Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Yield 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 289 records · Page 16
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Accelerating defect predictions in semiconductors using graph neural networks

First-principles computations reliably predict the energetics of point defects in semiconductors but are constrained by the expense of using large supercells and advanced levels of theory. Machine learning models trained on computational data, especially ones that sufficiently encode defect coordination environments, can be used to accelerate defect predictions. Here, we develop a framework for the prediction and screening of native defects and functional impurities in a chemical space of group IV, III–V, and II–VI zinc blende semiconductors, powered by crystal Graph-based Neural Networks (GNNs) trained on high-throughput density functional theory (DFT) data. Using an innovative approach of sampling partially optimized defect configurations from DFT calculations, we generate one of the largest computational defect datasets to date, containing many types of vacancies, self-interstitials, anti-site substitutions, impurity interstitials and substitutions, as well as some defect complexes. We applied three types of established GNN techniques, namely crystal graph convolutional neural network, materials graph network, and Atomistic Line Graph Neural Network (ALIGNN), to rigorously train models for predicting defect formation energy (DFE) in multiple charge states and chemical potential conditions. We find that ALIGNN yields the best DFE predictions with root mean square errors around 0.3 eV, which represents a prediction accuracy of 98% given the range of values within the dataset, improving significantly on the state-of-the-art. We further show that GNN-based defective structure optimization can take us close to DFT-optimized geometries at a fraction of the cost of full DFT. The current models are based on the semi-local generalized gradient approximation-Perdew–Burke–Ernzerhof (PBE) functional but are highly promising because of the correlation of computed energetics and defect levels with higher levels of theory and experimental data, the accuracy and necessity of discovering novel metastable and low energy defect structures at the PBE level of theory before advanced methods could be applied, and the ability to train multi-fidelity models in the future with new data from non-local functionals. The DFT-GNN models enable prediction and screening across thousands of hypothetical defects based on both unoptimized and partially optimized defective structures, helping identify electronically active defects in technologically important semiconductors.

Rahman, Md Habibur (ORCID:000000027705984X)↗
DOE OSTI2023

Experimental and computational design tools for industrial drying processes: Challenges in process‐limit prediction

Abstract The coating and drying of inks and slurries are important steps to manufacture a plethora of products. Drying processes, particularly, comprise energy‐intensive steps that affect product cost and quality. Prior work has highlighted failures of various multicomponent diffusivity models to conserve mass in dryer modeling and challenges in predicting process limits given variability in published values of key thermodynamic parameters. Herein, we develop a computational model and benchtop drying experiments to investigate these concerns for drying polymer‐laden coatings. Model predictions of process limits in a single‐zone drying oven demonstrate that published variability in Flory–Huggins parameter yields large variations in predicted operating temperatures above which blistering occurs. This indicates that caution should be exercised when choosing approaches to obtain or predict the Flory–Huggins parameter, and that both benchtop drying experiments and a set of additional experiments, such as sorption experiments, are needed to fully characterize and optimize a given drying process.

Parrish, Chance↗
DOE OSTI2021

New Fe 59 Stellar Decay Rate with Implications for the Fe 60 Radioactivity in Massive Stars

The discrepancy between observations from γ-ray astronomy of the 60 Fe/ 26 Al γ-ray flux ratio and recent calculations is an unresolved puzzle in nuclear astrophysics. The stellar β-decay rate of 59 Fe is one of the major nuclear uncertainties impeding us from a precise prediction. The important Gamow-Teller strengths from the low-lying states in 59 Fe to the 59 Co ground state are measured for the first time using the exclusive measurement of the 59 Co(t,3He+γ) 59 Fe charge-exchange reaction. The new stellar decay rate of 59 Fe is a factor of 3.5±1.1 larger than the currently adopted rate at T=1.2 GK. Stellar evolution calculations show that the 60 Fe production yield of an 18 solar mass star is decreased significantly by 40% when using the new rate. Our result eliminates one of the major nuclear uncertainties in the predicted yield of 60 Fe and alleviates the existing discrepancy of the 60 Fe/ 26 Al ratio.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2024

