SUMMIT FUN3D RETROPROPULSION DATASETS ================================================ Introduction ============ These datasets contain the results from 6 retropropulsion simulations, where a vehicle with a 16.4m diameter heat shield is using 8 rocket engines pointing in the direction of travel in order to decelerate. The ultimate goal of this research effort is to develop systems capable of delivering payloads large enough to support human exploration on the surface of Mars. The 6 datasets are comprised of 3 different free-stream Mach number conditions: 2.4, 1.4, and 0.8 (subdirectories A, B and C, respectively). For each Mach number there is a low-resolution (approximately 145M vertices) case and high resolution (approximately 1.14B vertices) case. These datasets were generated on Summit at Oak Ridge Leadership Computing Facility (OLCF) using FUN3D, a computational fluid dynamics code developed at NASA Langley Research Center. The mesh is unstructured, with a mix of tetrahedra, pyramid, and prism cells. The solution data are calculated at the mesh vertices (i.e., the nodes, also known as "vertex-centered data"). The data are written per subdomain, with successive snapshots concatenated to the same corresponding files. Science Background ================== The entry, descent, and landing (EDL) systems for NASA's eight successful landings on Mars have all relied on technology developed for the Viking missions of the mid-1970s. The most ambitious of these missions, the Mars Science Laboratory, delivered a rover the size of a small automobile to the surface of Mars in 2012. While incremental improvements to these technologies, namely rigid aeroshells, supersonic parachutes, and subsonic propulsive terminal descent, have increased payload mass capability, new approaches to EDL are necessary to support human exploration at Mars. Recent work at NASA has identified and begun developing technologies that enable delivery of significantly larger payloads to the surface of Mars. A goal of this effort is to assist in advancing the understanding of the underlying flow physics associated with novel approaches to propulsive deceleration for atmospheric descent at Mars. Specifically, this work supports NASA's efforts to characterize environments and requisite computational approaches to enable implementation of this technology into a flight system [K+:2020a]. Data ==== Each of the six cases follows a similar pattern in terms of what data are provided. We focus on the Mach 2.4 low-resolution case (subdirectory A/143M) here, the others are analogous. In the low-resolution cases the domain decomposition consists of 72 subdomains. In the high-resolution cases, there are 552 subdomains. The full domain grid for A/143M is provided in the file dAgpu0145_Fa.lb8.ugrid64. This file is in UGRID format, little-endian, binary, 8-byte reals, and with the integers written as 8-byte rather than 4-byte values. Here is the output from a simple utility showing the file header: % ugrid_show_header dAgpu0145_Fa.lb8.ugrid64 # UGRID: binary file # header_size: 56 # UGRID::INTEGER_8_FORMAT_FLAG # UGRID::FLOATING_POINT_8_FORMAT_FLAG # n_nodes: 143445568 (== n_vertices) # n_surface_triangles: 1142360 # n_surface_quadrilaterals: 17570 # n_tetrahedra: 755843155 # n_pyramids: 47306 # n_prisms: 32951050 # n_hexahedra: 0 The vehicle surface is defined by the surface triangles and quadrilaterals in this file. The full domain is decomposed into 72 subdomains. A submesh is provided for each subdomain, with the name dAgpu0145_Fa_mesh.lb4.N, where N is in [1 .. 72]. The files are little-endian, binary, with 4-byte reals and integers, e.g., for subdomain 1: % ugrid_show_header dAgpu0145_Fa_mesh.lb4.1 # UGRID: binary file # header_size: 28 # n_nodes: 2340458 # n_surface_triangles: 0 # n_surface_quadrilaterals: 0 # n_tetrahedra: 8760942 # n_pyramids: 1134 # n_prisms: 1459669 # n_hexahedra: 0 Each subdomain also has a corresponding metadata file, with the name dAgpu0145_Fa_meta.N, e.g., for subdomain 1: % cat dAgpu0145_Fa_meta.1 n_owned_tetrahedra 8006438 n_owned_pyramids 1079 n_owned_prisms 1350808 n_owned_hexahedra 0 The subdomain meshes contain volume cells (tetrahedra, pyramids and prisms, no hexahedra in this dataset) on the boundary that also exist in other subdomains, sometimes referred to as "ghost" or "halo" cells. In order to avoid processing these duplicate cells more than once, we introduce the concept of *owned* cells. Each volume cell is owned by exactly one subdomain mesh, and for a given submesh and cell type, the owned cells occur first. The corresponding metadata file specifies the number of owned cells for each type. For example, for subdomain 1 above, the first 8006438 of the 8760942 tetrahedra are owned by that subdomain. For this dataset, FUN3D wrote separate solution files for each subdomain, and successive snapshots in time were concatenated to the same corresponding files. The solution files are stored in subdirectories named with the maximum time iteration that may be written to the solution files in that subdirectory. In this dataset, each 200th time iteration is saved to disk (for the high-resolution data, every 50th timestep is saved). So, for example, subdirectory 2000unsteadyiters contains data up to time iteration 2000, and subdirectory 5500unsteadyiters contains iterations up to 5400, since iteration 5500 was not written out. For subdomain 1, we have: # Show the data file for subdomain 1 in the first data directory: % fun3d_show_native_volume_header \ 2000unsteadyiters/dAgpu0145_Fa_volume_data.1 # FUN3D::native_volume_scalar_real4: 1 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: I/O