AI-Enhanced System Level Design and Optimization of Inertial Fusion Energy (IFE) Power Plants using the Integrated Process Model (IPM)
Explore the source record for details and available documents.
Nuclear science and engineering↗
SEARCH · Search NASA
Results for “Integrated Modeling”
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
An Overview of Electric Vehicle Load Modeling Strategies for Grid Integration Studies
The adoption of electric vehicles (EVs) has emerged as a solution to reduce greenhouse gas emissions in the transportation sector, which has motivated the implementation of public policies to promote their use in several countries. However, the high adoption of EVs poses challenges for the electricity sector, as it would imply an increase in energy demand and possible impacts on the power quality (PQ) of the power grid. Therefore, it is important to conduct EV integration studies in the power grid to determine the amount that can be incorporated without causing problems and identify the areas of the power sector that will require reinforcements. Accurate EV load patterns are required for this type of study that, through mathematical modeling, reflect both the dynamic behavior and the factors that influence the decision to recharge EVs. This article aims to present an overview of EVs, examine the different factors considered in the literature for modeling EV load patterns, and review modeling methods. EV load modeling methods are classified into deterministic, statistical, and machine learning. The article shows that each modeling method has its advantages, disadvantages, and data requirements, ranging from simple load modeling to more accurate models requiring large datasets.
Computer Science↗
SimH 2 : an integrated techno-economic modeling framework for hydrogen pipeline infrastructure and network optimization
Large-scale hydrogen (H 2 ) pipeline transport design and network optimization have seldom been reported due to the lack of a cost model accounting for the relationship between transport cost and hydrogen mass flow rate. Here, this work introduced a system-level cost model for hydrogen pipeline transport at supercritical state and integrated it with an existing CO 2 pipeline network tool, SimCCS, for hydrogen-specific pipeline design and optimization. The Intermountain West (I-West) region of the U.S., historically dependent on fossil fuel-based economies, is chosen to demonstrate the capabilities of our H 2 pipeline cost model and transport network optimization platform called SimH 2 . Two scenarios are examined: one where the pipeline is not allowed to pass through disadvantaged communities and the other where it is permitted. The results highlight that incorporating disadvantaged-community constraints lead to longer pipeline routes and increased transport costs, reflecting the trade-offs involved in equitable infrastructure development. It is demonstrated that the newly developed SimH 2 tool not only enables the efficient design of H 2 transportation pipelines but also optimizes the network by accounting for local terrain and the presence of disadvantaged areas.
08 HYDROGEN↗
Experimental observation and integrated modelling of proton-beryllium fusion in He and D plasmas at JET
Validated integrated modelling of JET ITER-like wall experiments in which fusion performance is driven by reactions between fast ions and intrinsically present metal wall impurities is presented. A steady-state L-mode plasma with dominant proton-beryllium fusion and neutron yields of up to ≈ 6·10 13 s -1 is developed in He and D, via radiofrequency heating of a H minority. The fusion drive is unambiguously confirmed by the neutral particle analyser, fast ion loss detector, and γ-ray diagnostics. Experiments are analysed via an integrated modelling framework, developed to model the two-stage proton beryllium-fusion chain and produce high-fidelity fusion product source terms. The modelling chain comprises TRANSP and JETTO for plasma core modelling, LOCUST for full orbit product tracking and collisional slowing-down, DRESS to resolve two- and three-body fusion kinematics, and MCNP for neutron transport calculations. Modelling shows that the primary 9 Be(p,n) 9 B reaction is the dominant neutron emitter at naturally present concentrations of beryllium in these experiments. The yield contribution of secondary reactions between fusion products and beryllium, 9 Be(d,n) 10 B and 9 Be(α,n) 12 C, is found to be negligible. The proton-deuteron knock-on effect in D plasmas is modelled, which is calculated to contribute ≈ 25% to the total neutron yield. For both He and D discharges the total computed neutron rates match fission chamber (FC) measurements within the combined experimental and computational uncertainty, with an average discrepancy of ≈ ± 20%. Realistic proton-beryllium neutron sources are propagated through JET’s MCNP neutron transport model which shows that 235 U FCs’ response is sensitive to p–Be source changes, with up to ≈ 10% variation compared to a D–D neutron source. We show that the high-energy tail of the fast proton minority can be studied with multi-foil neutron activation. The framework is also applied to the study of interactions between fast protons and boron impurities, of relevance to ITER. We calculate that in JET conditions a significant alpha source with DT-like energies could be generated through 11 B(p, α)2α fusion, and detected via γ-emission in secondary interactions between fast alphas and boron. The work represents an important step towards validating predictive integrated modelling capabilities for non-standard fusion reactions.
