SEARCH · Engineering Papers
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.
A machine learning model and graph generator were able to accurately predict for the presence of nearly 1000 substructures and the connectivity of small organic molecules from experimental 1D NMR data.
The report documents the DOE Town Halls held during 2019 at Argonne National Laboratory, Oak Ridge National Laboratory, Lawrence Berkeley National Laboratory, and in Washington, DC. From July to October 2019, the Argonne, Oak Ridge, and Berkeley National Laboratories hosted a series of four town hall meetings attended by more than 1,000 U.S. scientists and engineers. The goal of the town hall series was to examine scientific opportunities in the areas of artificial intelligence (AI), Big Data, and high-performance computing (HPC) in the next decade, and to capture the big ideas, grand challenges, and next steps to realizing these opportunities. In this report and in the Department of Energy (DOE) laboratory community, we use the term “AI for Science” to broadly represent the next generation of methods and scientific opportunities in computing, including the development and application of AI methods (e.g., machine learning, deep learning, statistical methods, data analytics, automated control, and related areas) to build models from data and to use these models alone or in conjunction with simulation and scalable computing to advance scientific research. The AI for Science town hall discussions focused on capturing the transformational uses of AI that employ HPC and/or data analysis, leveraging data sets from HPC simulations or instruments and user facilities, and addressing scientific challenges unique to DOE user facilities and the agency’s wide-ranging fundamental and applied science enterprise.
The See Change survey was designed to make z > 1 cosmological measurements by efficiently discovering high-redshift Type Ia supernovae (SNe Ia) and improving cluster mass measurements through weak lensing. This survey observed twelve galaxy clusters with the Hubble Space Telescope (HST) spanning the redshift range z = 1.13–1.75, discovering 57 likely transients and 27 likely SNe Ia at z ~ 0.8–2.3. As in similar previous surveys, this proved to be a highly efficient use of HST for supernova observations; the See Change survey additionally tested the feasibility of maintaining, or further increasing, the efficiency at yet higher redshifts, where we have less detailed information on the expected cluster masses and star formation rates. Furthermore, we find that the resulting number of SNe Ia per orbit is a factor of ~8 higher than for a field search, and 45% of our orbits contained an active SN Ia within 22 rest-frame days of peak, with one of the clusters by itself yielding 6 of the SNe Ia. We present the survey design, pipeline, and supernova discoveries. Novel features include fully blinded supernova searches, the first random forest candidate classifier for undersampled IR data (with a 50% detection threshold within 0.05 mag of human searchers), real-time forward-modeling photometry of candidates, and semi-automated photometric classifications and follow-up forecasts. We also describe the spectroscopic follow-up, instrumental in measuring host galaxy redshifts. The cosmology analysis of our sample will be presented in a companion paper.
Publicly accessible weather radar data have significant capabilities for meteorological measurements and predictions and, further, have the potential to measure nonmeteorological events that include smoke, ash, and debris plumes as well as explosions. The ability to identify and track nonmeteorological events can be of assistance in emergency response, hazard mitigation, and related activities in locations where radar coverage both exists and is recorded and accessible to the user. Here, in this study, events from multiple locations in the United States that are reported in news outlets are assessed using a manual inspection process of Level 2 weather radar data to identify anthropogenic and nonbiological returns. Explosive events are also identified, and a large high-altitude debris cloud from the intentional destruction of the SpaceX Starship is tracked across a wide area. Finally, future efforts using a machine learning model are discussed as a means of automating the process and potentially enabling near-real-time nonmeteorological event identification in the same areas where the data are accessible. Using weather radar data can be a valuable new tool for Department of Defense systems to aid in military awareness, and for interagency emergency response and forensic mission experts to consider national weather service data in their mission profiles. Radar data can be effective in detecting several common types of emergencies and inform and aid response personnel.
