Search NASA⌕ Search

SEARCH · Search NASA

Results for “adaptive model”

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.

At least 163 records · Page 9
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Rapid Adaptation of Chemical Named Entity Recognition Using Few-Shot Learning and LLM Distillation

Named entity recognition (NER) has been widely used in chemical text mining for the automatic identification and extraction of chemical entities. However, existing chemical NER systems primarily focus on scenarios with abundant training data, requiring significant human effort on annotations. This poses challenges for applications in the chemical field, such as catalysis, where many advancements have traditionally relied on trial-and-error investigations and incremental adjustment of variables. This hinders catalysis science and technology progress in addressing emerging energy and environmental crises. In this work, we propose a few-shot NER model that can quickly adapt to extract new types of chemical entities by using only a limited number of annotated examples. Our model employs a metric-learning approach to transfer entity similarity knowledge from high-resource chemical domains (with abundant annotations) to enable effective entity recognition in low-resource specialized domains (limited annotation). We validate the effectiveness of our model on a few-shot chemical NER benchmark built based on six existing chemical NER data sets. Experiments show that the proposed few-shot NER model can achieve reasonable performance with only 5 examples per entity type and shows consistent improvement as the number of examples increases. Furthermore, we demonstrate how the proposed model can be trained with large language model (LLM) annotated data, opening a new pathway for rapid adaptation of NER systems. Furthermore, our approach leverages the knowledge broadness of large language models for chemistry while distilling this knowledge into a lightweight model suitable for efficient and in-house use.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Contrastive learning for robust representations of neutrino data

In neutrino physics, analyses often depend on large simulated datasets, making it essential for models to generalize effectively to real-world detector data. Contrastive learning, a well-established technique in deep learning, offers a promising solution to this challenge. By applying controlled data augmentations to simulated data, contrastive learning enables the extraction of robust and transferable features. This improves the ability of models trained on simulations to adapt to real experimental data distributions. In this paper, we investigate the application of contrastive learning methods in the context of neutrino physics. Through a combination of empirical evaluations and theoretical insights, we demonstrate how contrastive learning enhances model performance and adaptability. Additionally, we compare it to other domain adaptation techniques, highlighting the unique advantages of contrastive learning for this field. Published by the American Physical Society 2025

Wilkinson, Alex (ORCID:0000000253404506)↗
DOE OSTI2025

GeoLoRA: Geometric integration for parameter efficient fine-tuning

Low-Rank Adaptation (LoRA) has become a widely used method for parameter-efficient fine-tuning of large-scale, pre-trained neural networks. However, LoRA and its extensions face several challenges, including the need for rank adaptivity, robustness, and computational efficiency during the fine-tuning process. We introduce GeoLoRA, a novel approach that addresses these limitations by leveraging dynamical low-rank approximation theory. GeoLoRA requires only a single backpropagation pass over the small-rank adapters, significantly reducing computational cost as compared to similar dynamical low-rank training methods and making it faster than popular baselines such as AdaLoRA. This allows GeoLoRA to efficiently adapt the allocated parameter budget across the model, achieving smaller low-rank adapters compared to heuristic methods like AdaLoRA and LoRA, while maintaining critical convergence, descent, and error-bound theoretical guarantees. The resulting method is not only more efficient but also more robust to varying hyperparameter settings. We demonstrate the effectiveness of GeoLoRA on several state-of-the-art benchmarks, showing that it outperforms existing methods in both accuracy and computational efficiency.