JET↗
Integral Nuclear Data and Benchmarking Needs for Fusion Energy Systems
Fusion energy systems are currently being designed and optimized using radiation transport codes. To deal with the unique environment inside a fusion-based system, many of these designs incorporate novel materials able to withstand the high radiation fields, ensure adequate cooling and thermal protection, and produce tritium. Validation plays a vital role in building trust in the predictive power of these models and computational methods. Validation of a code consists of modeling documented real-world experiments and comparing the code-predicted response to the measured response. Adequate validation requires measured responses from real-world experiments, also known as integral data, that mimic the system being designed, including materials, impinging radiation, and temperature, among other variables. The most trusted integral data are experimental responses that have been through a rigorous benchmarking process that develops a recommended computational model and evaluates all experimental uncertainties. Finally, there are a few research groups around the world that have been producing integral data for fusion applications, but a substantial investment is needed to address the unique validation needs of the fusion community.
Fusion↗
Integrated Neutronics Modeling for Inertial Fusion Energy Systems: Development and Application to LD-FIRST
Lawrence Livermore National Laboratory (LLNL) is proposing a new Laser Driven Fusion Integration Research and Science Test Facility (LD-FIRST) with the goal of providing an experimental testbed for future Inertial Fusion Energy (IFE) systems. However, IFE systems require detailed and accurate multiphysics modeling to quantify material damage, thermal loading, and tritium breeding within complex chamber environments. This article presents the first step in an integrated multiphysics framework that couples meshed CAD-based geometry within Monte Carlo neutronic simulations to enable high-fidelity analysis of IFE chamber concepts, with future coupling to external codes. The neutronics workflow utilizes OpenMC and its third-party capability to use CAD-based geometries through DAGMC and tally on unstructured meshes with Libmesh to evaluate neutron transport behavior, geometric fidelity, and material performance under reactor-relevant conditions. The use of tailored tallies on unstructured meshes in this framework allows direct transfer without interpolating to CFD simulation tools. Two IFE chambers were evaluated, both conceived by LLNL: HYLIFE-II and Laser IFE (LIFE). This work produced high-fidelity conformal surface and volumetric meshes of the HYLIFE-II and LIFE chambers with mapped spatial insight into material damage, thermal loading, and tritium breeding. The HYLIFE-II model was built utilizing available resources and used as a test case to verify that the neutronics framework can handle complex geometries. The LIFE chamber CAD was provided by LLNL and was the main focus of this work. This work analyzes multiple ternary alloy breeding materials for the LIFE chamber, across different 6 Li enrichments to produce data relevant to the LD-FIRST project. This work also investigates the level of model fidelity for the LIFE chamber, and results show that inclusion of detailed first wall and coolant structures increased the predicted tritium breeding ratio (TBR) by ~30%, highlighting the sensitivity of tritium breeding and the need for a high-fidelity simulation framework for IFE chambers. These developments provide a scalable toolset for the design and optimization of next-generation IFE chambers, forming a solid foundation for future coupled multiphysics analysis.
70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
Time-dependent scenario modeling for the ST-E1 fusion power plant
ST-E1 is a low aspect ratio fusion power plant being designed by Tokamak Energy targeting 1.5 GW of fusion power. Characterization of the ST-E1 flat-top scenario is described elsewhere McNamara et al (2026 Nucl. Fusion 66 086008); here we focus on addressing the question of how to ramp-up the ST-E1 plasma from an initial state following breakdown and flux-surface formation to the target flat-top state. Being low-aspect ratio, the available solenoid flux of ST-E1 is limited. Therefore, particular consideration is placed on developing ramp-up scenarios that predominantly use inductive flux provided by external vertical field coils. Through time-dependent modeling with METIS, we show that this is possible when the ramp-up is performed at relatively high plasma density: although auxiliary current drive efficiency is reduced, this is significantly outweighed by (1) higher electron-ion collisional equilibration, (2) higher fusion power once ions become sufficiently hot, (3) higher poloidal beta for increased vertical-field flux, and (4) potentially favorable exhaust compatibilities. Ultimately, we show the target ST-E1 flat-top performance can be reached after a ramp-up period lasting 150 s using less than 40 Vs of solenoid flux (with vertical field providing ∼ 90 Vs of flux). The sensitivity to model assumptions are presented, with the general observation that deleterious effects can be mitigated through minor alterations of the auxiliary power temporal waveform and/or total auxiliary power level. The impact of a solenoid and the auxiliary power mix (electron cyclotron heating only versus electron and ion cyclotron heating) on the ST-E1 ramp-up success are also discussed in appendices. On this latter topic, we show that the effect of direct-ion heating during ramp-up is obscured by the uncertainty in the pedestal dynamics, identifying a clear line of future work required to make a definite decision on the ST-E1 auxiliary power mix.