Integrated, distributed hydrologic models leverage advances in computational power and data accessibility to improve predictive understanding of the water cycle. While impressive advances in this area of environmental modeling have been accomplished, such models are still rarely used, partially because of difficulty integrating model and data. This research describes the release of Watershed Workflow version 1.2, a new library aiming to automate and enable complex workflows defining inputs to high resolution, integrated, distributed hydrologic models. Watershed Workflow provides tools enabling the discovery, acquisition, mapping, and coordination of watershed geometry, land cover, soil properties, and meteorological data. It enables the construction of unstructured meshes that incorporate this data, and provides tools for automating a “first” simulation on any watershed in the United States. We present the design of the workflow tool, and describe best practices for its usage, culminating in a final example from watershed specification to simulation at the Coweeta Hydrologic Laboratory.
A high-fidelity neutronics model of the Oak Ridge National Laboratory High Flux Isotope Reactor (HFIR) with the low-enriched uranium (LEU) high-density silicide dispersion Optimized fuel design was updated and analyzed to generate reactor physics-based metrics to support follow-on thermal hydraulic and transient analyses of this design. The Python HFIR Analysis and Measurement Engine (PHAME) was also updated to enhance the automation capabilities of the framework developed and maintained to perform these reactor physics modeling and simulation efforts. The automated framework significantly increases the efficiency and reproducibility to design and thoroughly analyzes HFIR LEU core designs, changes, and uncertainties. Reactor physics metrics evaluated include but are not limited to fuel depletion, cycle length, fission rate density distributions, axial power peaking factors, kinetics data, reactivity coefficients, control element worths, heat deposition rates, and decay heat. These neutronics results provide essential input to follow-on steady state thermal, thermal hydraulic and reactor transient analyses, which are subject of other reports. The Optimized design operates at 95 MW to maintain HFIR’s current highly enriched uranium core performance level at 85 MW.
Artificial intelligence is reshaping scientific exploration, but most methods automate procedural tasks without engaging in scientific reasoning, limiting autonomy in discovery. We demonstrate that hierarchical agentic large language model reasoning can efficiently drive simulation and scientific exploration. Across two chemical applications, CO adsorption on Cu surface transition metal adatoms and on M–N–C catalysts, reasoning-guided exploration reduces required atomistic simulations by up to 90% relative to heuristic or random selection. Comparisons across single-agent, multi-agent, and stochastic baselines show that hierarchical strategies yield more coherent and information-efficient search trajectories. Reasoning traces reveal chemically grounded decisions that cannot be explained by semantic bias or stochastic sampling. We realize these agentic reasoning strategies in Materials Agents for Simulation and Theory in Electronic-structure Reasoning (MASTER), a multimodal system that translates natural language into density functional theory workflows. Altogether, multi-agent collaboration accelerates heterogeneous catalyst discovery and marks a step toward more autonomous, reasoning-guided scientific exploration.
Advancements were made in automating microscopy-based material characterization, particularly in studying irradiation effects on additively manufactured (AM) materials using machine learning (ML) and computer vision (CV). These automation efforts address the challenges of analyzing complex microstructures, accelerating the detection of irradiation-induced defects. Two CV models were developed at Argonne National Laboratory (ANL) to enhance transmission electron microscopy (TEM) analysis of irradiated AM 316 stainless steel. The first model focused on the detection of irradiation-induced dislocation loops, which contribute to material hardening and embrittlement. These loops, categorized as faulted or perfect, were automatically detected and classified using a Mask R-CNN model trained on TEM images from both in-situ and ex-situ ion irradiation experiments. The model achieved high accuracy, with precision, recall, and F1 scores of 0.839, 0.734, and 0.776, respectively, demonstrating its effectiveness in analyzing dislocation loops in irradiated AM materials. The second CV model was developed to analyze the size and wall thickness of dislocation cells in laser powder bed fusion (LPBF) 316 stainless steel. Using a U-Net++ architecture with EfficientNet as the encoder, the model was trained on TEM images to segment and measure cell size and wall thickness.