Schotthoefer, Steffen [ORNL] (ORCID:00000002156965↗
DOE OSTI2025

Virtual to Physical: Reinforcement Learning to Optimize SNS Particle Accelerator Controls

Complex accelerators must have control systems that can handle dynamic nonlinear environments. This makes traditional control methods unsuitable as they can struggle to adapt to these uncertainties. This provides an ideal environment for reinforcement learning algorithms as they are adaptable and generalizable. We present a reinforcement learning pipeline that can effectively handle the dynamics of a complex accelerator. We test and prove our pipelines capabilities on multiple environments including the Spallation Neutron Source (SNS) and the Beam Test Facility (BTF) at Oakridge National Lab (ORNL). Due to the limited time available to train an online algorithm like reinforcement learning on a real accelerator, we utilize a virtual twin accelerator (VIRAC) developed by ORNL to pretrain the policy and show its ability to converge in the virtual environment. We then test the adaptability of the pretrained RL model by applying it on the real accelerator and comparing the results. Utilizing our Scientific Optimization and Controls Toolkit (SOCT) and open-source standards such as Gymnasium we create and solve for a MEBT orbit correction problem in the SNS and an emittance maximization problem in the BTF. We show how Twin Delayed Deep Deterministic Policy Gradient (TD3) can solve this optimization environment in the virtual accelerator and transfer this policy onto the real accelerator for inference and model retraining. We show how reinforcement learning can be utilized as a control system for complex accelerators and provide a model pipeline for how an implementation performs and can be adapted to new accelerator control problems.

Kasparian, Armen [Thomas Jefferson National Accele↗
DOE OSTI2026

HydraGNN_GFM_FineTuning4Materials v1.0

This repository enables fine-tuning of the HydraGNN Predictive GFM 2026 — an open-source ensemble of pre-trained graph foundation models for atomistic materials modeling, developed at Oak Ridge National Laboratory. The GFM 2026 is freely available and downloadable via Globus from the OLCF Data Constellation (DOI: 10.13139/OLCF/2562660). Starting from these pre-trained weights, this repository provides a complete transfer learning pipeline for adapting the GFM ensemble to domain-specific molecular and materials property prediction tasks. It includes: 1) Utilities for ensemble fine-tuning with task-specific output heads 2) Example pipelines for eight widely-used materials and molecular datasets 3) Tools for model adaptation and head configuration 4) Data preprocessing utilities for each supported dataset 5) Benchmarking and evaluation scripts

Ungerboeck, Linda↗
DOE OSTI2025

Population structure limits the use of genomic data for predicting phenotypes and managing genetic resources in forest trees

There is overwhelming evidence that forest trees are locally adapted to climate. Thus, genecological models based on population phenotypes have been used to measure local adaptation, infer genetic maladaptation to climate, and guide assisted migration. However, instead of phenotypes, there is increasing interest in using genomic data for gene resource management. We used whole-genome resequencing and common-garden experiments to understand the genetic architecture of adaptive traits in black cottonwood. We studied the potential of using genome-wide association studies (GWAS) and genomic prediction to detect causal loci, identify climate-adapted phenotypes, and inform gene resource management. We analyzed population structure by partitioning phenotypic and genomic (single-nucleotide polymorphism) variation among 840 genotypes collected from 91 stands along 16 rivers. Most phenotypic variation (60 to 81%) occurred among populations and was strongly associated with climate. Population phenotypes were predicted well using genomic data (e.g., predictive abilityr> 0.9) but almost as well using climate or geography (r> 0.8). In contrast, genomic prediction within populations was poor (r< 0.2). We identified many GWAS associations among populations, but most appeared to be spurious based on pooled within-population analyses. Hierarchical partitioning of linkage disequilibrium and haplotype sharing suggested that within-population genomic prediction and GWAS were poor because allele frequencies of causal loci and linked markers differed among populations. Given the urgent need to conserve natural populations and ecosystems, our results suggest that climate variables alone can be used to predict population phenotypes, delineate seed zones and deployment zones, and guide assisted migration.

