Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Model selection”

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 307 records · Page 17
All sourcesNASA NTRSDOE OSTI
NASA NTRS2019

Shockwave/Boundary-Layer Interaction Studies Performed in the NASA Langley 20-Inch Mach 6 Air Tunnel

This paper highlights results from a collaborative study performed by The University of Tennessee Space Institute (UTSI) and NASA Langley Research Center on the Shockwave/Boundary-Layer Interaction (SWBLI) generated by a cylindrical protuberance on a flat plate in a Mach 6 flow. The study was performed in the 20-Inch Mach 6 Air Tunnel at NASA Langley Research Center and consisted of two separate entries. In the first entry, simultaneous high-speed schlieren and high-speed pressure-sensitive paint (PSP) imaging – which was performed for the first time in the 20-Inch Mach 6 facility at NASA Langley – were performed as well as simultaneous high-speed schlieren and oil-flow imaging. In the second entry, the model configuration was modified to increase the size of the interaction region. High-speed schlieren and infrared thermography (IR) surface imaging were performed in this second entry. The goal of these tests was to characterize the SBLI in the presence of a laminar, transitional, and turbulent boundary layer using high-speed optical imaging techniques. AoA = sting angle-of-attack (°) dcylinder = cylinder diameter (mm) dtrip = cylindrical tripping element diameter (mm) Δshock = shock stand-off distance (mm) hcylinder = cylinder height (mm) htrip = cylindrical tripping element height (mm) HSS = high-speed schlieren M∞ = freestream Mach number PSP = pressure-sensitive paint Re∞ = freestream unit Reynolds number (m-1) SWBLI = shockwave/boundary-layer interaction θplate = model plate angle (°) Introduction his paper highlights two experimental entries performed in the 20-Inch Mach 6 Air Blowdown Tunnel at NASA Langley Research Center in collaboration with The University of Tennessee Space Institute (UTSI). The purpose of these entries was to characterize the dynamic shockwave/boundary-layer interaction (SWBLI) between a vertical cylinder on a flat plate and laminar, transitional (XSWBLI), and turbulent (SWTBLI) boundary layers with a freestream Mach number of 6 using non-intrusive optical diagnostics. Experiments performed by Murphree et al.1,2 were among the first to specifically characterize XSWBLI induced by a vertical cylinder on a flat plate geometry using several optical measurement techniques. Recent optical studies of XSWBLI phenomenon have been performed by UTSI at Mach 2 in their low-enthalpy blow wind tunnel3-8 and by Texas A&M University and UTSI at Mach numbers of 6 and 7 in their Adjustable Contour Expansion wind tunnel.9 The experiments described in this paper were intended to complement previous studies by expanding the freestream unit Reynolds number range, Re∞, over which the XSWBLI phenomena has been observed. Additionally these experiments, made possible under NASA’s new facility funding model under the Aeronautics Evaluation and Test Capabilities (AETC) project, promoted collaboration between university and NASA researchers. The initial entry in the 20-Inch Mach 6 Air Tunnel at NASA Langley occurred in December of 2016. Originally, testing was to occur in November of 2016 in the 31-Inch Mach 10 Air Tunnel at NASA Langley. This facility was chosen so that the XSWBLI phenomenon could be observed at much higher Mach numbers than had previously been attempted in ground test experiments. The model selected for this experiment, a 10° half-angle wedge with a sharp leading edge (described in detail in section II.B), had previously been used by Danehy et al. [10] for boundary layer transition studies using the nitric oxide planar laser-induced fluorescence (NO PLIF) flow visualization technique. In that work, it was determined that transition could be induced downstream of a single htrip = 1-mm tall, dtrip = 4-mm diameter cylindrical tripping element and that the streamwise location of the transition could be changed for a single Re∞ by changing the model angle-of-attack (AoA) (see Fig. A3 in Ref. [10] for more details). Based on the findings of that work, a decision was made to use the wedge model with the cylindrical tripping element to trip the boundary layer flow ahead of a cylindrical protuberance in order to achieve a XSWBLI. Unfortunately, the 31-Inch Mach 10 facility had been taken offline for repairs in October of 2016 and a decision was made to move the test to the 20-Inch Mach 6 facility. Since the behavior of the boundary layer with the chosen model configuration had not been studied before in that facility and the available test time was limited, the entry was considered to be exploratory and was used to collect spatially-resolved and time-resolved flow and surface visualization data that would be used to inform a second entry. Test techniques included simultaneous high-speed schlieren (HSS) captured at 160 kHz and