Photosynthesis responds quickly to changes in light, increasing with incoming photosynthetic photon flux density (PPFD) until the leaves become light saturated. This instantaneous response to PPFD, which is widely studied and incorporated into models of photosynthesis, is overlaid on non-instantaneous photosynthetic changes resulting from the acclimation of plants to average PPFD over intermediate timescales of a week to months ($\overline{PPFD}$). Such photosynthetic light acclimation is not typically incorporated into models, due to the lack of observational constraints. In this work, we use eddy covariance observations from globally distributed and automated sensor networks, along with photosynthesis estimates from nine terrestrial biosphere models (TBMs), to quantify and assess photosynthetic acclimation to light in natural environments. We also use recent theoretical developments to incorporate light acclimation in a TBM. Our results show widespread light acclimation of ecosystem photosynthesis. On average, a 1 μmol m -2 s -1 increase in $\overline{PPFD}_{10}$ (ten-day average PPFD) leads to a 0.031 ± 0.013 μmol C m -2 s -1 increase in the maximum photosynthetic assimilation rate (A max ), with croplands having stronger acclimation rates than grasslands and forests. Our analysis shows that the TBMs examined either neglect or substantially underestimate light acclimation. By updating a TBM to include photosynthetic acclimation, successfully reproducing the $\overline{PPFD}_{10}$-A max relationship, we provide a robust method for the incorporation of photosynthetic light acclimation in future models.
The deployment of artificial intelligence systems in critical applications requires higher levels of assurance for safety, security, and interpretability. While neurosymbolic (NESY) approaches combining neural networks with symbolic reasoning offer potential advantages for assured AI, existing differentiable neurosymbolic frameworks face significant limitations including computational overhead and performance constraints. This report investigates the ISED (InferSampleEstimateDescend) framework as an alternative approach that enables neurosymbolic learning without requiring endtoend differentiability. We evaluate ISED’s utility for geointelligence applications by comparing neurosymbolic models against standard neural networks on aircraft classification tasks using the RarePlanes and MTARSI imagery datasets. Our results demonstrate that while ISEDbased models achieve slightly lower accuracy (89.7% vs 92.1% on RarePlanes; 91.1% vs 92.5% on MTARSI), they provide critical explainability capabilities that enable tracing incorrect predictions back to specific attribute misclassifications. We also present an automated pipeline that generates both attributeclass mappings and neurosymbolic model architectures from natural language descriptions, significantly reducing the manual effort required for NESY model deployment. These findings suggest that ISED offers a promising direction for developing assured AI systems where interpretability and reasoning transparency are prioritized alongside performance.
Simulating subsurface contaminant transport at the kilometer-scale often entails modeling reactive flow and transport within and through complex geologic structures. These structures are typically meshed by hand and as a result geologic structure is usually represented by one or a few deterministically generated geological models for uncertainty studies of flow and transport in the subsurface. Uncertainty in geologic structure can have a significant impact on contaminant transport. In this study, the impact of geologic structure on contaminant tracer transport in a shale formation is investigated for a simplified generic deep geologic repository for permanent disposal of spent nuclear fuel. An open-source modeling framework is used to perform a sensitivity analysis study on transport of two tracers from a generic spent nuclear fuel repository with uncertain location of the interfaces between the stratum of the geologic structure. The automated workflow uses sampled realizations of the geological structural model in addition to uncertain flow parameters in a nested sensitivity analysis. Concentration of the tracers at observation points within, in line with, and downstream of the repository are used as the quantities of interest for determining model sensitivity to input parameters and geological realization. Finally, the results of the study indicate that the location of strata interfaces in the geological structure has a first-order impact on tracer transport in the example shale formation, and that this impact may be greater than that of the uncertain flow parameters.
Machine learning (ML)-accelerated discovery requires large amounts of high-fidelity data to reveal predictive structure–property relationships. For many properties of interest in materials discovery, the challenging nature and high cost of data generation has resulted in a data landscape that is both scarcely populated and of dubious quality. Data-driven techniques starting to overcome these limitations include the use of consensus across functionals in density functional theory, the development of new functionals or accelerated electronic structure theories, and the detection of where computationally demanding methods are most necessary. When properties cannot be reliably simulated, large experimental data sets can be used to train ML models. In the absence of manual curation, increasingly sophisticated natural language processing and automated image analysis are making it possible to learn structure–property relationships from the literature. Finally, models trained on these data sets will improve as they incorporate community feedback.
