Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “performance prediction”

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 217 records · Page 12
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Multifidelity Active Learning for Failure Estimation of TRISO Nuclear Fuel

The Tristructural isotropic (TRISO)-coated particle fuel is a robust nuclear fuel proposed to be used for multiple modern nuclear technologies. Therefore, characterizing its safety is vital for the reliable operation of nuclear technologies. However, the TRISO fuel failure probabilities are small and the computational model is time consuming to evaluate them using traditional Monte Carlo-type approaches. In the paper, we present a multifidelity active learning approach to efficiently estimate small failure probabilities given an expensive computational model. Active learning suggests the next best training set for optimal subsequent predictive performance and multifidelity modeling uses cheaper low-fidelity models to approximate the high-fidelity model output. After presenting the multifidelity active learning approach, we apply it to efficiently predict TRISO failure probability and make comparisons to the reference results.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2023

Discovery of multi-functional polyimides through high-throughput screening using explainable machine learning

Polyimides have been widely used in modern industries because of their excellent mechanical and thermal properties, e.g., high-temperature fuel cells, displays, and aerospace composites. However, it usually takes decades of experimental efforts to develop a successful product. Aiming to expedite the discovery of high-performance polyimides, we utilize computational methods of machine learning (ML) and molecular dynamics (MD) simulations. Our study provides compelling evidence for the effectiveness of a data-driven approach in discovering novel polyimides. We first build a comprehensive library of more than 8 million hypothetical polyimides based on the polycondensation of existing dianhydride and diamine/diisocyanate molecules. Then we establish multiple ML models for the thermal and mechanical properties of polyimides based on their experimentally reported values, including glass transition temperature, Young’s modulus, and tensile yield strength. The obtained ML models demonstrate excellent predictive performance in identifying the key chemical substructures influencing the thermal and mechanical properties of polyimides. The use of explainable machine learning describes the effect of chemical substructures on individual properties, from which human experts can understand the cause of the ML model decision. Applying the well-trained ML models, we obtain property predictions of the 8 million hypothetical polyimides. Then, we screen the whole hypothetical dataset and identify three (3) best-performing novel polyimides that have better-combined properties than existing ones through Pareto frontier analysis. For an easy query of the discovered high-performing polyimides, we also create an online platform https://polyimide-explorer.herokuapp.com/ that embeds the developed ML model with interactive visualization. Furthermore, we validate the ML predictions through all-atom MD simulations and examine their synthesizability. The MD simulations are in good agreement with the ML predictions and the three novel polyimides are predicted to be easy to synthesize via Schuffenhauer’s synthetic accessibility score. Following the proposed ML guidance, we successfully synthesized a novel polyimide and the experimentally obtained high glass transition/thermal decomposition temperature demonstrated its excellent thermal stability. Here our study demonstrates an efficient way to expedite the discovery of novel polymers using ML prediction and MD validation. The high-throughput screening of a large computational dataset can serve as a general approach for new material discovery in other polymeric material exploration problems, such as organic photovoltaics, polymer membranes, and dielectrics.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Spectral Effects in Albedo and Rearside Irradiance Measurement for Bifacial Performance Estimation

We investigate the impact of spectral dependence of ground surface reflectivity on albedo and rearside irradiance measurements necessary for bifacial photovoltaic (PV) module performance estimation and monitoring. Because PV modules are spectrally selective, albedo and irradiance measurements performed with common irradiance sensors may require spectral mismatch corrections when used for performance prediction. We investigate via simulation the differences in spectrally responsive albedo measured with thermopile pyranometers and crystalline silicon PV reference cells in comparison to a typical crystalline-silicon bifacial PV module. Simulations are performed for nine different representative ground surface materials using simulated solar spectra together with spectral reflectivity data distributed with the SMARTS simulation software. For the materials considered, the results show that albedo spectral mismatch relative to the bifacial module is distributed over a range of ±9.2% for thermopile pyranometers versus only ±3.7% for a typical PV reference cell. We consider the impact of this spectrally-responsive albedo mismatch on bifacial PV module rearside irradiance measurements. Using synthesized rearside spectral irradiance distributions, we find that for the nine different ground surface materials the predicted rearside irradiance measurement deviates from the effective irradiance observed by the PV module by on the order of 16.5 W/m 2 for the pyranometer and 3.6 W/m 2 for the PV reference cell. We discuss the implications for bifacial albedo and irradiance measurement.