Side-by-Side Comparison of Subhourly Clipping Models

Over the past several years there have been numerous attempts at quantifying the inherent power clipping of inverters due to subhourly irradiance variability that is not captured in hourly PV performance models. Different models have been proposed to correct for these clipping losses in PV performance estimates, including matrix lookup models, distribution modeling of the PV power performance within a given hour, and machine learning methods. To date, there have been few comprehensive quantitative comparisons of these inverter clipping correction modeling approaches to evaluate the effectiveness of these approaches in predicting the actual behavior of PV system inverter clipping. In this study, we perform such a comparison, evaluating the Allen and Walker correction loss modeling approaches recently implemented in the System Advisor Model (SAM) against clipping losses modeled with 1-minute climate data. These comparisons were performed across a variety of climate locations and inverter loading ratios to thoroughly analyze the effectiveness of these modeling approaches relative to each other. Results from this analysis reveal that both clipping correction approaches improve annual energy accuracy to within 2% of 1-minute modeled energy yield. The two models predict annual clipping loss more accurately than simple hourly power limit clipping, with the Allen method typically being slightly more accurate at typical ILR values and the Walker method often being slightly more accurate at high ILR values The models can improve accuracy over the status quo clipping approach up to 3 percentage points in systems with ILR of 2.0, showing the importance of this modeling factor in energy yield estimates.

accuracy↗
DOE OSTI2025

Observation of kinetic mix enhancement in thin-shell OMEGA implosions

Recent separated reactant experiments for thin-shell (6 µ⁢m) shock-driven implosions on OMEGA have demonstrated significant mix from a buried deuterated layer of the shell into the hot spot. Time resolved D 3 He-p reaction history data demonstrate a (50 ± 20)⁢ ps shift earlier in peak nuclear emission for separated reactant experiments relative to control, in contrast to past experimental data for thicker, 20 µ⁢m shells with no laser burn through that show a 75 ps delay due to the time required for hydrodynamic instabilities to develop. This contrast suggests that the mix mechanism was not hydrodynamic. Ion kinetic simulations utilizing fall line analyses show much closer agreement with mix yield and temperature than diffusion models, predicting a D 3 He-p mix yield of 1.7 × 10 9 as compared to the experimental value of 9.3⁢ (±2.1) × 10 8 . This is three orders of magnitude closer than the fall line analysis from a hydrodynamic simulation with an inline diffusive mix model, which suggests minimal mix and D 3 He-p yields of 5×10 5 . This makes kinetic mechanisms the only feasible explanation for the mix seen, demonstrating impact of a non-standard mix mechanism. An analytical model of this kinetic mix mechanism suggests that it can remain significant in situations when the shell expands significantly to low densities, and diffusive models predict negligible mix. Finally, kinetic mix will impact multiple types of high energy density, laser-driven fusion experiments including high-adiabat direct drive cryoexperiments, nuclear cross section experiments, and thin-shell polar direct drive experiments used to tune heat conduction models.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2021

Assessing cathode property prediction via exchange-correlation functionals with and without long-range dispersion corrections

In this work, we benchmark calculated interlayer spacings, average topotactic voltages, thermodynamic stabilities, and band gaps in layered lithium transition-metal oxides (TMOs) and their de-lithiated counterparts, which are used in lithium-ion batteries as positive electrode materials, against available experimental data. Specifically, we examine the accuracy of properties calculated within density functional theory (DFT) using eight different treatments of electron exchange-correlation: the strongly constrained and appropriately normed (SCAN) and Perdew–Burke–Ernzerhof (PBE) density functionals, Hubbard-U-corrected SCAN and PBE (i.e., SCAN+U and PBE+U), and SCAN(+U) and PBE(+U) with added long-range dispersion (D) interactions (i.e., DFT(+U)+D). van der Waals interactions are included respectively via the revised Vydrov-Van Voorhis (rVV10) for SCAN(+U) and the DFT-D3 for PBE(+U). We find that SCAN-based functionals predict larger voltages due to an underestimation of stability of the MO2 systems, while also predicting smaller interlayer spacings compared to their PBE-based counterparts. Furthermore, adding dispersion corrections to PBE has a greater effect on voltage predictions and interlayer spacings than with SCAN, indicating that DFT-SCAN – despite being a ground-state theory – fortuitously captures some short and medium-range dispersion interactions better than PBE. While SCAN-based and PBE-based functionals yield qualitatively similar band gap predictions, there is no significant quantitative improvement of SCAN-based functionals over the corresponding PBE-based versions. Finally, we expect SCAN-based functionals to yield more accurate property predictions than the respective PBE-based functionals for most TMOs, given SCAN's stronger theoretical underpinning and better predictions of systematic trends in interlayer spacings, intercalation voltages, and band gaps obtained in this work.

