Method for Parameter Tuning of Hybrid Optimization Algorithms for Problems with High Computational Costs of Objective Function Evaluations

In this paper, the authors focus on presenting the methodology for tuning optimization algorithm parameters, with a special focus on evolutionary algorithm applications. The problem considered concerns the phenomenon of nonlinear buckling of the automotive shock absorber, which itself is solved using a commercial application of the finite element method (FEM) simulation. These analyses are usually time-consuming; therefore, the authors decided to use a surrogate model, which mimics the behavior of the actual nonlinear FEM simulation. Surrogate modeling (metamodeling) is utilized to drastically shorten the simulation time, and thus study numerous algorithm parameter combinations, allowing for tuning them and providing a robust and efficient tool for optimization. The example shown in this paper is related to the minimization of the shock absorber weight, taking into account the stability of the system. The presented method can be used in any optimization problem where the high computational cost of objective function evaluations prevents tuning of the algorithm parameters.