41 EE - Solar Energy Technologies Office (EE-4S)↗
DOE OSTI2025

Pavement condition and climatic data in southeast Texas: A dataset for evaluating flood impacts on pavement performance

Effective pavement maintenance is essential for economic stability, optimal network performance, and roadway safety. Achieving this requires thorough evaluation of pavement conditions, including structural integrity, surface roughness, and distress characteristics. Pavement performance indicators play a critical role in influencing vehicle safety and ride quality. Recent advances have emphasized the use of data-driven modeling to anticipate pavement behavior, with the goal of optimizing resource allocation and refining Maintenance and Rehabilitation (M&R) strategies through accurate condition assessment. A foundational requirement for these modeling efforts is the availability of standardized, high-quality datasets that can support robust and reproducible infrastructure analysis. This data article presents a comprehensive dataset assembled to facilitate pavement performance prediction, with a geographic focus on Southeast Texas, particularly the flood-vulnerable area of Beaumont. The dataset encompasses pavement and traffic attributes, meteorological records, flood simulation outputs, ground deformation measurements, and topographic indices, enabling detailed examination of both load-associated and non-load-associated degradation mechanisms. Data preprocessing was performed using ArcGIS Pro, Microsoft Excel, and Python to ensure consistency and usability in data-driven modeling applications, including machine learning workflows. Key contributions of this dataset include its utility in analyzing the climatic and environmental factors affecting pavement conditions, identifying critical predictive features, and enabling in-depth correlation analysis across diverse variables. By filling existing gaps in input variable selection resources, this dataset supports the development of predictive tools for estimating future maintenance demand and enhancing the resilience of pavement networks in flood-impacted areas. The resource highlights the importance of standardized datasets for advancing pavement management practices and provides a robust foundation for ongoing infrastructure performance modeling.

42 ENGINEERING↗
DOE OSTI2021

Evaluating causal‐based feature selection for fuel property prediction models

Abstract In‐silico screening of novel biofuel molecules based on chemical and fuel properties is a critical first step in the biofuel evaluation process due to the significant volumes of samples required for experimental testing, the destructive nature of engine tests, and the costs associated with bench‐scale synthesis of novel fuels. Predictive models are limited by training sets of few existing measurements, often containing similar classes of molecules that represent just a subset of the potential molecular fuel space. Software tools can be used to generate every possible molecular descriptor for use as input features, but most of these features are largely irrelevant and training models on datasets with higher dimensionality than size tends to yield poor predictive performance. Feature selection has been shown to improve machine learning models, but correlation‐based feature selection fails to provide scientific insight into the underlying mechanisms that determine structure–property relationships. The implementation of causal discovery in feature selection could potentially inform the biofuel design process while also improving model prediction accuracy and robustness to new data. In this study, we investigate the benefits causal‐based feature selection might have on both model performance and identification of key molecular substructures. We found that causal‐based feature selection performed on par with alternative filtration methods, and that a structural causal model provides valuable scientific insights into the relationships between molecular substructures and fuel properties.

Nguyen, Bernard↗
DOE OSTI2020

Transition Core Planning and Safety Analyses in Support of LEU Fuel Conversion of the University of Missouri Research Reactor (MURR)