current ramp-up↗
Validation Data for Benchmarking Wire Arc Additive Manufacturing Process Simulations
Residual stresses cause geometric distortion and affect mechanical performance of additively manufactured structures, yet they are notoriously difficult to assess and predict. Distortion (warpage) can drive parts outside dimensional tolerance limits, leading to part rejection or rework. For parts that meet tolerance, locked-in residual stress fields can affect structural integrity during operation, particularly subcritical cracking by fatigue, creep, or corrosion. This work develops benchmark data for a common additive manufacturing process (Wire Arc Additive Manufacturing) that can be applied for calibration and validation of physical process models that predict residual stress fields. The work includes design of two different samples of differing geometry, detailed manufacturing records for a set of physical samples, and an extensive set of residual stress measurement data developed using two diverse techniques (the contour method and neutron diffraction). An initial application of the work is also reported, where a modeling challenge was issued to secure residual stress model predictions from two independent laboratories that were blind to residual stress measurement data. These initial blind residual stress predictions show significant discrepancies relative to the measurement data, illustrating the potential value of the underlying validation data. An open repository for this work, including the sample designs, manufacturing process records, and the residual stress data, is also provided for future application in non-blind validation efforts.
36 MATERIALS SCIENCE↗
Computational modeling of coupled mechanical damage and electrochemistry in ternary oxide composite electrodes
Performance degradation of ternary layered oxide cathodes largely originates from their loss of structural integrity in cyclic usage. Mechanical damage, such as intergranular fracture of the active particles, is not only a mechanical cleavage process but also interferes with electrochemical kinetics such as infiltration of liquid electrolyte, surface corrosion of the constituent primary particles, and may eventually isolate the primary grains from the electron conducting network. Here, in this work, we develop a computational framework that integrates electrochemistry of a LiNi x Mn y Co 1−x−y O 2 (NMC) composite cathode with mechanical damage of the active particles. To fully examine the intricate chemomechanical behavior of the electrode, we evaluate the effects of the anisotropic material properties, the influence of mechanical potential on Li transport, and the concurrent intergranular fracture and electrolyte penetration along the grain boundaries upon multiple cycles. Electrolyte infiltration benefits capacity retention but aggravates further mechanical damage by corrosion. Structural failure mostly occurs in the first charging due to the anisotropic mechanical strain between the primary grains, while the resulting damage remains stable in the later few cycles. The results are consistent with experimental observations and the integration of electrochemistry and mechanical failure enables a step further understanding of the complex mechanism of battery degradation.
Battery degradation↗
Joint scheduling of energy, fast and primary frequency response reserves in integrated transmission–distribution networks
Inverter-based distributed energy resources (DERs) connected to distribution networks (DNs) can provide fast frequency support, but their reserve deliverability depends on feeder constraints and differs from synchronous primary frequency response (PFR). Existing transmission–distribution coordination studies usually treat reserve generically or neglect feeder-level feasibility, while frequency-security scheduling studies rarely represent distribution feeders explicitly. This paper develops a bi-level day-ahead scheduling framework for integrated transmission–distribution networks that jointly clears energy, transmission-side PFR, and distribution-side fast frequency response (FFR) under exogenous hourly inertia and largest-loss inputs from an external unit commitment (UC) schedule. The transmission problem is modeled with DC-optimal power flow (OPF) and closed-form second-order cone (SOC) frequency-security constraints, whereas each DN is represented by a reserve-aware branch-flow AC-OPF so that scheduled fast reserves remain deliverable during activation. The bi-level problem is reformulated through Karush–Kuhn–Tucker (KKT) conditions into a mixed-integer SOC program, and a penalty term is used to tighten the distribution-network relaxation. In the reduced test system, lower exogenous inertia increased the required primary response from 179.64 MW to 191.08 MW, distribution-side fast response reduced total frequency-response procurement by up to 4.9%, and neglecting distribution constraints overstated the combined distribution-side energy and reserve award by up to 18%. In the expanded study, the largest case was solved in 2.02 s with a 0.00% optimality gap. Time-domain simulations kept the frequency nadir above 59.0 Hz in all tested hours. These results demonstrate the value of fast-response modeling and distribution-feasible reserve delivery in coordinated market clearing.
