Advanced Computational Center for Entry System Simulation (ACCESS)

The Advanced Computational Center for Entry System Simulation (ACCESS) is a comprehensive team of world-leading experts from five universities (Colorado/CU; Illinois/UIUC; Kentucky/UKY; Minnesota/UMN; New Mexico/UNM). Our vision for ACCESS is to revolutionize analysis and design of planetary entry systems through development of a fully integrated, interdisciplinary, simulation capability employing high fidelity, validated physics models, including evaluation of uncertainty and reliability, that is enabled by innovative scalable algorithms and high performance computing. Entry Systems are essential to many of NASA’s highest priority space exploration missions, see Table 1. Based on the key attributes of each mission, the critical physical phenomena that drive Entry System design involve flow phenomena (e.g., chemistry, radiation, turbulence) and material and structural response (e.g., ablation, fracture). Table 1: Challenges for Design of Entry Systems for NASA reference missions NASA Mission Key Attribute(s) Key Physical Phenomena Quantities of Interest (QoIs) Artemis (Earth) High speed (11 km/s) Large size Air Chemistry, Radiation, Turbulence, Ablation, Fracture Surface Heating, Material Recession, Structural Failure Mars Return Entry Vehicle (Earth) Very high speed (13 km/s) Air Chemistry, Radiation, Ablation, Fracture Surface Heating, Material Recession, Structural Failure Crewed Lander (Mars) Mild speed (5 km/s) Large size and weight Tight landing ellipse CO2 Chemistry, Turbulence, Fracture Aerodynamics, Structural Failure DragonFly (Titan) Mild speed (7 km/s) Atmosphere: N2, CH4 CN Radiation, Fracture Surface Heating, Structural Failure A first significant limitation with state-of-the-art (SOA) analysis capabilities for Entry Systems is that the uncertainties associated with predicting each of the QoIs are so large that it is not always possible to close on a design cycle. For example, a margin of 100% for turbulent surface heating augmentation is typically employed for Mars entry [1], and a margin of 40% was used for radiative surface heating for lunar return at 11 km/s [2]. Such large uncertainties arise directly from limitations in the accuracy of modeling the key physical phenomena and represent a significant challenge for meeting design requirements, e.g., the Mars Return Entry Vehicle has a reliability requirement of less than 1 in 106 that cannot be met by SOA analysis capabilities [3]. A second significant challenge for the design of Entry Systems for NASA reference missions concerns the currently available analysis tools. NASA employs a number of computational codes for modeling the hypersonic flow (e.g., DPLR [4], LAURA [5], FUN3D [6]) and the material response (e.g., FIAT [7], CHAR [8], ICARUS [9]) relevant to Entry Systems. However, these tools are labor intensive to apply for analysis and design, their computational performance is limited by not taking advantage of Peta/Exa scale computer architectures, and they do not integrate uncertainty and reliability. Motivated by these two significant challenges, the ACCESS research plan involves four tightly coupled tasks that are summarized in the following. 2 Task-1: Kinetic Rate and Physical Process Modeling (Lead: Marco Panesi) Task-1 will develop models and quantify uncertainty for a number of critical physical processes induced by hypersonic entry through three sub-tasks: (1a) Gas-phase kinetics; (1b) Gassurface interaction; and (1c) Turbulence modeling. Gas-phase chemical kinetics models will be developed for the atmospheres of Earth, Mars, and Titan, complemented by the products of material ablation. Research priorities will be guided by sensitivity analysis. Electron driven kinetics and ablation chemistry are expected to be areas of particular focus. In-depth analysis of the key reactions will rely on ab-initio scattering calculations leveraging high-fidelity quantum mechanical data. The models will be validated using both new and existing experimental datasets that include crossbeam scattering (CU) and shock tube measurements (e.g., NASA EAST, Oxford T6). Gas-surface interaction modeling will rely heavily on experiments. Reaction data will be obtained using molecular beam scattering experiments (CU) on NASA-relevant materials. New gas-surface reaction models will be constructed based on the measured data with a particular focus on characterizing the associated uncertainties in the resulting model. The models will be further refined and validated by meso-scale and macro-scale experiments in flow-tube reactors (UIUC) and plasmatron facilities (UIUC). Ongoing research will be leveraged in direct numerical simulation (DNS) and wall-modeled large eddy simulation (LES) of hypersonic flows to develop optimal turbulence modeling strategies for NASA-relevant applications. We plan