Aplicação de Inteligência Artificial na Otimização e Análise de Incertezas em Materiais Compósitos Laminados

Abstract

The demand for high-performance projects has led to the utilization of composite materials due to their exceptional mechanical properties and lightweight nature. However, customizing these materials to meet specific project requirements and dealing with manufacturing variability pose challenges in finding the optimal parameter combination. To address these uncertainties, there is a need to develop robust optimization methods that consider design and parameter uncertainties. In this study, a methodology and computational tool were developed to optimize laminated composite materials, considering uncertainties in their properties. This approach reduces design time and costs while providing insights into the response of structures to uncertainties and specific application demands. The methodology combines artificial intelligence techniques with a structural analysis program. Genetic Algorithm and Fuzzy Logic are employed to optimize and analyze uncertainties, while a Structural Analysis Program (PAE) based on finite element methods provides mechanical responses. The results obtained by the PAE were compared with previous studies, including Navier’s closed method based on the Third Order Shear Deformation Theory (TSDT) and the Inverse Hyperbolic Shear Deformation Theory (IHSDT), known for their accuracy. The calculated displacements showed an average deviation of only 1.22% compared to literature values, with even smaller deviations on thinner plates and with more layers. Similarly, the average difference in stresses was less than 1%. These findings demonstrate the excellent precision of the developed model in estimating displacements and stresses, especially in thin plates composed of multiple layers. Using the Genetic Algorithm for optimization, the search space consisting of 65,536 possible combinations for the laminate was explored to find structures with lower weight and vertical displacement. The algorithm demonstrated high computational performance, with a probability greater than 95% of identifying all optimal combinations in a single run. Out of the search space, 251 optimal structures were identified, forming the Pareto Optimum set. The results highlight the robustness and efficiency of the developed methodology, particularly in situations where information about materials and their application is uncertain or imprecise. This approach has significant potential in addressing complex design problems for composite structures.