Noh, Seung-Gil↗
Hazard and risk analysis framework for nuclear power plant–based integrated energy systems
Employing integrated energy systems (IESs) with nuclear power plants (NPPs) can improve NPP utilization by leveraging dedicated thermal and electric power delivery, but it may also increase operational safety risks. This paper presents a framework to identify and quantify hazards and risks for such IESs. The framework combines accidentology to review past industrial accidents with failure modes and effects analysis (FMEA) to identify potential future incidents. Hydrogen explosion and toxic chemical release hazards are of particular concern. Explosion consequences are quantified using the Bauwens-Dorofeev (Bauwens) and trinitrotoluene equivalent mass (TNT-EM) methods, while chemical release consequences are computed using the Gaussian atmospheric dispersion method. Operational disturbances from direct electrical and thermal integration that may affect NPP safety are modeled using probabilistic risk analysis (PRA). Hazards and risks are then evaluated for regulatory compliance. The framework is applied to IESs comprising pressurized or boiling water reactors supplying three levels of thermal and electrical power to industrial customers. Case studies include high-temperature steam electrolysis hydrogen plants of varying capacities and a synthetic fuel production plant. Sensitivity analysis examines piping component failures in the PRA model as a precursor to cost estimation for thermal extraction line design. Additionally, Fussel-Vessely (FV) and risk increase importance (RII) measures identify risk-informed design improvements for the thermal extraction system. FMEA highlights hazards such as loss of offsite power, prompt loss of electrical load, loss of thermal output, and immediate steam diversion, in addition to hydrogen explosions and toxic chemical releases. Both Bauwens and TNT-EM methods suggest maintaining several hundred meters of separation between the NPP and hydrogen facility to mitigate explosion risks. PRA results show a maximum initiating event frequency increase of 1.15% and an overall risk increase of 0.28%. Importance measure analysis identifies upstream pipe leak isolation components as critical. Evaluating the results against safety regulations, it is concluded that hazards and risks can be managed to comply with regulations through risk-informed thermal and electrical connection designs, component selection, maintenance programs, and safe separation distances between NPPs and integrated industrial facilities.
08 - HYDROGEN↗
Case Study of Integrating High-Temperature Heat Pump with LiBr-H2O Absorption Chiller for Data Center Liquid Cooling
Data centers (DCs) are physical infrastructures that support artificial intelligence workloads. The rapid growth of artificial intelligence is putting substantial pressure on the US power grid. Most of electricity consumed by IT equipment, accounting for 50%-60% of total DC power, ultimately becomes waste heat. This heat is dissipated by DC’s cooling facilities, accounting for an additional 30%-40% of total DC power. Recovering and repurposing this waste heat offers a significant opportunity to enhance energy efficiency and reduce operating costs of DCs. One potential pathway is converting heat to cold using thermal-driven absorption chillers, therefore, reducing the power consumption in DC cooling facilities. Existing studies mainly demonstrate the technical and economic feasibility of repurposing DC’s waste heat for cooling applications but provide limited technical details on how to integrate the thermal-driven absorption chillers with DC cooling systems. In addition, the low-grade waste heat available from DCs must be upgraded to higher temperatures suitable for absorption chillers. This paper presents a case study on integrating high-temperature heat pumps with a LiBr-H2O absorption chiller to use DC waste heat for cooling. A thermodynamic model of single-effect, LiBr-H2O absorption chiller and an empirical model of high-temperature heat pumps were built. The case study considers ASHRAE W17 liquid-cooled DC, with facility service water supplied at 17.0℃ and returned at 25.3℃. The thermal behaviors of absorption chiller components were predicted for the generation temperature ranging from 75.0℃ to 115.0℃. Based on the available waste heat in the integrated system, two waste heat recovery strategies were evaluated: a facility service water-based strategy and cooling water-based strategy. Results indicated that the cooling water-based strategy achieves higher Coefficient of Performance (COPs) than the facility service water-based strategy. The relatively low cooling COPs of single-effect LiBr-H2O absorption chillers could be offset by high heating COP of high temperature heat pumps. The maximum cooling COP of absorption chiller and the overall COP of integrated systems occur at lower generation temperatures, but these conditions also yield lower cooling capacities. In practice, system operation should balance the trade-off between the COP and cooling capacity
Wang, Pengtao [ORNL] (ORCID:0000000214713429)↗
