5.819% and 7.137% on peak damping force.