Science & Technology - Other Topics↗
DOE OSTI2025

Data-Driven Mean-Corrected Recursive Estimation-Based Optimal DER Dispatch for Distribution System Voltage Control

Recent advances in smart inverters offer opportunities to mitigate adverse grid impacts caused by high penetrations of distributed photovoltaics (PV) in distribution grids, such as voltage violations. Here, this paper proposes a novel measurement-driven optimal power flow (OPF)-based distributed energy resource management system (DERMS) voltage regulation via recursive sensitivity estimation informed coordinated control of distributed PV inverters. The proposed approach leverages available grid and controllable DER measurements, eliminating reliance on system model information while being adaptive and robust to volatile operating conditions. A mean-corrected recursive ridge regression (MCRRR) algorithm is proposed for sensitivity estimation, continuously refining the sensitivity model through a closed-form solution. It effectively manages varying grid operating conditions, such as changes in power injections and topology reconfiguration, to facilitate a time-varying update of the Load Sensitivity Factors (LSF). The proposed approach is formulated as a linear programming (LP) problem and is thus scalable to larger-scale distribution systems. Its effectiveness and efficiency are demonstrated on a realistic distribution feeder with high PV penetrations in Southern California, USA.

14 SOLAR ENERGY↗
DOE OSTI2026

Analytical model for balling defects in laser melting using rivulet theory and solidification

In laser welding and additive manufacturing communities, the balling (humping) defect is primarily attributed to the Plateau–Rayleigh fluid instability (PRI) with a few authors suggesting fluid jetting and volume conservation as alternative mechanisms. As analytical descriptions of these mechanisms are unavailable, combining them into a single formalism is unfeasible. We present a new model of PRI with higher accuracy, accounting for competition with solidification, to compare the expected behavior with known experimental trends when fluid jetting is neglected. We adapt a rivulet instability model from fluid physics to account for the stabilizing effects of the substrate, which the traditional cylindrical-jet geometry does not account for and estimate the instability growth rate. Our model yields a continuous transition from non-balling to balling hitherto lacking in current literature and predicts instability growth at higher wavelengths with strong sensitivity to the solidification front curvature. While the fluid surface is most unstable for shallow melt pools, the absolute magnitude of balling relevant to printing defects scales with melt pool depth and has a maximum for a given melt pool geometry. Synchrotron-based x-ray radiography of thin samples indicates that PRI growth rates and solidification can be comparable in magnitude and thus compete, as we find in this work. We predict that deviations between model predictions and our experimental results demonstrate the importance of fluid flows and heat transport in the balling process. Our experiments further demonstrate at least one mechanism by which the melt pool length and the balling wavelength are not equivalent, as commonly claimed.

Taylor, Z. [Stanford Univ., CA (United States); SL↗
DOE OSTI2024

Enhancing Biopreparedness through a Model System to Understand the Molecular Mechanisms that Lead to Pathogenesis and Disease Transmission: NW-BRaVE