high-speed pressure sensitive paint captured at 10 kHz as well as oil flow visualization, captured at 750 Hz. The second entry in the 20-Inch Mach 6 facility occurred in June and July of 2017. In this follow-on test, modifications to the wind tunnel model were made based on observations made during the first entry and included removing the cylindrical tripping element, increasing the size of the cylinder used to induce the SWBLI to increase the size of the interaction while simultaneously improving spatial resolution, and using a swept ramp array, similar to that described in Ref. [11], to trip the flow to turbulence. Simultaneous HSS (captured at 140 kHz, 100 kHz, and 40 kHz) and conventional IR thermography (captured at 30 Hz) imaging were performed simultaneously in this follow-on entry. This paper is intended to serve as a summary of the work performed during these two entries, to detail lessons learned from each entry, and to highlight some of the datasets acquired. Details on the experimental setup, model configuration, and techniques used are provided. Papers providing a more rigorous analysis of data acquired during the second entry, including statistical, spectral, and modal decomposition methods, can be found in Refs. [12,13]. An entry examining XSWBLI in the 31-Inch Mach 10 Blowdown Wind Tunnel facility is currently planned for mid-to-late calendar year 2019, pending the success of facility repairs. The work performed and described in this paper and the upcoming entry in the 31-Inch Mach 10 facility at NASA Langley have been made possible by NASA’s new facility funding model under the Aeronautics Evaluation and Test Capabilities (AETC) project. Wind Tunnel Facility All experiments discussed in this paper were performed in the 20-Inch Mach 6 Air Tunnel at NASA Langley Research Center. Specific details pertaining to this facility can be found in Refs. [14,15], with only a brief description of the facility provided here. For both entries, the nominal freestream unit Reynolds number was varied between 1.8×106 m-1 (0.5×106 ft-1) and 26.3×106 m-1 (8×106 ft-1). The nominal stagnation pressure was varied between 0.21 MPa and 3.33 MPa and the nominal stagnation temperature was varied between 480 K and 520 K to achieve the desired Re∞ condition. For all runs, the nominal freestream Mach number was 6. The nearly square test section is 520.7-mm (20.5-inches) wide by 508-mm (20-inches) high. Two 431.8-mm (17-inch) diameter windows made of Corning 7940, Grade 5F schlieren-quality glass serve as the side walls of the tunnel and provide optical access for the high-speed schlieren measurements. A rectangular window made of the same material as the side windows served as the top wall of the test section and provided optical access for the high-speed PSP and oil flow measurements. For the second entry, this top window was replaced with a Zinc Selenide (ZnSe) window with an anti-reflection coating capable of passing IR wavelengths between 8μm and 12μm with greater than 98% transmittance. The model was sting supported by a strut attached to a hydraulic system that allows for the model pitch angle to be adjusted between -5° to +55°. For the first entry, an initial pitch/pause sweep of the model AoA was performed to observe the resulting SWBLI. Ultimately, however, the sting pitch angle for this entry was fixed at +10.0° so that the angle of the top surface of the wedge relative to the streamwise axis of the tunnel (referred to herein as the plate angle, θplate), was θplate = 0°. For the second entry, θplate = 0° and θplate = -13.25° were initially tested with the swept ramp array (discussed in the following section) to determine which orientation produced conditions most favorable for XSWBLI to occur based on the heating signatures observed over the top surface of the model in the IR thermography images. Based on these initial tests, θplate = -13.25° was set for the remainder of the runs in the second entry. For both entries, any model changes were performed in a housing located beneath the closed test section. Prior to performing a run of the tunnel, the housing was sealed and the tunnel started. Once the appropriate freestream conditions were achieved, the model was injected into the test section using a hydraulic injection system. B. Model Geometry For all runs, a 10° half-angle (20° full-angle) wedge model with a sharp leading edge was used. The model is described in detail in Refs. [10,16]. The top surface of the sharp leading edge of the model extended 47.8 mm from its upstream-most edge to a junction with the upstream edge of a stainless steel top plate that then extended an (a) (c) (b) Fig. 1 (a) Schematic of top surface of wedge model with gas seeding insert, (b) perspective view of the model in the 20-Inch Mach 6 tunnel with centerline pressure orifices on sharp leading edge, and (c) a perspective view of the model with stainless steel (top) and SLA middle insert (bottom) during the first entry. Flow occurs from left to right.

