A Surrogate Modeling Framework for Airfoil Design and Optimization
Muhammad Muneeb Safdar, Apurva Anand, Koushik Marepally, Bumseok Lee, James D. Baeder · 2024
The behavior of aerodynamic geometries in a fluid environment is governed by higher order nonlinear partial differential equations, which are solved numerically through computationally intensive processes and architectures. Analyzing numerous geometric designs poses the issue of impractical requirement of computational resources due to the inherent cost associated with the analysis as well as the curse of dimensionality associated with the possible design space. These limitations present a major challenge in design and optimization of airfoil in aerospace applications. Surrogate Modeling has the potential to obviate such limitations. Surrogate models are capable of handling large data, identifying complex relationships that may not be apparent otherwise, and optimizing system designs involving large number of variables. Due to its competitive advantage over traditional tools and techniques, surrogate model is making rapid advancements in airfoil design and analysis. For this purpose, an integrated surrogate modeling framework is developed to optimize airfoil geometry using Field Inversion and Machine Learning augmented CFD data. The framework is capable to make flow predictions precisely, estimate aerodynamic coefficients accurately, converge the candidate geometries to an optimized shape efficiently, and quantify the geometric and flow uncertainties and associated effects reliably. This article aims to provide an overview of the integral elements of the developed airfoil design and optimization framework.