The University of Missouri Research Reactor (MURR®) is one of six U.S. High Performance Research Reactors (USHPRR), including one critical facility, that is working with the National Nuclear Security Administration (NNSA) Office of Material Management and Minimization (M3) Reactor Conversion Program to convert from highly enriched uranium (HEU) to low-enriched uranium (LEU) fuel. The M3 Reactor Conversion USHPRR Project objectives include the development of LEU fuel element designs that will ensure safe reactor operations and to maintain the existing experimental performance of each facility. The work is being conducted through many inter-related activities being completed by four Project Pillars: Fuel Qualification (FQ), Fuel Fabrication (FF), Reactor Conversion (RC), and Cross Cutting (CC). A new type of LEU fuel based on an alloy of uranium-10 wt% molybdenum (U-10Mo) is expected to allow the conversion of those USHPRR, like MURR, requiring higher density fuels. The very-high-density LEU U-10Mo monolithic fuel is currently undergoing irradiation testing and post-irradiation examination under a planned and documented fuel qualification effort. The FQ Pillar will document fuel property and fuel performance data and qualify the fuel for use in these reactors. The FF Pillar is fabricating fuel for ongoing and future irradiation tests, as well as conducting fabrication demonstrations to validate or update preliminary fabrication assumptions. The FF Pillar is also working to develop and install commercial manufacturing capacity with the U-10Mo monolithic fuel to produce prototypic fuel. Working with the RC Pillar at Argonne, MURR has progressed through a preliminary fuel element design using preliminary data for the proposed monolithic alloy of U-10Mo. Analyses were completed in previous work that found for typical equilibrium operations with the preliminary LEU fuel element design, in conjunction with a power uprate to 12 MW and appropriate changes to the MURR Limiting safety system settings (LSSS), MURR will have adequate margins to safety for steady-state operations and postulated transient accidents and will have experimental performance in key locations that meets or exceeds current operations with HEU fuel. The purpose of this work is to develop a sequence of transition cycles that will enable MURR to transition from operation with the reactor core loaded with fresh LEU fuel elements only to typical equilibrium operations with mixed-burnup cores following conversion while meeting operational requirements on safety and experimental performance. It is expected that the use of fresh LEU fuel at conversion and subsequent low burnup of the LEU fuel elements that will initially be available for use following conversion will result in critical control blade positions that will substantially change the axial power distribution in the core and the neutron flux available in key experimental locations relative to equilibrium LEU operations. Given the constraints of MURR safety margins, operational practices, and production and research, a novel method has been developed to identify a transition sequence that minimizes the time MURR operates atypically compared to the current prototypic cycles using HEU fuel. The proposed transition sequence moves quickly to the same sort of equilibrium cycles for the LEU fuel that have already been evaluated in documented preliminary safety analyses. Although shifting the neutron flux peak to the lower half of the core during initial cycles with LEU at 12 MW reduces the experiment performance in some key locations relative to current HEU operations at 10 MW, all LEU cores provide an average performance that meets or exceeds that of HEU. An LEU cycle is reached that meets or exceeds the level of experimental performance predicted for current HEU and equilibrium LEU operations in more than 450 key locations identified by a reactor specialist at MURR by the 23rd cycle following conversion and that afterwards will enable MURR to consistently meet its experimental performance requirements. The proposed transition sequence only requires the fabrication of 34 fresh LEU elements in the first year of operation and does not exceed the anticipated availability of fresh elements that can be produced by the fuel fabricator. By the third year after conversion, 22 fresh LEU elements will be required each year, which is the same as expected for equilibrium LEU operations and the same as current operations with HEU fuel. The proposed transition sequence thus combines a relatively short time period before equilibrium burnup is achieved, a temporary increase of fuel elements needed annually relative to typical operations that are within the production capabilities of the fuel fabricator, and demonstrates comparable experimental performance of the LEU cores relative to current HEU operations. Further measures may be taken to reduce any initial experimental performance penalty even further, where possible, by repositioning certain experiments to leverage the increased performance in the lower axial experimental positions in the initial cycles following conversion or leaving the experiments in the irradiation facilities longer in order to achieve the required neutron fluence. This analysis may require refinement depending on the experimental facilities in use at the time of conversion. Nonetheless, the results presented here, including the experimental performance, core burnup, and critical control blade positions throughout the transition cycles, show that the proposed transition cycle fuel management patterns are consistent with what is expected and desired for MURR operation with LEU U-10Mo fuel. Detailed core power distributions from the neutronics models were also used to evaluate safety margins during steady-state operations for the selected transition cycles and the equilibrium LEU core. It is shown that there are adequate safety margins for both steady-state operations and postulated accident scenarios. For the steady-state operations with the preliminary LEU fuel element design the analysis predicts at least 2.49 MW margin to the onset of flow instability at the LSSS power of 15 MW. Considering the LSSS power is 125% of full license power, the margin to OFI is sufficient. In addition, the critical heat flux ratio at LSSS power is well above the requirement of CHFR > 2.0 from NUREG-1537 for all considered cases. For postulated transient accidents, the minimum margin to the fuel temperature safety limit is at least 109 °C. In summary, the proposed sequence of core loadings for MURR operations following conversion to LEU fuel and a power uprate to 12 MW provides sufficient safety margins for both steady-state operations and postulated transient accidents during a proposed sequence of transition cycles to equilibrium operations. Analysis has shown that there are some local experimental performance penalties during the initial cycles. Although there are local shifts in the experimental performance, on average all LEU cores at 12 MW have equal or higher performance than HEU at 10 MW. Temporary adjustments are being planned that will produce suitable experimental performance during these cycles. The results indicate that for the equilibrium LEU core the experimental performance exceeds that of current HEU operations in all key locations while also demonstrating sufficient safety margins.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2020