Bathel, Brett F.↗
DOE OSTI2023

Elucidating Biomass-Derived Pyrolytic Lignin Structures from Demethylation Reactions through Density Functional Theory Calculations

Pyrolytic lignin is a fraction of pyrolysis oil that contains a wide range of phenolic compounds that can be used as intermediates to produce fuels and chemicals. However, the characteristics of the raw lignin structure make it difficult to establish a pyrolysis mechanism and determine pyrolytic lignin structures. Herein this study proposes dimer, trimer, and tetramer structures based on their relative thermodynamic stability for a hardwood lignin model in pyrolysis. Different configurations of oligomers were evaluated by varying the positions of the guaiacyl (G) and syringyl (S) units and the bonds βO4 and β5 in the hardwood model lignin through electronic structure calculations. The homolytic cleavage of βO4 bonds is assumed to occur and generate two free radical fragments. These can stabilize by taking hydrogen radicals that may be in solution during the intermediate liquid (pathway 1) formation before the thermal ejection. An alternative pathway (pathway 2) could occur when the radicals use intramolecular hydrogen, turning themselves into stable products. Subsequently, a demethylation reaction can take place, thus generating a methane molecule and new oligomeric lignin-derived molecules. The most probable resulting structures were studied. We used FTIR and NMR spectra of selected model compounds to evaluate our calculation approach. Thermophysical properties were calculated using group contribution methods. The results give insights into the lignin oligomer structures and how these molecules are formed. They also provide helpful information for the design of pyrolysis oil separation and upgrading equipment.

09 BIOMASS FUELS↗
DOE OSTI2023

Chaconne: A Statistical Approach to Nonlocal Compression for Supervised Learning, Semi-Supervised Learning, and Anomaly Detection

This project developed a novel statistical understanding of compression analytics (CA), which has challenged and clarified some core assumptions about CA, and enabled the development of novel techniques that address vital challenges of national security. Specifically, this project has yielded the development of novel capabilities including 1. Principled metrics for model selection in CA, 2. Techniques for deriving/applying optimal classification rules and decision theory to supervised CA, including how to properly handle class imbalance and differing costs of misclassification, 3. Two techniques for handling nonlocal information in CA, 4. A novel technique for unsupervised CA that is agnostic with regard to the underlying compression algorithm, 5. A framework for semisupervised CA when a small number of labels are known in an otherwise large unlabeled dataset. 6. The academic alliance component of this project has focused on the development of a novel exemplar-based Bayesian technique for estimating variable length Markov models (closely related to PPM [prediction by partial matching] compression techniques). We have developed examples illustrating the application of our work to text, video, genetic sequences, and unstructured cybersecurity log files.