Detecting cracks in cylinder heads traditionally relies on manual inspection, which is time-consuming and susceptible to human error. As an alternative, automated object detection utilizing computer vision and machine learning models has been explored. However, these methods often face challenges due to a lack of sufficiently annotated training data, limited image diversity, and the inherently small size of cracks. Addressing these constraints, this paper introduces a novel automated crack-detection method that enhances data availability through a synthetic data generation technique. Unlike general data augmentation practices, our method involves copying cracks from one location to another, guided by both random and informed engineering decisions about likely crack formations due to cyclic thermomechanical loads. The innovative aspect of our approach lies in the integration of domain-specific engineering knowledge into the synthetic generation process, which substantially improves detection accuracy. We evaluate our method’s effectiveness using two metrics: the F2 score, which emphasizes recall to prioritize detecting all potential cracks, and mean average precision (MAP), a standard measure in object detection. Experimental results demonstrate that, without engineering insights, our method increases the F2 score from 0.40 to 0.65, while maintaining a stable MAP. Incorporating detailed engineering knowledge further enhances the F2 score to 0.70 and improves MAP to 0.57, representing increases of 63% and 43%, respectively. These results confirm that our approach not only mitigates the limitations of traditional data augmentation but also significantly advances the reliability and precision of crack detection in industrial settings.
Bottom-up urban energy models are crucial for understanding current energy use patterns and informing design strategies. However, accurately characterizing these models to represent different communities remains a challenge due to the extensive data needed for simulating existing energy use behavior. This data includes information related to human activities and building characteristics, all of which correlate with socioeconomic factors. To overcome this challenge, we developed an automated framework that utilizes both top-down and bottom-up data, to predict unknown building and occupant characteristics that are needed for more accurate and equitable modeling and analytics. Our framework, integrated into the URBANopt district energy modeling platform, uses statistical data models from ResStock. URBANopt models co-located buildings and neighborhoods. At this scale there are data gaps in building characteristic data, such as materials, insulation, occupancy, income, and energy usage of the buildings. To address this data gap, we use ResStock data, representative at the census tract scale, and develop machine-learning and deeplearning techniques to disaggregate it to individual buildings. By mapping unique occupant, building and economic properties to URBANopt energy models, we gain detailed insights into the variability of building energy use across different neighborhoods. This insight helps deploy technologies for co-located buildings and supports targeted upgrades for communities with unique economic and demographic characteristics, ensuring energy equity. Accurate characterization of energy models allows us to develop equitable strategies tailored to diverse neighborhoods, whether underserved or affluent. Our automated framework streamlines energy modeling and provides a reliable tool for building energy characterization.
In this research, lean mixed-mode combustion is numerically investigated using computational fluid dynamics (CFD) in a spark-ignition engine. A new E30 fuel surrogate is developed using a neural network model with matched octane numbers. A skeletal mechanism is also developed by automated mechanism reduction and by incorporating a NO x submechanism. A hybrid approach that couples the G-equation model and the well-stirred reactor model is employed for turbulent combustion modeling. The developed CFD model is shown to well predict pressure and apparent heat release rate (AHRR) traces compared with experiment. Two types of combustion cycles (deflagration-only and mixed-mode cycles) are observed. The mixed-mode cycles feature early flame propagation and subsequent end-gas auto-ignition, leading to two distinctive AHRR peaks. The validated CFD model is then employed to investigate the effects of NO x chemistry. The NO x chemistry is found to promote auto-ignition through the residual gas, while the deflagration phase remains largely unaffected. Sensitivity analysis is finally performed to understand effects of fuel properties, including heat of vaporization (HoV) and laminar flame speed (S L ). An increased HoV tends to suppress auto-ignition through charge cooling, while the impact of HoV on flame propagation is insignificant. In contrast, an increased S L is found to significantly promote both flame propagation and end-gas auto-ignition. The promoting effect of S L on auto-ignition is not a direct chemical effect; it is rather caused by an advancement of the combustion phasing, which increases compression heating of the end-gas.