Hybrid deep learning architecture for general disruption prediction across tokamaks

In this paper, we present a new deep learning disruption prediction algorithm based on important findings from explorative data analysis which effectively allows knowledge transfer from existing devices to new ones, thereby predicting disruptions using very limited disruptive data from the new devices. Here, the explorative data analysis conducted via unsupervised clustering techniques confirms that time-sequence data are much better separators of disruptive and non-disruptive behavior than the instantaneous plasma state data with further advantageous implications for a sequence-based predictor. Based on such important findings, we have designed a new algorithm for multi-machine disruption prediction that achieves high predictive accuracy on the C-Mod (AUC=0.801), DIII-D (AUC=0.947) and EAST (AUC=0.973) tokamaks with limited hyperparameter tuning. Through numerical experiments, we show that boosted accuracy (AUC=0.959) is achieved on EAST predictions by including in the training only 20 disruptive discharges, thousands of non-disruptive discharges from EAST, and combining this with more than a thousand discharges from DIII-D and C-Mod. The improvement of predictive ability obtained by combining disruptive data from other devices is found to be true for all permutations of the three devices. Furthermore, by comparing the predictive performance of each individual numerical experiment, we find that non-disruptive data are machine-specific while disruptive data from multiple devices contain device-independent knowledge that can be used to inform predictions for disruptions occurring on a new device.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2021

Hybrid deep learning architecture for general disruption prediction across tokamaks

In this paper, we present a new deep learning disruption prediction algorithm based on important findings from explorative data analysis which effectively allows knowledge transfer from existing devices to new ones, thereby predicting disruptions using very limited disruptive data from the new devices. The explorative data analysis conducted via unsupervised clustering techniques confirms that time-sequence data are much better separators of disruptive and non-disruptive behavior than the instantaneous plasma state data with further advantageous implications for a sequence-based predictor. Based on such important findings, we have designed a new algorithm for multi-machine disruption prediction that achieves high predictive accuracy on the C-Mod (AUC=0.801), DIII-D (AUC=0.947) and EAST (AUC=0.973). tokamaks with limited hyperparameter tuning. Through numerical experiments, we show that boosted accuracy (AUC=0.959) is achieved on EAST predictions by including in the training only 20 disruptive discharges, thousands of non-disruptive discharges from EAST, and combining this with more than a thousand discharges from DIII-D and C-Mod. The improvement of predictive ability obtained by combining disruptive data from other devices is found to be true for all permutations of the three devices. Furthermore, by comparing the predictive performance of each individual numerical experiment, we find that non-disruptive data are machine-specific while disruptive data from multiple devices contain device-independent knowledge that can be used to inform predictions for disruptions occurring on a new device.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2022

Longitudinal Trajectories of Memory Performance in Patients with Early-Stage Breast Cancer

Background. While breast cancer and its treatments may affect cognition, the longitudinal trajectories of cognition among those receiving differing cancer treatment types remain poorly understood. Prior research suggests hippocampal-prefrontal cortex network integrity may influence cognition, although how this network predicts performance over time remains unclear. Methods. We conducted a prospective trial including 69 patients with early-stage breast cancer receiving adjuvant therapy and 12 controls. Longitudinal cognitive testing was conducted at four visits: pretreatment-baseline, 6-7 months, 14-15 months, and 23-24 months. Cognitive composite scores of episodic memory, executive functioning, and processing speed were assessed at each timepoint. Baseline structural MRI was obtained in a subset of these participants, and hippocampal and prefrontal cortex regional volumes were extracted. Results. Longitudinal linear mixed modeling revealed significant group by time interactions on memory performance, controlling for age and education. Post hoc analyses revealed this effect was driven by patients treated with chemotherapy or chemotherapy plus hormone therapy, who demonstrated the least improvement in memory scores over time. Treatment group did not significantly influence the relationship between time and processing speed or executive functioning. Neither pretreatment hippocampal nor prefrontal volume differed between groups, and there were no significant group by time by baseline regional volume effects on cognition. Conclusion. Patients with early-stage breast cancer treated with chemotherapy or chemotherapy plus hormone therapy benefit less from practice effects seen in healthy controls on memory tests. Loss of longitudinal practice effect may be a new and clinically relevant measure for capturing patients’ experience of cognitive difficulties after treatment.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Experimental validation of an organic rankine-vapor compression cooling cycle using low GWP refrigerant R1234ze(E)