25 ENERGY STORAGE↗
DOE OSTI2025

Deep potential molecular dynamics simulations of ion-enhanced etching of silicon by atomic chlorine

The continued development of plasma-assisted processing techniques requires a fundamental understanding of plasma-surface interactions. Molecular dynamics (MD) simulations have been employed to complement experimental studies and better understand the properties of such systems. Recently, machine learning (ML) methods have enabled the development of ab initio-based interatomic potentials, which can be generalized to complex combinations of multiple atom types. In this work, we use ML potentials developed using the Deep Potential Molecular Dynamics (DeepMD) framework to provide a model of ion-enhanced etching of Si by Cl atoms. We demonstrate the importance of proper selection of the training data set to the accuracy of the DeepMD model and compare our results to MD results using empirical potentials, as well as to experimental measurements. Exposure of undoped Si at 300 K to thermal Cl atoms yields a steady-state Cl coverage of 1.25 monolayers, which is slightly lower than the value obtained in previous experimental studies. Predictions of Si etch yields by simultaneous Cl atom and Ar + ion impacts as a function of ion energy, neutral to ion flux ratio, and angle of incidence of the ions are in reasonably good agreement with classical MD results and experimental measurements. Finally, etch yields and SiCl x mixed layer thicknesses during simultaneous bombardment of the Si(100) surface by Cl atoms and Cl + ions are in good agreement with experimental data. In conclusion, the present work is a necessary condition for the extension of the DeepMD procedure to more complex systems of interest in plasma-surface interactions.

Artificial neural networks↗
DOE OSTI2022

A Model for Radiolysis in a Flowing-Water Target during High-Intensity Proton Irradiation

At the Facility for Rare Isotope Beams (FRIB), interactions between heavy-ion beams and beam-dump water will create a wide variety of radionuclides which can be accessed by a technique known as “isotope harvesting”. However, irradiation of water is always accompanied by the creation of numerous radical, ionic, and molecular radiolysis products. Some of the radiolysis products have sufficiently long lifetimes to accumulate in the irradiated water and affect the harvesting chemistry. Here we investigate the formation of hydrogen peroxide, molecular hydrogen, and molecular oxygen during a high-intensity proton irradiation of a flowing-water isotope-harvesting target and compare the experimental results to simulations. The simulations kinetically model the chemical reactions occurring in the homogeneous phase of radiolysis in flowing water and establish an “effective yield”. In both the experiment and simulations, the bulk quantities of H 2 , H 2 O 2 , and O 2 are considerably lower than predicted by primary radiolysis yields (escape yields), meaning that in the high beam intensity regime the homogeneous phase reactions have a considerable impact on the overall chemical composition of the water. Further, it could be shown that for radiation which is characterized by a limited linear energy transfer, such as the here applied protons, the bulk outcome of the microscopic kinetic modeling could be estimated by a simplified steady-state model.

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

Ion Transport in Concentrated Crosslinked Solid Polymer Electrolytes

Crosslinking polymers is a common approach to create mechanically stable solid materials such as polymer electrolytes for lithium batteries. In conventional liquid electrolytes, the solvent molecules move freely to accommodate the field-induced motion of ions. However, in crosslinked polymer electrolytes, the rearrangement of polymer segments is constrained by the deformation limits of the network. Herein, we develop a new transport model that accounts for both the formation of concentration gradients and the elasticity of the electrolyte. The elasticity is incorporated by adding an additional term related to the entropy of crosslinked strands to the electrochemical potential of the salt. The resulting Crosslink Model contains two adjustable parameters: $\mathcal{N}$, the average number of monomers in a strand, and λ crit , the maximum strain the network can sustain. These solid-like constraints produce singularities in the governing transport equations, fundamentally altering the concentration profiles. Plateaus in salt concentrations emerge near the electrodes, and network elasticity introduces a strain overpotential. When compared to a Baseline Model ($\mathcal{N}$ → ∞, equivalent to concentrated solution theory), which predicts steepest gradients near the electrodes, both models yield similar current–voltage relationships. Model predictions are compared to electrochemical data for a poly(ethylene oxide)-based crosslinked polymer electrolyte.

