Flutter Optimization of Carbon/Epoxy Plates Based on a Fast Tree Algorithm

This study focuses on optimizing carbon/epoxy laminate configurations to maximize the flutter speed of composite structures using a Fast Tree Regression algorithm. Initially, a seed dataset was created, using finite element method (FEM) modal analysis for common stack-ups used in composite fins and UAV components. The FEM analysis, based on the Lanczos algorithm for extracting modal frequencies in bending and torsion, was verified through experimental modal analysis using an AS-4/3501-6 composite system. Custom software was developed to interface with the FEA modal software, enabling the generation and augmentation of laminate dataset scenarios. The seed dataset was expanded until the coefficient of determination (R2) reached at least 0.95. Various regression algorithms, including Fast Forest Regression, Fast Tree Regression, Sdca Regression, and Lbfgs Poisson Regression, were evaluated. The Fast Tree Regression algorithm was selected for further analysis due to its superior performance. This algorithm was applied to a design space of nearly 2000 potential laminate candidates, focusing on symmetric lay-ups to avoid undesirable coupling between bending and torsion in UAV and missile control surfaces. The final optimized lay-ups, exhibit the highest Delta function values (the squared difference of modal frequencies in torsion and bending), indicating the expected highest flutter speeds. The results demonstrate the efficacy of tailored composite materials in achieving specific aerodynamic performance goals.