Development of a multi-GPU solver for atmospheric entry flows with gas-surface interactions

The understanding of atmospheric re-entry is fundamental in the aerospace engineering field. The heat load experienced by a space vehicle while entering in the atmosphere is extreme and its correct prediction is necessary in the view of an appropriate design of the thermal shield. Technology progression allows to exploit sophisticated facilities able to reproduce the macroscopic features of entry flows. However, high fidelity experimental reproduction are still hard due to two main reasons, namely the cost of an experiment and the difficulty in reproducing each aspect of the flight conditions. This led many companies to invest more and more in numerical tool, representing a valid alternative to provide accurate prediction of interesting information, such as heat flux, pressure distribution or shock stand-off distance. Of course, the development of an efficient numerical tool is not trivial and requires particular attention. Indeed, dealing with hypersonic flows, one must account for 'real gas' effects, known as non-equilibrium phenomena. By the years, many researchers have been devoting efforts to the development of physical models able to describe the correct evolution of the challenging conditions encountered during the re-entry. The high velocities of a space vehicle induce the formation of strong shock waves in front of it, across which the temperature reaches values of the order of 10000 K. It is immediate to understand that these extreme conditions implicate the conversion of the kinetic energy into internal energy, whose total content involves translational, rotational, vibrational and electronic modes. Also, molecular dissociation occurs due to the particle collisions in the shock layer and, if the temperature is large enough, ionization occurs. The latter is a relevant aspect of re-entry flows as the presence of electrons in the mixture is responsible for the well-known blackout. For the purpose of heat mitigation, several strategies are adopted. The employment of ablative material for the Thermal Protection System (TPS) has become very common. Thanks to material degradation, the heat flux on the surface of the vehicle is reduced, even if this introduces further complexity in the numerical modeling. The material directly interacts with the species in the mixture, leading the the occurrence of gas-surface interactions (GSI) such as catalysis and ablation. Classical numerical approaches exploit finite-volume method applied in a body-fitted multi-block grids, very common in Computational Fluid Dynamics (CFD). Nevertheless, when dealing with complex and/or moving geometries, the employment of body-conformal domains can be complicated due to the need of run time remeshing procedures. In this context, Immersed Boundary Methods (IBM) are suitable for a more versatile numerical solver. Such an approach allows for a unique Cartesian grid generation, that can be refined in the most critical region to increase the accuracy of the numerical solution. Taking into account all the above mentioned phenomena is a complex task as the numerical model employed must be accurate and cheap at the same time. Indeed, given the huge computational cost required by these kind of numerical simulations, an affordable strategy must be thought in order to speed-up the calculations. Graphics Processing Units (GPUs) provide high performances for general purposes in the scientific field. NVIDIA Corporation is still actively working in the development of efficient interfaces between hardware and software. The most famous one is Compute Unified Device Architecture (CUDA) that allows a very easy interface with basic programming languages such as C/C++ or Fortran. Thanks to GPU programming, very fast simulations are possible even in the most demanding configurations. All the aforementioned aspects are addressed in this manuscript, which aims at illustrating the main challenges in modeling hypersonic flows. A comparison of the current tools is presented for interesting aerospace applications, with the hope it can inspire further developments for technology progression.