Data-Driven Modeling of Hypersonic Reentry Flow with Heat and Mass Transfer

Abstract

The entry phase constitutes a design driver for aerospace systems that include such a critical step. This phase is characterized by hypersonic flows encompassing multiscale phenomena that require advanced modeling capabilities. However, because high-fidelity simulations are often computationally prohibitive, simplified models are needed in multidisciplinary analyses requiring fast predictions. This work proposes data-driven surrogate models to predict the flow and mixture properties along the stagnation streamline of hypersonic flows past spherical objects. Surrogate models are designed to predict the velocity, pressure, temperature, density, and air composition as functions of the object’s radius, velocity, reentry altitude, and surface temperature. These models are trained with data produced by numerical simulation of the quasi-one-dimensional Navier–Stokes formulation and a selected Earth atmospheric model. Physics-constrained parametric functions are constructed for each flow variable of interest, and artificial neural networks are used to map the model parameters to the model’s inputs. Surrogate models were also developed to predict surface quantities of interest for the case of nonreacting or ablative carbon-based surfaces, providing alternatives to semiempirical correlations. A validation study is presented for all the developed models, and their predictive capabilities are showcased along selected reentry trajectories of space debris from low Earth orbits.