The science of biopreparedness to counter biological threats hinges on understanding the fundamental principles and molecular mechanisms that lead to pathogenesis and disease transmission. Our vision to address this challenge is to create a powerful and user-friendly platform to elucidate the fundamental principles of how molecular interactions drive pathogen-host relationships and host shifts. We will enable groundbreaking discoveries by integrating a wide range of structural, genomics, proteomics, and other advanced omics measurements, along with evolutionary and artificial intelligence predictions. To make sure the system is applicable to real-world problems, we will develop it in the context of a tractable model system, the small, abundant, and accessible photosynthetic cyanobacteria and their constantly co-adapting viral pathogens, cyanophages. This model will maintain the system’s applicability to real-world problems and techniques, but the overall focus will be on elucidating general principles of detecting, assessing, and surveilling molecular interaction, adaptation, and coevolution that are system agnostic and therefore extensible to other viral-host interactions. Our overall objectives are to (1) identify the molecular complexes that comprise the cyanobacteria redox macromolecular subsystem and how they dynamically change with bacteriophage infection in situ, using cryo-electron tomography; (2) profile regulatory changes during infection using proteomics, multiomics, and experimental validation, and integrate the data with in situ structures; (3) use genomics and metagenomics to determine environmental and population factors across time scales that impact the interactions between marine cyanobacteria and their cyanophage parasites, predicting the evolutionary origins of in situ structural and functional interactions, convergence and coevolution; and (4) develop a data integration and transformation platform that facilitates the integration of in situ, proteomic, and evolutionary measurements of molecular interactions to surveil diverse hosts and parasites in various environmental contexts. These objectives address Focus Area 2 Reveal Molecular Interactions Across Biological Scales for Design of Targeted Interventions. Our powerful and user-friendly platform will enhance connections between the often-siloed fields of structure, molecular phenotype, and evolutionary genomics that are key to biopreparedness, but in need of integration (Figure 1). We will build an integrated navigation tool to facilitate the effective use of globally distributed experimental data for integrated analysis and predictive modeling. The project will develop, implement, and test a platform to assess host-pathogen molecular interactions, adaptation to hosts and host shifts, and coevolution between hosts and pathogens, successfully impacting the research community by revolutionizing abilities to study any host-pathogen interaction, encourage diverse community contributions, and gain fundamental insights into how proteins adapt to new contexts. This ability will be critical for designing early interventions to address future threats. We will build surveillance training capability, aiming for a fair and equitable response to future pandemics and biothreats.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Adaptive Dynamic Digital Twin for Test Scenario Generation

Vehicle testing has been an important part in the development of both highly automated vehicles (HAV) and advanced driving assistant systems (ADAS). Obtaining a good representation of the Vehicle Under Test (VUT) is crucial for test scenario library generation (TSLG). Current vehicle testing methods often involve calibrating car-following models using vehicle trajectory data to create static representations that cannot be dynamically updated. For instance, when multiple vehicle trajectories are collected, it is difficult to automatically determine whether a new trajectory improves the model's representativeness or degrades its accuracy. In this paper, we introduce a dynamically updated digital twin modeling framework featuring an adaptive mechanism that evaluates new trajectory data. This mechanism can decide whether to incorporate newly collected data into the current model or create a separate digital twin model when the trajectory significantly differs from prior data. Vehicle location, speed, and acceleration extracted from the newly collected trajectory data are used to support the dynamic update decision. By integrating this digital twin model into the test library generation process, we demonstrate its ability to assist in generating test libraries while effectively handling newly collected data.

Chen, Hanlin [ORNL] (ORCID:0000000165087715)↗
DOE OSTI2025

Multifrequency Models of Black Hole Photon Rings from Low-luminosity Accretion Disks

Images of black holes encode both astrophysical and gravitational properties. Detecting highly lensed features in images can differentiate between these two effects. We present an accretion disk emission model coupled to the Adaptive Analytical Ray Tracing (AART) code that allows a fast parameter space exploration of black hole photon ring images produced from synchrotron emission from 10 to 670 GHz. As an application, we systematically study several disk models and compute their total flux density, average radii, and optical depth. The model parameters are chosen around fiducial values calibrated to general relativistic magnetohydrodynamic (GRMHD) simulations and observations of M87*. For the parameter space studied, we characterize the transition between optically thin and thick regimes and the frequency at which the first photon ring is observable. Our results highlight the need for careful definitions of photon ring radius in the image domain, as in certain models the highly lensed photon ring is dimmer than the direct emission at certain angles. We find that at low frequencies the ring radii are set by the electron temperature, while at higher frequencies the magnetic field strength plays a more significant role, demonstrating how multifrequency analysis can also be used to infer plasma parameters. Lastly, we show how our implementation can qualitatively reproduce multifrequency black hole images from GRMHD simulations when adding time variability to our disk model through Gaussian random fields. This approach provides a new method for simulating observations from the Event Horizon Telescope and the proposed Black Hole Explorer space mission.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2026