99 GENERAL AND MISCELLANEOUS↗
DOE OSTI2026

Scaling open-weight large language models for hydropower regulatory information extraction: A systematic analysis

Information extraction from regulatory and technical documents using large language models (LLMs) involves practical trade-offs between extraction quality and computational cost. We evaluate eight open-weight LLMs spanning 0.6B–70B parameters on hydropower licensing documents and report deployment-oriented evidence under a unified extraction schema and evaluation protocol. Across the model set, we observe clear scale-dependent trends in both baseline extraction quality and the effectiveness of reflective reasoning (self-checking) under our fixed-prompt, no-augmentation setting. Mid-scale models often provide a favorable balance of accuracy and efficiency, whereas the smallest models show limited or inconsistent gains from the reasoning variants tested. Larger models achieve the highest overall F1 scores but incur substantially greater compute and infrastructure requirements. We further find that reliability failure modes can distort conventional metrics in this domain: in particular, high recall can coincide with systematic extraction errors when models fabricate values for fields that are absent from the source text, underscoring the importance of conservative null handling and evidence-grounded evaluation. Overall, our study provides a reproducible resource–performance comparison for open-weight LLM-based extraction in hydropower regulatory documentation and offers practical guidance for model selection under different deployment constraints.

Evaluation protocol↗
DOE OSTI2022

Atacama Cosmology Telescope: Constraints on prerecombination early dark energy

The early dark energy (EDE) scenario aims to increase the value of the Hubble constant (H0) inferred from cosmic microwave background (CMB) data over that found in the standard cosmological model (Λ CDM ), via the introduction of a new form of energy density in the early Universe. The EDE component briefly accelerates cosmic expansion just prior to recombination, which reduces the physical size of the sound horizon imprinted in the CMB. Previous work has found that nonzero EDE is not preferred by Planck CMB power spectrum data alone, which yield a 95% confidence level (C.L.) upper limit f EDE < 0.087 on the maximal fractional contribution of the EDE field to the cosmic energy budget. In this paper, we fit the EDE model to CMB data from the Atacama Cosmology Telescope (ACT) data release 4. We find that a combination of ACT, large-scale Planck TT (similar to WMAP), Planck CMB lensing, and BAO data prefers the existence of EDE at >99.7 % C .L . : f EDE = 0.091$^{+0.020}_{-0.036}$, with H 0 = 70.9$^{+1.0}_{-2.0}$ km/s/Mpc (both 68% C.L.). From a model-selection standpoint, we find that EDE is favored over Λ CDM by these data at roughly 3 σ significance. In contrast, a joint analysis of the full Planck and ACT data yields no evidence for EDE, as previously found for Planck alone. We show that the preference for EDE in ACT alone is driven by its TE and EE power spectrum data. The tight constraint on EDE from Planck alone is driven by its high-ℓ TT power spectrum data. Understanding whether these differing constraints are physical in nature, due to systematics, or simply a rare statistical fluctuation is of high priority. Here, the best-fit EDE models to ACT and Planck exhibit coherent differences across a wide range of multipoles in TE and EE, indicating that a powerful test of this scenario is anticipated with near-future data from ACT and other ground-based experiments.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2023

Explosive Yield Estimation Using Regional Seismic Moment Tensors

Here, we use the Pasyanos and Chiang (2022) data set to calculate the seismic moment M 0 for each explosion and use the measured explosive yield W to validate the W~M 0 relationship in Denny and Johnson (1991; hereafter, DJ91). The M 0 is corrected by transforming to a potency tensor and applying more appropriate near-source geophysical parameter values in the moment estimate. The mean residual between observed and predicted yield is near zero; however, the standard deviation of the residuals results in an F-value (a 95% confidence factor) of about 5. We re-estimate the coefficients in the DJ91 model and find similar values and only a slight improvement in the F-value. Next, we embark on a similar model selection process as DJ91, allowing for non-cube-root yield scaling and other plausible near-source elastic moduli. As was found by DJ91, the yield dependence is not significantly different from unity, and a cube root assumption is valid. Therefore, we yield scale the seismic moment and test the significance of all plausible explanatory variables. Isotropic moment performs better in the response variable than total moment. The preference for isotropic moment could be due to its relationship to volume change, which would be more directly affected by explosive yield. Surprisingly, we find that the overburden pressure, which is a function of depth, is not a significant parameter in the model. We hypothesize that this is due to the competing depth effects on source asymmetry and the incorporation of depth in the Green’s functions used to calculate the seismic moment tensors. Importantly, this emphasizes that only seismic moment tensor-derived moments should be used in these models. After removing insignificant model parameters, we are left with a simple model to predict explosive yield $\widehat{W}$ in kt from isotropic moment M I in N·m, $\widehat{W}$=κ –1.4132 10 0.035626GP M I , in which κ and GP are the near-source bulk modulus and gas porosity in Pa and %, respectively. The F-value for this model is approximately 3.