to focus on wall-modeled LES to balance computational cost and simulation accuracy, so that high-enthalpy effects can be included. Data from high-fidelity simulations can be used to inform and understand the range of applicability of lower-cost RANS models. Uncertainty will be quantified in (1a-1c) via close collaboration between modeling and experiments. The model uncertainty will feed into the material response modeling Task-3, and the uncertainty quantification (UQ) Task-4. The sub-tasks will also work closely with Task-2 to transition the final models to the overall integrated simulation framework. Task-2: Tightly Coupled Multidisciplinary Analysis Capability (Lead: Graham Candler) Task-2 will develop the fully coupled flow-surface-material-structural modeling capability for analysis of Entry Systems: the Integrated Simulation Framework (ISF). Four sub-Tasks will be pursued to achieve this goal: (2a) Development of the overall integrated modeling capability including the products from Tasks 1, 3, and 4; (2b) Development of coupling approaches for all relevant physical phenomena; (2c) Implementation allowing execution on peta/exa-scale computer architectures; and (2d) Investigation of innovative and non-traditional methods. The ISF will be based on the widely used CFD code US3D [10], currently in joint development between UMN and NASA Ames. The physical phenomena that may be coupled include aerodynamics, aerothermodynamics, radiation, material response, and structural response to thermal and aerodynamic loads. The proposed development will build on existing US3D plugins (implemented via inter-operable function pointers), including a finite element thermal response solver, radial basis functions for grid motion, and unsteady free-flight simulations. Non-linear sensitivity analysis methods will be integrated to guide the degree of coupling required for all relevant physical phenomena. For those that require close coupling, current approaches will be assessed and employed where satisfactory. Novel approaches will be developed for those processes where coupling is problematic. We plan to develop an exascale capability through refactoring key components of US3D. Critical time-consuming loops will be threaded, and then moved to GPU hardware [11]. An incremental approach will add performance 3 improvements to US3D as progress is made. Innovative hybridizable discontinuous Galerkin (DG) methods and Jacobian-free implicit methods will be explored as potential game-changers. Integration of the physics modeling products from Task-1 is expected to be straightforward due to widespread use of US3D and its plug-in capability [e.g., 12]. Integration of the models from Task-3 for the material and structural response of the thermal protection system will present challenges in terms of coupling time-scales, the need for robust treatment of moving and deforming surface meshes, and effective implementation on advanced computer hardware. However, integration of ICARUS [9] with US3D is underway, and this coupling infrastructure will be adopted for ISF. The integration of the UQ and reliability products from Task-4 requires the propagation of uncertainty across disparate physical phenomena in a computationally efficient manner. Plugin code allows UQ drivers to access simulation parameters without intruding in the core code, enabling non-invasive UQ analysis running concurrently with simulations. A challenge in developing ISF concerns Verification & Validation (V&V). As new capabilities are added to ISF, there will be a continuing need to V&V all of the legacy functionality along with the newly introduced capabilities. Existing US3D regression tests will be augmented to drive down verification complexity. Task-3: High Fidelity Modeling of TPS Features, Damage and Failure (Lead: Alexandre Martin) Due to the complexity of real thermal protection system (TPS) materials and the time and length scales of hypersonic flows, first principles atomistic modeling of material response is infeasible. Therefore, an innovative stochastic approach will be adopted to model the TPS using four scales (nano, micro, meso and macro) where models are built up from the smaller scale, and validated at every step using "modelable" experiments. Task-3 will develop and assess high fidelity TPS models through three sub-tasks: (3a) Macro/meso-scale modeling; (3b) Micro/nano-scale modeling; and (3c) Experimental characterization. For macro/meso-scale Task-3a, a fully coupled 3D simulation capability will be developed for anisotropic materials using the full momentum equation, multi-species gas chemistry, gasmaterial interaction, and heat transfer. The macroscale model will rely on meso-scale, stochastic properties, and be validated using high enthalpy facilities (UIUC and NASA). All meso-scale properties will be a function of the thermodynamic state (temperature and pressure), degree of char (morphology), impurities and defects, and length scale. These properties will be obtained through synthetic representative volume elements and Computed Tomography scans, based on micro-scale properties. Validation will use a suite of experiments (UK, CU, and UIUC), as part of Task-3c. Effects of surface degradation, including a melt layer, will be studied both at the meso and macro scales. The ablation kinetics data obtained from Task-1 will be integrated. In the micro/nano-scale Task-3b, single fiber properties will be derived. Oxidation models, including pitting dynamics, will be developed and the effects on the shape and structural properties will be assessed. Fiber-to-fiber contact will be studied to develop new models for de-bonding (fracture) and particle shedding to inform the meso-scale. Surface chemistry models will be constructed, using innovative phonon techniques that account for surface effects. As part of Task- 3c, the models will be validated using single fibers, or bundle of fibers experiments. Finally, the nanostructure of the fibers will be studied in order to understand the crystalline structure of the carbon, and relate the structure to the manufacturing technique, and assess the effects on fiber degradation. This theoretical approach will be combined with nanoscale in-situ experiments performed at Oak Ridge National Laboratory through user proposals. The stochastic approach developed in Task-3a will generate data that provides the mean and standard deviation for each property. This method will enable seamless integration into the Task4 2 modeling tool of uncertainty analyses at every scale, and quantify the overall errors of the models. The key TPS material and structural properties and processes will be identified using nonlinear sensitivity analyses. Finally, using the RVE approach, physical imperfections, such as damage caused by micro meteoroids and orbital debris, will be included at every scale. This capability, combined with off-nominal conditions and uncertainty analysis, will enable predictive modeling of reliability. Task-2 enables Task-3 by providing the overall analysis framework into which all of the physical models developed in Task-3 will be integrated. Task-4: Uncertainty Quantification and Reliability Estimation (Lead: Alireza Doostan) Task-4 will investigate UQ and reliability through three sub-Tasks: (4a) Development of UQ techniques applicable to the multi-physics analysis of entry systems; (4b) Integration of UQ into the core simulation framework; and (4c) Development and integration of a capability for evaluating Entry System reliability. Accurate UQ and reliability analyses of entry systems are made difficult by assumptions in the choice of physics models and their imperfect specifications (model parameters, constitutive properties, and boundary/initial conditions). While these shortcomings can be represented by uncertainties, their quantification and propagation using standard techniques are infeasible for entry systems due to the extreme computational cost and the large number of significant uncertainty sources. ACCESS therefore aims to develop enhancement to the state-of-the-art computational approaches that enable efficient evaluation of uncertainty and overall reliability. For UQ within a single area, such as TPS modeling, high order approximation techniques will be used, such as polynomial chaos and multi-wavelet representations, that allow accurate estimation of highly non-linear and non-Gaussian QoIs. Advanced sparse approximation techniques, such as compressed sensing, and sampling strategies will be deployed to efficiently explore the uncertainty space and minimize the computational cost. Key advantages of such representations include rapid generation of global (non-linear) sensitivity indices via the analysis of variance (ANOVA) to rank model inputs according to their contributions to the variability of the QoIs, and exploration of failure surfaces for reliability estimation and failure rates of O(10-7). Full system UQ and reliability analysis will rely on novel multi-fidelity and multi-level strategies to enable scalable propagation of uncertainty through the coupled, multi-physics models. Multi-fidelity modeling uses a combination of a few expensive high-fidelity simulations and a larger number of inexpensive simulations. A multi-fidelity framework provides a tractable path toward scalable UQ and reliability analysis and yields predictive accuracy comparable to the highfidelity model with a computational cost close to the low-fidelity model [13]. Such strategies also enable scalable sampling of the uncertainty space, thereby facilitating global sensitivity analysis and accurate estimation of overall system failure probabilities of O(10-4). After five years, ACCESS will deliver the first ever integrated capability for predicting entry system reliability derived from high fidelity modeling of all physical phenomena, uncertainty quantified by experimental data, and performance accelerated by high performance computing. 