DAART: Deloitte AI-Accelerated Reentry Thermodynamics

In this paper, we present the Deloitte AI-Accelerated Reentry Thermodynamics (DAART) framework, a novel machine learning approach designed to predict key aerothermal properties for heat shields during atmospheric reentry. Traditional high-fidelity Computational Fluid Dynamics (CFD) simulations, while accurate, are computationally expensive and time-consuming. Our machine learning approach leverages a Multi-layer Perceptron (MLP) to rapidly and accurately predict aerothermal metrics using limited high- fidelity CFD data. It achieves an r2 accuracy of ~0.99 with as few as 40 datasets in under a minute, compared to a week for CFD simulations. DAART is not intended to replace CFD but to complement it by enabling strategic parameter selection for detailed analysis, thus reducing overall time and resource requirements. This framework is readily extensible to other use cases and geometries, such as the Orion spacecraft planned for NASA’s Artemis program.