mode: BINARY_IO_MODE 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: version: 13.4-d296c34 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: n_nodes: 2340458 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: n_nodes_0: 2025135 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: variable_names[ 0]: rho 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: variable_names[ 1]: u 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: variable_names[ 2]: v 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: variable_names[ 3]: w 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: variable_names[ 4]: p 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: variable_names[ 5]: turb1 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: variable_names[ 6]: vort_mag 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: local_to_global[510944, \ 510943, 510523, ... 143442426, 143443042, 143443759] 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: header size: 18723744 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[0]: 200 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[1]: 400 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[2]: 600 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[3]: 800 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[4]: 1000 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[5]: 1200 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[6]: 1400 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[7]: 1600 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[8]: 1800 2000unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[9]: 2000 # Show the data file for subdomain 1 in the second data directory: % fun3d_show_native_volume_header \ 5500unsteadyiters/dAgpu0145_Fa_volume_data.1 # FUN3D::native_volume_scalar_real4: 1 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: I/O mode: BINARY_IO_MODE 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: version: 13.4-d296c34 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: n_nodes: 2340458 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: n_nodes_0: 2025135 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: variable_names[ 0]: rho 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: variable_names[ 1]: u 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: variable_names[ 2]: v 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: variable_names[ 3]: w 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: variable_names[ 4]: p 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: variable_names[ 5]: turb1 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: variable_names[ 6]: vort_mag 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: local_to_global[510944, \ 510943, 510523, ... 143442426, 143443042, 143443759] 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: header size: 18723744 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[0]: 2200 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[1]: 2400 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[2]: 2600 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[3]: 2800 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[4]: 3000 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[5]: 3200 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[6]: 3400 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[7]: 3600 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[8]: 3800 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[9]: 4000 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[10]: 4200 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[11]: 4400 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[12]: 4600 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[13]: 4800 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[14]: 5000 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[15]: 5200 5500unsteadyiters/dAgpu0145_Fa_volume_data.1: snapshot[16]: 5400 See read_fun3d_native_volume_field.cxx for example code for reading a field from a FUN3D native volume file. Export Control ============== This dataset has been cleared by the NASA Langley Research Center Export Control Office for release to all users, domestic and foreign. References ========== @inproceedings{K+:2020a, title = "Effects of Spatial Resolution on Retropropulsion \ Aerodynamics in an Atmospheric Environment", author = "A. M. Korzun and E. J. Nielsen and A. C. Walden \ and W. T. Jones and J.-R. Carlson and P. J. Moran \ and C. Henze and T. A. Sandstrom", author1 = "Ashley M. Korzun and others", booktitle = "AIAA SciTech Conference", year = "2020", month = "January", note = "AIAA 2020-1749", } @misc{Jones:2019, title = "Summit supercomputer simulates how humans will 'brake' \ during {M}ars landing", author = "Katie E. Jones", year = "2019", month = "October", howpublished = "\url{https://phys.org/news/2019-10-summit-supercomputer-simulates-humans-mars.html}", } Acknowledgments =============== This work was supported by resources received through 2019 Innovative and Novel Computational Impact on Theory and Experiment (INCITE) and Summit Early Science Program awards at the Oak Ridge National Laboratory Leadership Computing Facility, which is a U.S. Department of Energy Office of Science User Facility supported under Contract DE-AC05-00OR22725. The scientific objectives of the effort are in collaboration with the NASA Langley High Performance Computing Incubator and the NASA Space Technology Mission Directorate's Descent Systems Studies Project and aligned with the NASA Aeronautics Research Mission Directorate's Aerosciences Evaluation and Test Capability Challenge.