OpenStudio®-MCP is a Model Context Protocol (MCP) server that lets AI assistants perform building energy modeling through natural language. Rather than requiring users to learn the OpenStudio® SDK, EnergyPlus® scripting, or Ruby/Python automation, the server translates conversational requests into sequences of tool calls that create models, design HVAC systems, run simulations, and extract results — all within a single chat session. The server's 124 tools are organized into a skills architecture where each skill encapsulates a domain of building energy modeling (envelope, HVAC, loads, weather, simulation, results) behind typed, LLM-friendly interfaces. High-leverage operations like applying ASHRAE 90.1 baseline systems or generating standards-compliant typical buildings are exposed as single tool calls that internally wire dozens of OpenStudio® objects. Bundled measures from ComStock™ and Openstudio® -common-measures-gem are wrapped with dedicated tools and typed arguments rather than exposed through a generic measure interface, so AI models get consistent, error-resistant recipes without needing to discover measure arguments at runtime. A key design decision is structured results extraction: six SQL-based tools return surgical ~300–1,000 token responses (end-use breakdowns, envelope summaries, HVAC sizing, timeseries data) instead of requiring the AI to parse ~100K-token raw HTML reports, making iterative design exploration practical within context window limits. The codebase is designed as a reference implementation — explicit, well-commented, and modular — so that other simulation engines (EnergyPlus® standalone, TRNSYS, DOE-2) can use it as a template for building their own MCP servers.
An automated platform has been developed to assist researchers in the rapid development of optical spectroscopy sensors to quantify species from spectral data. This platform performs calibration and validation measurements simultaneously. Real-time, in situ monitoring of complex systems through optical spectroscopy has been shown to be a useful tool; however, building calibration models requires development time, which can be a limiting factor in the case of radiological or otherwise hazardous systems. While calibration time can be reduced through optimized design of experiments, this study approached the challenge differently through automation. The ATLAS (Automated Transient Learning for Applied Sensors) platform used pneumatic control of stock solutions to cycle flow profiles through desired calibration concentrations for multivariate model construction. Additionally, the transients between desired concentrations based on flow calculations were used as validation measurements to understand model predictive capabilities. This automated approach yielded an incredible 76% reduction in model development time and a 60% reduction in sample volume versus estimated manual sample preparation and static measurements. The ATLAS system was demonstrated on two systems: a three-lanthanide system with Pr/Nd/Ho representing a use case with significant overlap or interference between analyte signatures and an alternate system containing Pr/Nd/Ni to demonstrate a use case in which broad-band corrosion species signatures interfered with more distinct lanthanide absorbance profiles. Both systems resulted in strong model prediction performance (RMSEP < 9%). Lastly, ATLAS was demonstrated as a tool to simulate process monitoring scenarios (e.g., column separation) in which models can be further optimized to account for day-to-day changes as necessary (e.g., baseline correction). Ultimately, ATLAS offers a vital tool to rapidly screen monitoring methods, investigate sensor fusion, and explore more complex systems (i.e., larger numbers of species).
History matching aims to find a numerical reservoir model that can be used to predict the reservoir performance. An engineer and model calibration (data inversion) method are required to adjust various parameters/properties of the numerical model in order to match the reservoir production history. In this study, we develop deep neural networks within the reinforcement learning framework to achieve automated history matching that will reduce engineers’ efforts, human bias, automatically and intelligently explore the parameter space, and remove the need of large set of labeled training data. To that end, a fast-marching-based reservoir simulator is encapsulated as an environment for the proposed reinforcement learning. The deep neural-network-based learning agent interacts with the reservoir simulator within reinforcement learning framework to achieve the automated history matching. Reinforcement learning techniques, such as discrete Deep Q Network and continuous Deep Deterministic Policy Gradients, are used toth, used to train the learning agents. The continuous actions enable the Deep Deterministic Policy Gradients to explore more states at each iteration in a a learning episode; consequently, a better history matching is achieved using this algorithm as compared to Deep Q Network. For simplified dual-target composite reservoir models, the best history-matching performances of the discrete and continuous learning methods in terms of normalized root mean square errors are 0.0447 and 0.0038, respectively. Furthermore, our study shows that continuous action space achieved by the deep deterministic policy gradient drastically outperforms deep Q network.