There is a significant global opportunity to capture and utilize low grade waste heat to reduce fossil fuel consumption, greenhouse gas emissions, and improve energy efficiency across a wide range of industries. In this work, an advanced type of thermally activated cooling system, an organic Rankine-vapor compression cycle (ORVC) with novel heat integration strategies, was designed and tested at a relevant scale for industrial waste heat recovery (300 kW th cooling capacity). The ORVC linked an organic Rankine power cycle and a vapor compression cooling cycle using a turbine and compressor that shared a single shaft. The ORVC test facility absorbed waste heat from a liquid stream at 91 °C to simulate engine coolant in diesel generator sets, rejected heat to a glycol stream at 30 °C, and generated chilled water at 7 °C. A cooling capacity of 264 kW ± 3.5 kW was experimentally validated with a COP of 0.56 ± 0.01 during steady-state operation at the design temperatures. The thermal efficiency, accounting for pump work, of the Rankine cycle was 7.7% ± 0.22% and the COP of the vapor compression cycle was 5.25 ± 0.09. The centrifugal turbo-compressor operated at 31.5 kRPM ± 0.3 kRPM, with a turbine and compressor isentropic efficiencies of 76.7% ± 0.90% and 84.8% ± 0.54%, respectively, with near-perfect power transmission between these components. The pressure drop in the piping and heat exchangers were significantly larger than expected which had a detrimental impact on the performance of the ORVC. In addition, the condenser on the cooling cycle could not deliver the subcooling as specified from the design point modeling. Furthermore, the results from the sensitivity analysis showed that the higher condenser glycol outlet temperature had the largest impact on performance, which is consistent with other analytical models in the literature. When the ORVC simulations were updated with experimental values for isentropic efficiencies of the turbomachinery, the thermal COP was 0.66 which represents an estimate of the predicted performance if test facility limitations are overcome.

30 DIRECT ENERGY CONVERSION↗
DOE OSTI2021

Desalination metamodels and a framework for cross-comparative performance simulations

There is an opportunity to save energy and reduce operational expenses when choosing a suitable desalination method aided by computational modeling. Existing models are not conducive to generalized comparisons between different desalination methods. Therefore, the work in this study developed metamodels for six desalination methods, grouped them into thermal and molecular transport families, and validated their predictive performance within 9% difference from published data. This validated framework allowed comparisons of desalination methods at their prescribed ranges of operational conditions that they were designed for. These conditions specify feed salinity ranges of 1.6 to 2.4 g/kg for Capacitive Deionization and Reverse Osmosis (RO), 2.8 to 4.2 g/kg for Electrodialysis, 28 to 42 g/kg for Thermovapor Compression and Humidification-Dehumidification, and 37 to 55 g/kg for Multi-Effect Distillation (MED). Despite different operational conditions, all models exhibit non-linear, positive correlation between energy consumption and system size in response to feed salinity and production rate. The framework is also employed in a cross-comparative analysis between MED and RO whose results suggest that energy intensity for MED is an order of magnitude greater than RO for the same operational conditions, but actual operational costs are comparable. Overall, the framework is ready for deployment in case studies of actual desalination plants.

42 ENGINEERING↗
DOE OSTI2024

Experimental and statistical study on the effect of process parameters on the quality of continuous fiber composites made via additive manufacturing