Community-Informed Urban Flood Modeling for Impact Mitigation

Climate change is intensifying the hydrologic cycle, leading to more frequent and severe rainfall-driven (pluvial) flooding in urban areas. In the mid-Atlantic US cities, aging and under-designed stormwater infrastructure is increasingly strained by these events, resulting in recurring damage to property and disruptions to transportation networks. In this study, we combine community engagement with hydrologic modeling to develop and evaluate potential urban flood adaptation strategies. Over a three-year period, local technical experts and community representatives met regularly to discuss flooding concerns, identify priorities, and co-develop adaptation strategies. These discussions informed the development of an urban flooding model (EPA Storm Water Management Model) for the Baltimore Harbor watershed, the focus location of this study. The flooding model integrates complex surface and subsurface stormwater infrastructure data, local expert knowledge, and community insights. We simulate stakeholder-prioritized adaptations, such as green and gray infrastructure strategies. Model results demonstrate that enhanced infrastructure maintenance is the most effective adaptation for reducing flood depths, but has varied effects across the watershed, and can increase flooding in some locations. Spatially concentrated greening provides limited benefit to the watershed as a whole, but moderate benefit in community priority areas. Together, these adaptations have the potential to reduce flood depths by as much as 58% in some locations, greatly reducing property damage and mobility impacts, primary concerns of stakeholders. Future work will implement robust optimization tools to search for adaptations which meet stakeholder objectives and perform highly under varied future climate conditions. This work contributes to the expanding literature on collaborative modeling and demonstrates that community-engaged approaches can enhance model credibility and generate more actionable insights for communities seeking to strengthen climate resilience.

Baltimore MD↗
DOE OSTI2026

Causality-respecting adaptive refinement for PINNs: enabling precise interface evolution in phase field modeling

Physics-informed neural networks (PINNs) have emerged as a powerful tool for solving physical systems described by partial differential equations (PDEs). However, their accuracy in dynamical systems, particularly those involving sharp moving boundaries with complex initial morphologies, remains a challenge. Here, this study introduces an approach combining residual-based adaptive refinement (RBAR) with causality-informed training to enhance the performance of PINNs in solving spatio-temporal PDEs. Our method employs a three-step iterative process: initial causality-based training, RBAR-guided domain refinement, and subsequent causality training on the refined mesh. Applied to the Allen-Cahn equation, a widely-used model in phase field simulations, our approach demonstrates significant improvements in solution accuracy and computational efficiency over traditional PINNs. Notably, we observe an ‘overshoot and relocate’ phenomenon in dynamic cases with complex morphologies, showcasing the method’s adaptive error correction capabilities. This synergistic interaction between RBAR and causality training enables accurate capture of interface evolution, even in challenging scenarios where traditional PINNs fail. Our framework not only resolves the limitations of uniform refinement strategies but also provides a generalizable methodology for solving a broad range of spatio-temporal PDEs. The enhanced performance of the RBAR–causality combined framework demonstrates its strong potential for advancing PINN-based modeling of physical systems characterized by complex, evolving interfaces.