Patel, Vivaan [University of California, Berkeley,↗
DOE OSTI2022

Efficient machine-learning model for fast assessment of elastic properties of high-entropy alloys

We combined descriptor-based analytical models for stiffness-matrix and elastic-moduli with mean-field methods to accelerate assessment of technologically useful properties of high-entropy alloys, such as strength and ductility. Model training for elastic properties uses Sure-Independence Screening (SIS) and Sparsifying Operator (SO) method yielding an optimal analytical model, constructed with meaningful atomic features to predict target properties. Computationally inexpensive analytical descriptors were trained using a database of elastic properties determined from density functional theory for binary and ternary subsets of Nb-Mo-Ta-W-V refractory alloys. The optimal Elastic-SISSO models, extracted from an exponentially large feature space, give an extremely accurate prediction of target properties, similar to or better than other models, with some verified from existing experiments. Here we also show that electronegativity variance and elastic-moduli can directly predict trends in ductility and yield strength of refractory HEAs, and reveals promising alloy concentration regions.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Subsurface Characterization and Machine Learning Predictions at Brady Hot Springs Results

Geothermal power plants typically show decreasing heat and power production rates over time. Mitigation strategies include optimizing the management of existing wells - increasing or decreasing the fluid flow rates across the wells - and drilling new wells at appropriate locations. The latter is expensive, time-consuming, and subject to many engineering constraints, but the former is a viable mechanism for periodic adjustment of the available fluid allocations. Data and supporting literature from a study describing a new approach combining reservoir modeling and machine learning to produce models that enable strategies for the mitigation of decreased heat and power production rates over time for geothermal power plants. The computational approach used enables translation of sets of potential flow rates for the active wells into reservoir-wide estimates of produced energy and discovery of optimal flow allocations among the studied sets. In our computational experiments, we utilize collections of simulations for a specific reservoir (which capture subsurface characterization and realize history matching) along with machine learning models that predict temperature and pressure timeseries for production wells. We evaluate this approach using an "open-source" reservoir we have constructed that captures many of the characteristics of Brady Hot Springs, a commercially operational geothermal field in Nevada, USA. Selected results from a reservoir model of Brady Hot Springs itself are presented to show successful application to an existing system. In both cases, energy predictions prove to be highly accurate: all observed prediction errors do not exceed 3.68% for temperatures and 4.75% for pressures. In a cumulative energy estimation, we observe prediction errors that are less than 4.04%. A typical reservoir simulation for Brady Hot Springs completes in approximately 4 hours, whereas our machine learning models yield accurate 20-year predictions for temperatures, pressures, and produced energy in 0.9 seconds. This paper aims to demonstrate how the models and techniques from our study can be applied to achieve rapid exploration of controlled parameters and optimization of other geothermal reservoirs. Includes a synthetic, yet realistic, model of a geothermal reservoir, referred to as open-source reservoir (OSR). OSR is a 10-well (4 injection wells and 6 production wells) system that resembles Brady Hot Springs (a commercially operational geothermal field in Nevada, USA) at a high level but has a number of sufficiently modified characteristics (which renders any possible similarity between specific characteristics like temperatures and pressures as purely random). We study OSR through CMG simulations with a wide range of flow allocation scenarios. Includes a dataset with 101 simulated scenarios that cover the period of time between 2020 and 2040 and a link to the published paper about this project, where we focus on the Machine Learning work for predicting OSR's energy production based on the simulation data, as well as a link to the GitHub repository where we have published the code we have developed (please refer to the repository's readme file to see instructions on how to run the code). Additional links are included to associated work led by the USGS to identify geologic factors associated with well productivity in geothermal fields. Below are the high-level steps for applying the same modeling + ML process to other geothermal reservoirs: 1. Develop a geologic model of the geothermal field. The location of faults, upflow zones, aquifers, etc. need to be accounted for as accurately as possible 2. The geologic model needs to be converted to a reservoir model that can be used in a reservoir simulator, such as, for instance, CMG STARS, TETRAD, or FALCON 3. Using native state modeling, the initial temperature and pressure distributions are evaluated, and they become the initial conditions for dynamic reservoir simulations 4....

