Optimization of the laminated composite beam with box section under pure shear load

Abstract

This study optimizes the buckling load in composite beams with box sections under pure shear load. A combination of finite element methods and optimization techniques was employed to develop an optimal model. A beam with a one-sided boundary condition, a tenlayer carbon-epoxy configuration was modeled to achieve this. The buckling load under shear force with critical limitations such as material, ply angles from 0 to 90 degrees, the number of wall layers, and the thickness of all section walls are considered in this study. We model this problem as an optimization one tackled by two population-based metaheuristics. The one is the well-known genetic algorithm and the other is the Bayesian Optimization which is powerful when computing the values of the objective function on the candidate solutions that are expensive or unavailable. In the case of applying the genetic algorithm, the mathematical formulation of the objective function is molded as the regression task. In contrast, in the case of applying the Bayesian optimization technique, we do not need any mathematical formulation of the objective function, and use the Abaqus simulator instead. The findings of this research showed that by changing the configuration of the layers, the buckling load capacity after optimization with 100 simulation calls can increase up to 1.5 times compared to the initial values under shear for asymmetric configurations.