Master thesis and internship[BR]- Master's thesis : Aerostructural optimization of an aircraft wing assisted by dimensionality reduction methods[BR]- Internship

Abstract

Advancements in numerical methods have increasingly allowed for the modeling and simulation of complex problems with greater accuracy. However, as these simulations grow in complexity, they become computationally expensive, especially as the growth in computer processing speed appears to have plateaued. In industrial applications, where computational resources are limited, it is essential to find techniques that reduce these costs without compromising accuracy. This thesis presents a framework that integrates dimensionality reduction methods into an aerostructural optimization process, aiming to improve the performances of the Onera M6 wing, a well-known wing that operates under transonic conditions where shockwaves occur. The wing geometry is modified using the Free-Form Deformation (FFD) method, which allows for smooth deformations based on a parameterization involving 125 design variables. However, this high-dimensional (HD) problem poses challenges for optimization, as the vast design space makes it difficult for the optimizer to locate the global maxima that would enhance the wing performances. To address this, the study employs dimensionality reduction techniques to simplify the optimization problem. Specifically, the feature selection method, particularly the filter method, is used to rank design variables by their impact on wing performances, setting aside the less influential ones. This approach reduces the dimensionality of the problem, making the optimization process more manageable and cost-effective. The research is organized into five key sections. The first section focuses on developing a Computational Fluid Dynamics (CFD) setup to accurately model the Onera M6 wing performances, using OpenFOAM to simulate the transonic, compressible, and steady-state fluid-body interaction. The second section details the FFD method and its parameterization. The third section introduces and applies dimensionality reduction methods to the problem at hand. The fourth section covers the optimization process, using two different algorithms to optimize the wing performances. Finally, the last section provides recommendations for further research and potential enhancements to the framework. This study demonstrates that even with a reduction in dimensionality, the optimization process can yield significant performance improvements while maintaining accuracy, offering valuable insights for industrial applications where computational efficiency is critical.