Used Nuclear Fuel Management Using the Next Generation System Analysis Model
The U.S. Department of Energy (DOE) is leading the National effort to manage the back end of the nuclear fuel cycle, encompassing the safe transportation, storage/staging, and/or eventual disposal of used nuclear fuel (UNF) and high-level radioactive waste. The Next Generation System Analysis Model (NGSAM) is DOE’s discrete-event, agent-based simulation tool designed to model the full life cycle of UNF from reactor discharge to final disposal. NGSAM supports the DOE Office of Spent Fuel and High-Level Waste Disposition by enabling a detailed, scenario-based analysis of logistics, infrastructure, and shipping strategies. NGSAM replaces legacy models with a modern, flexible platform built on Repast Simphony and enhanced by the Process Analysis Tool. NGSAM simulates the movement and interaction of individual fuel assemblies with system components such as canisters, casks, railcars, and facilities. The model integrates with the Java Transportation Operations Model to plan and execute transportation scenarios, supporting both constrained and unconstrained resource allocation. Key features include customizable allocation and acceptance algorithms, detailed facility-level operations, and a Quick Edit tool for rapid scenario adjustments. NGSAM supports multimodal transportation modeling (e.g. rail, road, barge) and provides comprehensive cost, schedule, and infrastructure data. NGSAM utilizes data from sources such as DOE’s STANDARDS UNF database and DOE’s Stakeholder Tool for Assessing Radioactive Transportation, while also allowing user-defined inputs for scenario customization. NGSAM enables stakeholders to evaluate complex UNF management strategies, assess system performance under varying assumptions, and inform decision making for future infrastructure investments. Its modular architecture and integration with other Integrated Waste Management System tools make it a critical asset for planning the safe and efficient disposition of the Nation’s growing UNF inventory.
Craig, Brian [Argonne National Laboratory (ANL)]↗
Integrated Simulation of Weld Residual Stress Evolution and Crack Propagation Using XFEM
Nuclear power plant components operate in environments that promote multiple degradation mecha- nisms, several of which involve crack initiation and growth. An ongoing effort in the U.S. Department of Energy’s Nuclear Energy Advanced Modeling and Simulation (NEAMS) program is developing a general capability within the Multiphysics Object Oriented Simulation Environment (MOOSE) framework for simulating three-dimensional crack growth under a range of driving conditions, including fatigue, stress corrosion cracking (SCC), brittle fracture, and stress-relaxation cracking. This report demonstrates an end-to-end workflow that uses this capability to model weld-residual-stress-driven SCC in the J-groove weld of a pressurized-water reactor control rod drive mechanism penetration in the vessel head. The workflow consists of a thermomechanical welding simulation with temperature-dependent plasticity, followed by cooldown to ambient conditions, and a restart of the simulation using the MOOSE extended finite element method (XFEM) module to propagate a three-dimensional crack through the residual stress field. New welding capabilities were developed to properly initialize newly activated elements in the weld region, and robustness improvements were made to the mesh-based algorithm for defining cutting planes in the 3D XFEM algorithm, allowing it to handle complex crack fronts and stress fields. Together these advances allowed the simulated SCC crack to grow from an initial elliptical flaw in the weld, across the weld, through the tube wall, and almost to the triple point (where the weld, tube, and reactor pressure vessel head intersect) over roughly 36 years of simulated service. These results demonstrate a workflow that can be extended to fully three-dimensional welding simulations and more complex crack interaction problems.
42 - ENGINEERING↗
Market participation strategy of hybrid energy resources: A New York ISO case study
Drawing on existing market designs with independent resource participation in electricity markets, this study analyzes participation models for hybrid resources combining renewable generation and storage. Two models are considered: in the first, the components operate independently, with the Independent System Operator (ISO) managing the storage state of charge (SoC); in the second, the hybrid resource acts as an integrated unit, submitting offers as a “black box” and managing its SoC internally. Using a production cost model for the zonal New York Bulk Power System, we evaluate trade-offs in system reliability, market efficiency, and asset profitability. Our results provide several key insights for policymakers, showing that the ISO-managed granular model enhances social welfare through explicit SoC management, while the simpler integrated model is more computationally efficient, but may cause more real-time violations and lower overall profits.