2. Research Relevance, Multidisciplinary Nature, Innovation, and Credibility The research proposed by ACCESS is directly responsive to all of the challenges for SOA modeling capabilities for Entry Systems. Our approach will significantly advance the modeling of all relevant flow, material and structural phenomena, reducing and quantifying uncertainty through validation and anchored by experiments. Effective design of Entry Systems must couple many of the flow and material phenomena, requiring a tightly integrated, multidisciplinary approach. Material exposure experiments conducted in unique facilities at UIUC will validate modeling at 5 UKY and quantify the associated uncertainty at CU. Validated flow models developed at UIUC and UMN will be implemented including quantified uncertainty into the Integrated Simulation Framework developed at UMN. The unprecedented opportunity for very close collaboration across the faculty and capabilities in ACCESS will significantly accelerate development of the integrated Entry Systems analysis and design capability sought by NASA. Innovative research approaches will be investigated throughout ACCESS for key activities where current methods are inadequate in terms of accuracy and/or efficiency. In Task-1, a rigorous probabilistic approach able to seamlessly combine experimental and ab-initio data will be used to construct predictive models for gas kinetics and gas-surface interaction with characterized uncertainty. In Task-2, the implementation of an advanced DG formulation in US3D has the goal of obtaining accurate heating predictions without labor-intensive re-meshing. In Task-3, the characterization of TPS properties extends far beyond the SOA by accounting for thermodynamic state (temperature and pressure), the scale at which they are applied (i.e., mesh size), the oxidation stage (morphology) as well as damage and defect density. In Task-4, the proposed multi-fidelity and multi-level approaches for UQ and reliability will enable accurate quantification of highdimensional uncertainties, global sensitivity analysis, and practical evaluation of design risk for a full entry system simulation, currently infeasible using SOA methods. These approaches will also enable the use of advanced machine learning techniques for large-scale, physics-based modeling and simulation that are currently in their infancy. ACCESS comprises an exceptionally strong and credible research team with decades of relevant experience. Many of the ACCESS participants (Boyd, Candler, Martin, Minton, Panesi, Schwartzentruber, Stephani) have received extensive NASA funding in recent years to advance the understanding of fundamental flow and material phenomena relevant to Entry Systems, including NASA Early Career awards to Panesi and Stephani, and a NASA PECASE award to Stephani. Most of these faculty have also conducted closely related flow and materials research over many years under funding from the Department of Defense and from private industry. Several of the ACCESS team (Boyd, Doostan, Panesi) have participated in multiple, large-scale, multiyear projects on UQ that were funded by the Department of Energy. 3. Institute Leadership Team and Management Plan The Institute Director will be Iain Boyd, University of Colorado. He is not serving as the PI on any other NSTRI proposal. The Director will work closely with four technical leads to develop the vision for ACCESS, and oversee the planning and execution of all research tasks: - Marco Panesi (UIUC) leads Task-1, contributes to sub-tasks (2a,2b,4a,4b). - Graham Candler (UMN) leads Task-2, contributes to sub-tasks (1c,4b,4c). - Alexandre Martin (UKY) leads Task-3, contributes to sub-tasks (2a,2b,4b,4c). - Alireza Doostan (CU) leads Task-4, contributes to sub-task (2a). The additional ACCESS participants and their sub-task areas are as follows: CU: Robyn Macdonald (1a,1c), David Marshall (3b,3c), Tim Minton (1a,1b,3c) UIUC: Harley Johnson (3a,3b), Francesco Panerai (1b,3c), Kelly Stephani (1b,3a) UKY: M. Beck (3a,3b), C. Brehm (1c,3a), H. Chen (3a), J. Maddox (1b), S. Poovathingal (1b,3a) UMN: Bernardo Cockburn (2d), Tom Schwartzentruber (1a,1b,3a), Joe Nichols (2c,2d) UNM: Hua Guo (1a,1b) All of the faculty listed above hold tenure track appointments at their institutions. As such, all participants meet the compliance requirements stated in the solicitation. UNM is a Hispanic Serving Institution (HSI) and UMN is an Asian American and Native American Pacific Islander-Serving Institution (AANAPISI). 6 4. References [1] Wright, M., Edquist, K., Tang, C., Krasa, P., Campbell, C., “A Review of Aerothermal Modeling for Mars Entry Missions,” AIAA Paper