Aerodynamic Risk Assessment using Parametric, Three-Dimensional Unstructured, High-Fidelity CFD and Adaptive Sampling

TL;DR

This paper presents an automated adaptive sampling method using high-fidelity CFD to efficiently estimate the probability of rare aerodynamic events for 3D airplane models, significantly reducing computational costs.

Contribution

It introduces a fully automated pipeline for geometry generation, meshing, and adaptive sampling to accurately and efficiently assess aerodynamic risks.

Findings

Adaptive sampling reduces computational cost by hundreds of times compared to Monte Carlo.

The method effectively estimates rare event probabilities in complex 3D geometries.

Automated geometry and meshing streamline the risk assessment process.

Abstract

We demonstrate an adaptive sampling approach for computing the probability of a rare event for a set of three-dimensional airplane geometries under various flight conditions. We develop a fully automated method to generate parameterized airplanes geometries and create volumetric mesh for viscous CFD solution. With the automatic geometry and meshing, we perform the adaptive sampling procedure to compute the probability of the rare event. We show that the computational cost of our adaptive sampling approach is hundreds of times lower than a brute-force Monte Carlo method.