Allen-Cahn equations↗
DOE OSTI2025

Knowledge Oriented Graph Unified Transformer (KOGUT) v0.1

KOGUT — Knowledge Oriented Graph Unified Transformer KOGUT implements the Relational Graph Transformer (RelGT) architecture for knowledge graph link prediction in biological domains, with a primary focus on microbial growth media prediction. While the original RelGT (arXiv:2505.10960) targets relational tables, time series, and multi-table databases, KOGUT adapts this architecture for heterogeneous biological knowledge graphs, providing first-in-class AI predictive models for microbial cultivation. Key Adaptations Beyond Original RelGT: - Knowledge Graph Focus: Applied to biological KGs with semantic node types (taxa, chemicals, media, phenotypes, environments) versus generic relational database tables, trained on the KG-Microbe knowledge graph (1.3M entities, 2.9M edges, 24 relation types). - Multimodal Node Encoding: Integrates node labels, categories, descriptions, and synonyms from KG metadata through learned embedding layers—adapting relational column features to graph node attributes with textual semantics. - Extended K-Hop Subgraph Strategy: Optimized neighborhood sampling (3-hop default, configurable up to 200 nodes) tuned for sparse biological networks, building on the original local-global attention framework with biological relation preservation. - Biolink Predicate Preservation: Type-specific transformations for 24 biological edge semantics (occurs_in, consumes, produces, has_phenotype, subclass_of) beyond standard relational foreign keys, enabling multi-relation link prediction. - Inductive Learning Support: Enables zero-shot predictions for novel taxa through feature-based embeddings (temperature, oxygen requirements, gram stain, cell shape), extending the original transductive relational benchmark scope to uncultured microorganisms. CheapSOTA Performance Optimizations (This Distribution): - VQ-EMA Centroid Attention: Vector quantization with exponential moving average for improved global context modeling (+5-10% MRR improvement). - HDF5 Precomputed Data Loading: One-time preprocessing of k-hop subgraphs to eliminate redundant graph traversals (2-5× training speedup). - Distributed Data Parallel Training: Multi-GPU support for scaling to larger knowledge graphs (tested on 4× NVIDIA A100 GPUs at NERSC Perlmutter). - Mixed Precision Training: Automatic mixed precision (AMP) for memory efficiency and faster training. Advantages Over Standard Knowledge Graph Embedding Models: Combines RelGT's proven multi-element tokenization (features, type, hop, structure) with graph-native biological representations, enabling interpretable link prediction across heterogeneous entities that standard embedding models (TransE, RotatE, ComplEx) and table-based transformers cannot directly model. Achieves near-perfect performance on microbial growth media prediction (MRR: 0.9966, Precision@1: 0.9932, Hit@10: 1.0000) while maintaining explainability through attention-based reasoning over biological pathways. Training Data: - KG-Microbe merged knowledge graph: 1,379,337 nodes, 2,960,472 edges - 24 biological relation types including taxonomic hierarchies, metabolic interactions, phenotype associations, and environmental relationships - Primary prediction task: Growth media suitability for microbial taxa (biolink:occurs_in, 50K edges) - Multi-relation capability: Predicts links for any of the 24 relation types, including chemical consumption/production, phenotype associations, and taxonomic classification Citation: Original RelGT Architecture: Dwivedi et al., "Relational Graph Transformer", arXiv:2505.10960, 2025 KOGUT Implementation: Knowledge Oriented Graph Unified Transformer for Microbial Growth Media Prediction Developed at Lawrence Berkeley National Laboratory (LBNL) Trained on NERSC Perlmutter supercomputer

Joachimiak, Marcin [Lawrence Berkeley National Lab↗
DOE OSTI2024

Propagule Pressure in Microbial Introductions

The use of potentially beneficial microorganisms in agriculture (microbial inoculants) has rapidly accelerated in recent years. For microbial inoculants to be effective as agricultural tools, these organisms must be able to survive and persist in novel environments while not destabilizing the resident community or spilling over into adjacent natural ecosystems. Here, we adapt a macroecological propagule pressure model to a microbial scale and present an experimental approach for testing the role of propagule pressure in microbial inoculant introductions. We experimentally determined the risk-release relationship for an IAA-expressing Pseudomonas simiae inoculant in a model monocot system. We then used this relationship to simulate establishment outcomes under a range of application frequencies (propagule number) and inoculant concentrations (propagule size). Our simulations show that repeated inoculant applications may increase establishment, even when increased inoculant concentration does not alter establishment probabilities. The dataset filed here includes the experimemtal datafile, and a RMarkdown file that includes all the code used in in both the modeling and anaylsis.

agriculture↗
DOE OSTI2025

REBOUND: Reverse Engineering Bidirectional Outflow Under Non-Equilibrium Diffusion