58 GEOSCIENCES↗
DOE OSTI2025

Chapter 4 - Recent Advances in Identification of Differential Equations from Noisy Data: IDENT Review

Differential equations and numerical methods are extensively used to model various real-world phenomena in science and engineering. With modern developments, we aim to find the underlying differential equation from a single observation of time-dependent data. If we assume that the differential equation is a linear combination of various linear and nonlinear differential terms, then the identification problem can be formulated as solving a linear system. The goal then reduces to finding the optimal coefficient vector that best represents the time derivative of the given data. We review some recent works on the identification of differential equations. We find some common themes for the improved accuracy: (i) The formulation of linear system with proper denoising is important, (ii) how to utilize sparsity and model selection to find the correct coefficient support needs careful attention, and (iii) there are ways to improve the coefficient recovery. We present an overview and analysis of recent developments on the topic.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

A Comparative Study of Machine Learning Algorithms for Industry-Specific Freight Generation Model

According to Bureau of Transportation Statistics, the U.S. transportation system handled 14,329 million ton-miles of freight per day in 2020. Understanding the generation of these freight shipments is crucial for transportation researchers, planners, and policymakers to design and plan for a more efficient and connected freight transportation system. Traditionally, the freight generation modeling has been based on Ordinary Least Square (OLS) regression, although more advanced Machine Learning (ML) algorithms have been evaluated and proven to have excellent performance in various transportation applications in recent years. Furthermore, one modeling approach applied for one industry might not always be applicable for another as their freight generation logics can be quite different. The objective of this study is to apply and evaluate alternative ML algorithms in the estimation of freight generation for each of 45 industry types. Seven alternative ML algorithms, along with the base OLS regression, were evaluated and compared. In addition, the study considered different combinations of variables in both the original and logarithmic form as well as hyperparameters of those ML algorithms in the model selection for each industry type. The results showed statistically significant improvements in the root mean square error reduction by the alternative ML algorithms over the OLS for over 80% of cases. The study suggests utilizing the alternative ML algorithms can reduce the root mean square error by about 30%, depending on industry types.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

Development and Implementation of a New AI-Based Tool to Support Fast Reactor Software Model Generation and Validation

This report summarizes FY26 work to develop Maggie, an artificial intelligence-based assistant designed to support software model generation and validation activities for fast reactor analysis codes. The project established a modular, code-agnostic software architecture that separates reusable agent capabilities from code-specific knowledge and tools, with initial implementation focused on the FRP-supported fast reactor safety analysis code SAS4A/SASSYS1 (SAS). A curated SAS-specific knowledge base was assembled from the code manual, training materials, historical analysis reports, and representative input files, and was integrated through retrieval-augmented generation to ground Maggie’s responses in authoritative sources. Maggie was deployed on the internal Argonne network, where it demonstrated practical user-facing capability as a chatbot for answering natural language questions about SAS and retrieving relevant technical information. Demonstration cases also showed that Maggie can generate useful snippets of SAS input for selected modeling tasks, while highlighting current limitations in reliability and consistency for more complex input generation tasks. Overall, the FY26 effort established the technical foundation for an AI-assisted capability intended to improve the efficiency, consistency, and accessibility of fast reactor software model development at Argonne and, with further improvements, to support eventual use by the broader fast reactor community, including industry users of FRP-supported analysis tools.

