Adaptive surrogate-based harmony search algorithm for design optimization of variable stiffness composite materials

摘要

Having the advancement in automated fiber placement (AFP) and three-dimensional (3D) printing, fabricating novel composites reinforced with curvilinear fibers, namely variable stiffness composites (VSCs), have gradually become feasible. Contrasted to the traditional straight fiber-reinforced composites, VSCs possess significantly enhanced mechanical properties and increased mechanical performance. However, the optimization of a VSC panel is still a challenging problem, which can be attributed to the high non-convexity of the feasible design space and highly involved computation resources. To accelerate the design and optimization of VSC structures, in this study, we propose an effective and efficient adaptive surrogate-based harmony search algorithm (ASBHSA). The VSC is modeled using the fiber path methodology through which the fiber trajectory is generated by a mathematical formulation. In addition, the manufacturability and quality of the product are guaranteed by considering the curvature constraint. For the ASBHSA, a distance-based infill criterion in combination with a radial basis function (RBF) is proposed to guide the adaptive sampling. The global best harmony search (GHS) with a good ability to find the global minimum is integrated into the developed optimization framework. After solving some popular benchmark optimization problems as well as making comparisons of the optimal designs of VSCs with those in published works, the proposed optimization framework is demonstrated being able to effectively and efficiently find the global minimum. Then, a series of designs and optimizations of laminated VSC plates under various boundary conditions and different loading conditions are carried out systematically. The reported interesting findings may provide some insights and guidance in the design of laminated VSC composites. Moreover, our proposed optimization model provides a useful tool for solving various kinds of optimization problems. (C) 2021 Elsevier B.V. All rights reserved.