Ongoing research in additive manufacturing towards structural and industrial application has led to the use of commingled roving as a manufacturing feedstock for printing high fiber volume fraction composites. The prospects of using this technology for high performance applications necessitates the need for a comprehensive experimental investigation into the effects of processing parameters on the quality of an additively manufactured composite printed from commingled roving feedstock. Here, in this work, transverse flexure and void fraction matrix pyrolysis testing are both performed to evaluate composite quality. The transverse flexure test is a testing approach that evaluates the quality of the interfacial fiber-matrix bond while the void fraction test estimates the void content in the printed composite. A full observational study consisting of 27 different test combinations is done to investigate the effects of three different process parameters namely, temperature, pressure, and print speed across three different levels. Composite samples were made from commingled roving of E-glass and amorphous PET using an in-house built continuous fiber composite digital manufacturing system. Least squares regression analysis is performed to study the main, interaction and quadratic effects of process parameters. A statistical regression model having an R2 adjusted value of 80.1% is generated from the transverse flexure study, which is used to explain main and interaction effects and also predict performance. Response surface plots are also generated and are used to optimize process parameters which can subsequently be of help in scaling up composite manufacturing. Results show that all three process parameters are highly statistically significant at the 0.01 level of significance. Pressure * Temperature and Pressure * Printspeed are significant interaction terms. Pressure plays a weightier role when print speed is increased or temperature is decreased as it closes more voids that would ordinarily have been introduced because of drop in polymer melt viscosity. Micrographic analysis is also performed.

36 MATERIALS SCIENCE↗
DOE OSTI2021

MADE3D: Enabling the next generation of high-torque density wind generators by additive design and 3D printing

Direct-drive wind turbine generators are increasing in popularity, thanks to recent project developments—especially offshore, where reliability and efficiency are major cost drivers. Yet, high capital costs are forcing many original equipment manufacturers to consider lightweight, high-torque density generators for next-generation multi-megawatt turbines that may be difficult to realize by traditional design or manufacturing methods. In this study, we present a new design framework enabled by advanced machine learning and multimaterial additive manufacturing to perform a magnetic topology optimization that maximizes the torque per rotor active mass for a 15-megawatt direct-drive permanent magnet wind generator. A comparison of the proposed approach against conventional topology optimization demonstrated a significant increase in computational efficiency and accuracy in performance predictions. Results using single and multimaterial compositions for rotor core and magnets identify a wider choice of 3D printable designs for a given specification. A hybrid combination of sintered and dysprosium-free polymer-bonded magnets shows good potential for torque performance by saving material costs up to 8.75%. More than 30% improvement in rotor torque densities is identified which can marginally improve the overall generator torque density. With the rapid evolution of multipowder deposition technolgies, this study can greatly inspire a new paradigm for design-driven manufacturing with novel material compositions and lightweight, low-cost, high-strength multimaterial geometries that were previously unexplored for direct-drive generators.

17 WIND ENERGY↗
DOE OSTI2022

Energetic Materials

Energetic materials comprise explosives, pyrotechnics, and propellants. The science of energetic materials is dedicated to developing a means to predict performance and safety characteristics with high fidelity. This is a particular challenge and is predicated on materials science and engineering, physics, chemistry, and dynamic response in extreme conditions. Fundamental elements of these complicated composite materials remain grand challenges—from the design of high-energy metastable molecules, to the engineering of composite formulations, to the processing parameters that link to safety and performance characteristics in as-yet undetermined ways. Key elements include crystalline mechanics, grain dynamics, multiphase interfaces, thermal and mechanical damage, and failure—all linked to multistep and high-rate chemistry and shock physics. A future revolution in our understanding and predictive capability for energetic materials behavior and responses is dependent upon sustained focus and advances in materials research and development.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Protein-ligand binding affinity prediction using multi-instance learning with docking structures

Recent advances in 3D structure-based deep learning approaches demonstrate improved accuracy in predicting protein-ligand binding affinity in drug discovery. These methods complement physics-based computational modeling such as molecular docking for virtual high-throughput screening. Despite recent advances and improved predictive performance, most methods in this category primarily rely on utilizing co-crystal complex structures and experimentally measured binding affinities as both input and output data for model training. Nevertheless, co-crystal complex structures are not readily available and the inaccurate predicted structures from molecular docking can degrade the accuracy of the machine learning methods. We introduce a novel structure-based inference method utilizing multiple molecular docking poses for each complex entity. Our proposed method employs multi-instance learning with an attention network to predict binding affinity from a collection of docking poses. We validate our method using multiple datasets, including PDBbind and compounds targeting the main protease of SARS-CoV-2. The results demonstrate that our method leveraging docking poses is competitive with other state-of-the-art inference models that depend on co-crystal structures. This method offers binding affinity prediction without requiring co-crystal structures, thereby increasing its applicability to protein targets lacking such data.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