Thomas, Rachel [Argonne National Laboratory (ANL),↗
NASA NTRS1986

Simulated flight acoustic investigation of treated ejector effectiveness on advanced mechanical suppresors for high velocity jet noise reduction

Ten scale-model nozzles were tested in an anechoic free-jet facility to evaluate the acoustic characteristics of a mechanically suppressed inverted-velocity-profile coannular nozzle with an accoustically treated ejector system. The nozzle system used was developed from aerodynamic flow lines evolved in a previous contract, defined to incorporate the restraints imposed by the aerodynamic performance requirements of an Advanced Supersonic Technology/Variable Cycle Engine system through all its mission phases. Accoustic data of 188 test points were obtained, 87 under static and 101 under simulated flight conditions. The tests investigated variables of hardwall ejector application to a coannular nozzle with 20-chute outer annular suppressor, ejector axial positioning, treatment application to ejector and plug surfaces, and treatment design. Laser velocimeter, shadowgraph photograph, aerodynamic static pressure, and temperature measurement were acquired on select models to yield diagnositc information regarding the flow field and aerodynamic performance characteristics of the nozzles.

Brausch, J. F.↗
DOE OSTI2021

Comparison of Deterministic and Statistical Models for Water Quality Compliance Forecasting in the San Joaquin River Basin, California

Model selection for water quality forecasting depends on many factors including analyst expertise and cost, stakeholder involvement and expected performance. Water quality forecasting in arid river basins is especially challenging given the importance of protecting beneficial uses in these environments and the livelihood of agricultural communities. In the agriculture-dominated San Joaquin River Basin of California, real-time salinity management (RTSM) is a state-sanctioned program that helps to maximize allowable salt export while protecting existing basin beneficial uses of water supply. The RTSM strategy supplants the federal total maximum daily load (TMDL) approach that could impose fines associated with exceedances of monthly and annual salt load allocations of up to $1 million per year based on average year hydrology and salt load export limits. The essential components of the current program include the establishment of telemetered sensor networks, a web-based information system for sharing data, a basin-scale salt load assimilative capacity forecasting model and institutional entities tasked with performing weekly forecasts of river salt assimilative capacity and scheduling west-side drainage export of salt loads. Web-based information portals have been developed to share model input data and salt assimilative capacity forecasts together with increasing stakeholder awareness and involvement in water quality resource management activities in the river basin. Two modeling approaches have been developed simultaneously. The first relies on a statistical analysis of the relationship between flow and salt concentration at three compliance monitoring sites and the use of these regression relationships for forecasting. The second salt load forecasting approach is a customized application of the Watershed Analysis Risk Management Framework (WARMF), a watershed water quality simulation model that has been configured to estimate daily river salt assimilative capacity and to provide decision support for real-time salinity management at the watershed level. Analysis of the results from both model-based forecasting approaches over a period of five years shows that the regression-based forecasting model, run daily Monday to Friday each week, provided marginally better performance. However, the regression-based forecasting model assumes the same general relationship between flow and salinity which breaks down during extreme weather events such as droughts when water allocation cutbacks among stakeholders are not evenly distributed across the basin. A recent test case shows the utility of both models in dealing with an exceedance event at one compliance monitoring site recently introduced in 2020.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

How reliable is distribution of relaxation times (DRT) analysis? A dual regression-classification perspective on DRT estimation, interpretation, and accuracy