Rare-earth elements (REEs) are essential for electronics, renewable energy, and defense technologies. However, the current supply of REEs relies on mining concentrated in a few countries and energy-intensive separations. DOE’s Basic Energy Sciences (BES) program has launched a grand challenge which aims to ensure a sustainable supply of critical REEs by developing innovative and environmentally friendly separation methods. As an alternative to costly and harmful traditional methods, the Non-Equilibrium Transport Driven Separations (NETS) initiative has created a microfluidic Y-channel co-flow method that applies external fields to exploit magneto- and electrohydrodynamic effects for separating dilute REE ions from complex feedstocks. Computational fluid dynamics (CFD) studies have identified a few operating conditions with promising ion selectivity and separation efficiency. However, challenges remain regarding Y-channel versatility across feedstocks and accurate incorporation of physical phenomena into CFD models. In this work, we develop a multi-fidelity modelling approach which integrates experimental results with CFD simulation to build a surrogate model for the dependence of separation efficiency to variation of design parameters. The surrogate model enables a reinforcement learning (RL) method to adaptively launch CFD and experimental runs, improving model fidelity around optimal Y-channel parameters.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Forecasting fluid-injection induced seismicity to choose the best injection strategy for safety and efficiency

Induced seismicity poses a challenge to the development of Enhanced Geothermal Systems (EGS). Improving monitoring and forecasting techniques is essential to mitigate induced seismicity and thereby fostering a positive perception of EGS projects among local authorities and population. Induced seismicity is the result of complex and coupled thermo-hydro-mechanical-chemical mechanisms. Injection flux and pressure are crucial controlling parameters for both hydraulic stimulation and circulation protocols. We develop a methodology combining a hydro-mechanical model with a seismicity rate model to estimate the magnitude and frequency of mainshocks and aftershocks induced by fluid injection. We apply the methodology to the case of the Basel EGS (2006, Switzerland) to compare the effects of progressive, cyclic and constant injections on the mechanical response of discrete faults. Results from the coupled hydro-mechanical models show that the pore pressure diffusion and consequent enhancement of fault permeability are limited to the vicinity of the injection well during cyclic injection. Additionally, constant injection induces seismicity from the start of the injection but enhances the permeability of most of the faults within a shorter duration, inducing less post-injection seismicity. The methodology can be adapted to any numerical model and allows new projects to be developed by anticipating the safest injection protocol. This article is part of the theme issue ‘Induced seismicity in coupled subsurface systems’.

58 GEOSCIENCES↗
DOE OSTI2025

The Artificial Scientist: in-Transit Machine Learning of Plasma Simulations

Large-scale simulations or scientific experiments produce petabytes of data per run. This poses massive challenges for I/O and storage when scientific analysis workflows are run manually offline. Unsupervised deep learning-based techniques to extract patterns and non-linear relations from these large amounts of data provide a way to build scientific understanding from raw data, reducing the need for manual pre-selection of analysis steps, but require exascale compute and memory to process the full dataset available. In this paper, we demonstrate a heterogeneous streaming workflow in which plasma simulation data is streamed directly to a Machine Learning (ML) application training a model on the simulation data in-transit, completely circumventing the capacity-constrained filesystem bottleneck. This workflow employs openPMD to provide a high level interface to describe scientific data and also uses ADIOS2, to transfer volumes of data that exceed the capabilities of the filesystem. We employ experience replay to avoid catastrophic forgetting in learning from this non-steady state process in a continual manner and adapt it to improve model convergence while learning in-transit. As a proof-of-concept, we approach the ill-posed inverse problem of predicting particle dynamics from radiation in a particle-incell (PIConGPU) simulation of the Kelvin-Helmholtz instability (KHI). We detail hardware-software co-design challenges as we scale PIConGPU to full Frontier, the Top-1 system as of June 2024 Top500 list.

Kelling, Jeffrey [Helmholtz-Zentrum Dresden Rossen↗