Bansal, Rajni Kant↗
Tracing U.S. fuel life-cycle greenhouse gas emissions in a multi-sector dynamics model using LC-GCAM
Model-based analysis of fuel pathways is essential for informing energy and environmental policy. Two major model types are typically used: multi-sector dynamics models, which capture the broader energy-economy, such as GCAM (Global Change Analysis Model), and life cycle assessment models, such as GREET (Greenhouse Gases, Regulated Emissions, and Energy Use in Transportation). Each has distinct strengths and limitations, and recent studies increasingly adopt hybrid approaches to harness the advantages of both. However, such integration is often time-consuming and complicated by inconsistencies in system boundaries and technology definitions. We present LC-GCAM, a new tool that enables estimation of life-cycle greenhouse gas emissions and primary energy use for any fuel pathway represented in GCAM. We apply LC-GCAM to 300 scenarios designed to explore key uncertainties affecting the life-cycle performance of future fuel options in the U.S. freight sector. To evaluate LC-GCAM, we compare its results with those from GREET for nine fuel types in a 2030 reference scenario. When input assumptions are modestly aligned, LC-GCAM and GREET estimates typically agree within 10% (absolute sum-based mean absolute percentage error). LC-GCAM offers a flexible and efficient approach to generating life-cycle metrics within an integrated modeling framework, supporting robust policy analysis across a wide range of interacting energy system uncertainties.
Wolfram, Paul↗
From Rules to Reasoning: A Survey of Large Language Model-Based Approaches to Scientific Hypothesis and Idea Generation
Scientific hypothesis generation represents a fundamental challenge in contemporary research due to exponentially expanding literature volumes and increasing disciplinary specialization. Large language models (LLMs) have emerged as transformative tools for automated scientific discovery, moving beyond traditional rule-based and literature-mining approaches. Four paradigmatic approaches define current LLM-driven hypothesis generation: direct prompting and fine-tuning methods, knowledge-enhanced frameworks integrating retrieval-augmented generation (RAG), multi-agent collaborative systems simulating research teams, and reasoning-focused approaches implementing cognitive architectures. Domain-specific applications demonstrate statistical equivalence to human expert performance in social psychology, experimental validation in biomedical research, and near-expert quality in astronomy. Evaluation methodologies encompass human expert assessment, LLM-as-judge frameworks, and comprehensive benchmarking systems. Technical challenges include hallucination management, knowledge integration limitations, and balancing novelty with feasibility. Future directions emphasize hybrid neural-symbolic architectures and sophisticated human-AI collaboration models for responsible scientific discovery acceleration.
AI-driven discovery↗
Improving the Performance of NEML2 with Modern Graph Compilation Backends
NEML2 vectorizes constitutive-model evaluation for large-scale multiphysics simulation, using PyTorch as its tensor backend so that a batch of material-point updates runs on CPU or GPU through a single implementation. In the two prior reports in this series it was a C++-native library, deployed through TorchScript tracing and just-in-time (JIT) compilation; it has since been rewritten from the ground up into a Python-native library deployed through Ahead-of-Time Inductor (AOTInductor), a modern PyTorch graph-compilation backend. The rewrite is driven by a persistent tension, not a language preference: NEML2 composes constitutive models at runtime from a registry of small, independently-authored pieces, and that flexibility is difficult to reconcile with the compile-time knowledge an efficient GPU kernel needs. This report documents the rewrite and the investment that accompanied it: the AOTInductor export pipeline that turns a Python-authored model into a portable, Python-free compiled artifact loadable from pure C++; the eager and compiled runtimes and the new implicit solver layer built on them; a head-to-head benchmark of legacy JIT against AOTInductor; the physics-model catalog and its worked examples; the developer tooling; and the corresponding overhaul of MOOSE’s NEML2 integration that lets MOOSE consume it. A central objective is to examine whether modern PyTorch graph-compilation backends are effective for MOOSE GPU integration. The benchmark answers directly: AOTInductor outperforms legacy JIT on every GPU scenario measured, by 1.0–4.5×. Modern graph-compilation backends are effective for MOOSE GPU integration, and AOTInductor specifically – not compilation in the abstract – is why.
Hu, Gary (Tianchen) [Argonne National Laboratory (↗