The distribution of relaxation times (DRT) has gained increasing attention and adoption in recent years as a versatile method for analyzing electrochemical impedance spectroscopy (EIS) data obtained from complex devices like fuel cells, electrolyzers, and batteries. The DRT deconvolutes the impedance without a priori specification of a generative model, which is especially useful for interpretation and model selection when the governing principles of the system under study are not fully understood. However, DRT estimation is an ill-posed inversion problem that must be addressed with a subjective choice of regularization and tuning, which leaves substantial risk of misleading interpretations of EIS data. In this work, we suggest a new classification view of the DRT inversion to clarify DRT estimation and interpretation. We introduce a dual regression-classification framework that unifies the classification and regression views of the DRT inversion with wide-reaching implications for DRT analysis. The dual framework is employed to demonstrate a new kind of DRT inversion algorithm and develop novel evaluation metrics that capture previously ignored aspects of DRT accuracy. These approaches are applied to both synthetic data and experimental spectra collected from a protonic ceramic fuel cell and a lithium-ion battery to illustrate their broad utility. The dual inversion algorithm shows promising performance for accurate DRT estimation and autonomous model identification, while the dual evaluation approach produces metrics that meaningfully assess the strengths and risks of DRT algorithms. Here this work provides valuable insight for both practical application of the DRT to experimental data and further development of EIS analysis methods.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Field evaluation of semi‐automated moisture estimation from geophysics using machine learning

Geophysical methods can provide three-dimensional (3D), spatially continuous estimates of soil moisture. However, point-to-point comparisons of geophysical properties to measure soil moisture data are frequently unsatisfactory, resulting in geophysics being used for qualitative purposes only. This is because (1) geophysics requires models that relate geophysical signals to soil moisture, (2) geophysical methods have potential uncertainties resulting from smoothing and artifacts introduced from processing and inversion, and (3) results from multiple geophysical methods are not easily combined within a single soil moisture estimation framework. To investigate these potential limitations, an irrigation experiment was performed wherein soil moisture was monitored through time, and several surface geophysical datasets indirectly sensitive to soil moisture were collected before and after irrigation: ground penetrating radar, electrical resistivity tomography (ERT), and frequency domain electromagnetics (FDEM). Data were exported in both raw and processed form, and then snapped to a common 3D grid to facilitate moisture prediction by standard calibration techniques, multivariate regression, and machine learning. A combination of inverted ERT data, raw FDEM, and inverted FDEM data was most informative for predicting soil moisture using a random regression forest model (one-thousand 60/40 training/test cross-validation folds produced root mean squared errors ranging from 0.025–0.046 cm 3 /cm 3 ). This cross-validated model was further supported by a separate evaluation using a test set from a physically separate portion of the study area. Machine learning was conducive to a semi-automated model-selection process that could be used for other sites and datasets to locally improve accuracy.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

Distributed out-of-memory NMF on CPU/GPU architectures

We propose an efficient distributed out-of-memory implementation of the non-negative matrix factorization (NMF) algorithm for heterogeneous high-performance-computing systems. The proposed implementation is based on prior work on NMFk, which can perform automatic model selection and extract latent variables and patterns from data. In this work, we extend NMFk by adding support for dense and sparse matrix operation on multi-node, multi-GPU systems. The resulting algorithm is optimized for out-of-memory problems where the memory required to factorize a given matrix is greater than the available GPU memory. Memory complexity is reduced by batching/tiling strategies, and sparse and dense matrix operations are significantly accelerated with GPU cores (or tensor cores when available). Input/output latency associated with batch copies between host and device is hidden using CUDA streams to overlap data transfers and compute asynchronously, and latency associated with collective communications (both intra-node and inter-node) is reduced using optimized NVIDIA Collective Communication Library (NCCL) based communicators. Benchmark results show significant improvement, from 32X to 76x speedup, with the new implementation using GPUs over the CPU-based NMFk. Good weak scaling was demonstrated on up to 4096 multi-GPU cluster nodes with approximately 25,000 GPUs when decomposing a dense 340 Terabyte-size matrix and an 11 Exabyte-size sparse matrix of density 10 -6 .

