Surrogate Based Multi Objective Optimization Of Variable Stiffness Composites With Embedded Defects Induced By Automated Fiber Placement PDF Download

"In composite materials, a variable stiffness design, where the fibers are allowed to follow curvilinear paths, gives freedom to tailor the properties in directions that are more favorable for carrying loads within the laminate. As a result, it yields improved structural performance and provides flexibility for a trade-off between structural properties. A variable stiffness composite laminate can be manufactured using Automated Fiber Placement (AFP) technology. Due to the characteristics of the manufacturing process by AFP, certain defects in the form of gaps and/or overlaps emerge within the composite laminate. The present thesis describes an efficient tool developed to design and optimize variable stiffness laminates with embedded defects manufactured by AFP.First, simultaneous maximization of the buckling load and in-plane stiffness is investigated for a variable stiffness laminate, where semi-analytical analysis is used to evaluate the structural properties. The results show that the use of a curvilinear fiber path allows the simultaneous improvement of both buckling load and in-plane stiffness of a variable stiffness laminate compared to a quasi-isotropic laminate. It is found that the optimum direction of varying the fiber angle depends on the loading and boundary conditions. For a plate with free transverse edges, the optimum direction is perpendicular to the loading direction, while for fixed transverse edges it is along the loading direction. It is demonstrated that the minimum turning radius constraint imposed by AFP significantly reduces the maximum achievable improvement in the buckling load of a variable stiffness laminate. Second, a novel method, namely Defect Layer Method, is proposed to capture the location and geometry of defects and build the finite element model of a variable stiffness laminate with embedded defects. The method is then used in a parametric study to evaluate the impact of design and manufacturing parameters on the amount of defects within the laminate. The results show that using a one-sided tow drop strategy results in a significant lower amount of gaps and overlaps within the laminate compared to a both-sided tow drop strategy. It is demonstrated that using the largest number of tows with the smallest tow width yields the minimum gap and overlap area percentages within a variable stiffness laminate.Third, a surrogate-based optimization algorithm is proposed to alleviate the high computational cost of the optimization of a variable stiffness laminate. To select an appropriate metamodel, the performance of alternative metamodeling methods is compared using multiple criteria. A metamodel performance is assessed in terms of its accuracy and robustness in constructing an approximation of the buckling load and in-plane stiffness, and its suitability for integration into an optimization framework. It is found that in terms of accuracy, both Kriging (KRG) and Radial Basis Function (RBF) provide the best results, where in terms of robustness KRG has been slightly better than RBF. In terms of suitability, KRG has shown the best performance for problems with a low number of design variables, whereas RBF has been the most appropriate method for a high number of variables.Finally, the surrogate-based multi-objective optimization is used to simultaneously maximize the buckling load and in-plane stiffness of a variable stiffness laminate with embedded defects. The results show that if the effect of defects is factored in the optimization problem, for a complete gap strategy the defect-free Pareto front shifts towards lower in-plane stiffness and buckling load, whereas for a complete overlap strategy the defect-free Pareto front moves towards higher structural properties. The difference between gap and overlap modified Pareto fronts is largest for laminates that provides the maximum buckling load improvement and decreases as the layup becomes closer to the case of maximum in-plane stiffness." --