MADE3D: Enabling the Next-Generation High-Torque- Density Wind Generators by Additive Design and 3D Printing

Direct-drive wind turbine generators are increasing in popularity, thanks to recent project developments - especially offshore, where reliability and efficiency are major cost drivers. Yet, high capital costs are forcing many original equipment manufacturers to consider lightweight, high-torque density generators for next-generation multi-megawatt turbines that may be difficult to realize by traditional design or manufacturing methods. In this study, we present a new design framework enabled by advanced machine learning and multimaterial additive manufacturing to perform a magnetic topology optimization that maximizes the torque per rotor active mass for a 15-megawatt direct-drive permanent magnet wind generator. A comparison of the proposed approach against conventional topology optimization demonstrated a significant increase in computational efficiency and accuracy in performance predictions. Results using single and multimaterial compositions for rotor core and magnets identify a wider choice of 3D printable designs for a given specification. A hybrid combination of sintered and dysprosium-free polymer-bonded magnets shows good potential for torque performance by saving material costs up to 8.75%. More than 30% improvement in rotor torque densities is identified which can marginally improve the overall generator torque density. With the rapid evolution of multipowder deposition technologies, this study can greatly inspire a new paradigm for design-driven manufacturing with novel material compositions and lightweight, low-cost, high-strength multimaterial geometries that were previously unexplored for direct-drive generators.

3D printing↗
DOE OSTI2025

An Initial Microstructurally Informed Model of High Burnup Structure Formation in UO 2 Fuel

The microstructure of a UO 2 fuel pellet changes as burnup increases, impacting fuel performance. Predicting and characterizing high burnup structure (HBS) and dark zone formation is a key part of supporting burnup limit extensions for light water reactors. This paper describes a model developed through fitting radially resolved pellet data obtained from recently published microstructural characterization data. The model predicts grain size and grain character, in addition to pore density and size, with fitting dependencies on power history variables. Separately fitting power history variables to microstructural parameters allows for insight into the underlying physical phenomena for future model development. Additionally, experimental data have been correlated to an HBS fraction to facilitate the development of a model capable of predicting a total fuel restructured fraction at the engineering scale. In conclusion, this two-step approach provides a coupling from reactor power history to microstructural data to fractional HBS and creates a basis to model HBS-dependent parameters in a fuel performance code.

High burnup structure↗
DOE OSTI2021

A Physics-Constrained Deep Learning Model for Simulating Multiphase Flow in 3D Heterogeneous Porous Media

Physics-based simulators for multiphase flow in porous media emulate nonlinear processes with coupled physics, and usually require extensive computational resources for software development, maintenance and simulation execution. As a result, a huge demand exists for fast modeling of coupled processes in a wide range of subsurface applications including geological sequestration, hydrocarbon recovery and geothermal energy extraction. In this work, an efficient physics-constrained deep learning model is developed for solving multiphase flow in 3-Dimensional (3D) heterogeneous porous media. The model fully leverages the spatial topology predictive capability of convolutional neural networks, specifically U-Net with successive contracting and expansive steps, and is coupled with an efficient continuity-based smoother to predict flow responses that need spatial continuity. Furthermore, the transient regions are penalized to steer the training process such that the model can accurately capture flow in these regions. The model takes inputs including properties of porous media, fluid properties and well controls, and predicts the temporal-spatial evolution of the state variables (pressure and saturation). While maintaining the continuity of fluid flow, the 3D spatial domain is decomposed into 2D images for reducing training cost, and the decomposition results in an increased number of training data samples and better training efficiency. Additionally, a surrogate model is separately constructed as a postprocessor to calculate well flow rate based on the predictions of state variables from the deep learning model. We use the example of CO 2 injection into saline aquifers, and apply the physics-constrained deep learning model that is trained from physics-based simulation data and emulates the physics process. The model performs prediction with a speedup of ~ 1400 times compared to physics-based simulations, and the average temporal errors of predicted pressure and saturation plumes are 0.27% and 0.099% respectively. Furthermore, water production rate is efficiently predicted by a surrogate model for well flow rate, with a mean error less than 5%. Therefore, with its unique scheme to cope with the fidelity in fluid flow in porous media, the physics-constrained deep learning model can become an efficient predictive model for computationally demanding inverse problems or other coupled processes.

58 GEOSCIENCES↗