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

Inverse design of hypoeutectoid pearlite steel microstructures using a deep learning and genetic algorithm optimization framework

Goal-oriented microstructure design in metallic materials is a challenging task due to complex structure-property relationships. Traditional experimental and computational approaches are time-intensive and economically inefficient, limiting their applicability for large-scale design space exploration. Here, in this work, we propose an end-to-end framework that integrates deep learning models with genetic optimization to design microstructures with targeted mechanical properties. Deep learning models enable accurate forward design, while their integration with genetic optimization enables efficient inverse design within a few hours, compared to days or weeks using conventional finite element simulations. The framework combines experimental characterization and finite element modeling to analyze the influence of microstructural features on the mechanical behavior of hypoeutectoid steels. Data from both experiments and simulations are used to train the deep learning models. To demonstrate its effectiveness, we apply the framework to 0.63% carbon steel with proeutectoid ferrite and pearlite phases, commonly used in industrial applications. In this study, 2D microstructures were used for modeling, selected primarily for computational efficiency and to establish proof of concept. The framework successfully optimizes microstructures for targeted yield strength, ultimate strength, and stress concentration factors while significantly reducing computational time. Beyond hypoeutectoid steels, this scalable framework can be extended to other material systems and integrated with additive manufacturing, offering an efficient approach for accelerating microstructure design for specific engineering applications.

ConvLSTM↗
DOE OSTI2026

Toward Rapid Actinium-225 Purification via Membrane Adsorbers with Covalently Tethered Diglycolamide Ligands

Extractive diglycolamide (DGA) resins are used in several state-of-the-art techniques for purifying 225 Ac, a promising radiometal for targeted alpha therapy. Unfortunately, separation processes that rely on resins are often limited to slow flow rates, high elution volumes, and long processing times. Membrane adsorbers functionalized with DGA ligands are an alternative separation material that may overcome these challenges. This work presents (1) the synthesis of an aminated tetrahexyldiglycolamide ligand, (2) the covalent tethering of the ligand to electrospun poly(vinylbenzyl chloride) fiber mats, and (3) the adsorption and desorption of La(III) and 225 Ac. Chemical and physical characterization supports the covalent tethering of the ligand to the fiber mat, as well as the preservation of the fiber surface area and porosity after functionalization. Equilibrium adsorption experiments were performed with stable La(III) and radioactive 225 Ac. Trends in affinity are consistent between commercial resins and the synthesized membrane adsorbers; however, the Langmuir constants and the maximum binding capacity of the membrane adsorbers were generally lower than the resins. Despite these differences, the modeled selectivity for an equimolar solution of La(III)/ 225 Ac in 10 M nitric acid is 57. Furthermore, 225 Ac is rapidly desorbed from the fibers in 10 M nitric acid (<20 min). The La(III)/ 225 Ac selectivity and rapid 225 Ac desorption indicate this class of materials is promising for rapid radioanalytical separations.

07 ISOTOPE AND RADIATION SOURCES↗
DOE OSTI2020

Forecasting the Decay of Hybrid Perovskite Performance Using Optical Transmittance or Reflected Dark-Field Imaging

The practicality and economic viability of hybrid perovskite solar cells hinge on their operational lifetime, and methods for forecasting the performance of perovskites under different operational stresses are urgently needed. Here, we explore the evolution of material-level optoelectronic properties as MAPbI 3 degrades and discover universal behaviors where the carrier diffusion length (L D ) decays before quasi-Fermi-level splitting (ΔE F ), regardless of the specific stress protocol (oxygen, humidity, thermal stress, or a combination). We employ a machine learning greedy feature selection model that uses initially measured properties to predict the time it takes L D to decrease to 85% of its initial value with a prediction accuracy of 12.8%. This model reveals a strong correlation between the initial rate of transmittance change and the time until loss of transport. We translate this material-level finding to photovoltaic device-level forecasting by demonstrating that the rate of change of transmittance is equivalent to the rate of change of the spatial standard deviation of dark-field image intensity (i.e., scattered light intensity) collected in reflection mode (and thus applicable to devices with opaque contacts). Furthermore, this work demonstrates that transmittance and scattering methods are highly effective for accelerated material (and device) stability evaluation and forecasting.

14 SOLAR ENERGY↗