15 GEOTHERMAL ENERGY↗
DOE OSTI2022

Modeling Subsurface Performance of a Geothermal Reservoir Using Machine Learning

Geothermal power plants typically show decreasing heat and power production rates over time. Mitigation strategies include optimizing the management of existing wells—increasing or decreasing the fluid flow rates across the wells—and drilling new wells at appropriate locations. The latter is expensive, time-consuming, and subject to many engineering constraints, but the former is a viable mechanism for periodic adjustment of the available fluid allocations. In this study, we describe a new approach combining reservoir modeling and machine learning to produce models that enable such a strategy. Our computational approach allows us, first, to translate sets of potential flow rates for the active wells into reservoir-wide estimates of produced energy, and second, to find optimal flow allocations among the studied sets. In our computational experiments, we utilize collections of simulations for a specific reservoir (which capture subsurface characterization and realize history matching) along with machine learning models that predict temperature and pressure timeseries for production wells. We evaluate this approach using an “open-source” reservoir we have constructed that captures many of the characteristics of Brady Hot Springs, a commercially operational geothermal field in Nevada, USA. Selected results from a reservoir model of Brady Hot Springs itself are presented to show successful application to an existing system. In both cases, energy predictions prove to be highly accurate: all observed prediction errors do not exceed 3.68% for temperatures and 4.75% for pressures. In a cumulative energy estimation, we observe prediction errors that are less than 4.04%. A typical reservoir simulation for Brady Hot Springs completes in approximately 4 h, whereas our machine learning models yield accurate 20-year predictions for temperatures, pressures, and produced energy in 0.9 s. This paper aims to demonstrate how the models and techniques from our study can be applied to achieve rapid exploration of controlled parameters and optimization of other geothermal reservoirs.

15 GEOTHERMAL ENERGY↗
DOE OSTI2022

SCALE 6.2.4: Validation Overview

SCALE is a multi-application code system with tools for reactor physics, criticality safety, radiation shielding, and spent fuel characterization for a range of nuclear systems. The last of the SCALE 6.2 series, version 6.2.4, was released in June 2020. Validation studies in nuclear criticality safety, reactor physics, and shielding applications were performed to confirm results, assess accuracy, and identify biases from SCALE simulations. Nuclear criticality safety applications yield small code biases for a wide range of systems, generally overestimating cross section uncertainty. Reactor physics applications improvements in nuclide inventory and full-core eigenvalue predictions, with similar performance in decay heat predictions. Radiation shielding applications yield sufficient agreement in comparisons to measurements and alternate simulations for eight shielding experiments.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2023

Adaptation Strategies Strongly Reduce the Future Impacts of Climate Change on Simulated Crop Yields

Abstract Simulations of crop yield due to climate change vary widely between models, locations, species, management strategies, and Representative Concentration Pathways (RCPs). To understand how climate and adaptation affects yield change, we developed a meta‐model based on 8703 site‐level process‐model simulations of yield with different future adaptation strategies and climate scenarios for maize, rice, wheat and soybean. We tested 10 statistical models, including some machine learning models, to predict the percentage change in projected future yield relative to the baseline period (2000–2010) as a function of explanatory variables related to adaptation strategy and climate change. We used the best model to produce global maps of yield change for the RCP4.5 scenario and identify the most influential variables affecting yield change using Shapley additive explanations. For most locations, adaptation was the most influential factor determining the projected yield change for maize, rice and wheat. Without adaptation under RCP4.5, all crops are expected to experience average global yield losses of 6%–21%. Adaptation alleviates this average projected loss by 1–13 percentage points. Maize was most responsive to adaptive practices with a projected mean yield loss of −21% [range across locations: −63%, +3.7%] without adaptation and −7.5% [range: −46%, +13%] with adaptation. For maize and rice, irrigation method and cultivar choice were the adaptation types predicted to most prevent large yield losses, respectively. When adaptation practices are applied, some areas are predicted to experience yield gains, especially at northern high latitudes. These results reveal the critical importance of implementing adequate adaptation strategies to mitigate the impact of climate change on crop yields.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Forest carbon sequestration on the west coast, USA: Role of species, productivity, and stockability

Forest ecosystems store large amounts of carbon and can be important sources, or sinks, of the atmospheric carbon dioxide that is contributing to global warming. Understanding the carbon storage potential of different forests and their response to management and disturbance events are fundamental to developing policies and scenarios to partially offset greenhouse gas emissions. Projections of live tree carbon accumulation are handled differently in different models, with inconsistent results. We developed growth-and-yield style models to predict stand-level live tree carbon density as a function of stand age in all vegetation types of the coastal Pacific region, US (California, Oregon, and Washington), from 7,523 national forest inventory plots. We incorporated site productivity and stockability within the Chapman-Richards equation and tested whether intensively managed private forests behaved differently from less managed public forests. We found that the best models incorporated stockability in the equation term controlling stand carrying capacity, and site productivity in the equation terms controlling the growth rate and shape of the curve. RMSEs ranged from 10 to 137 Mg C/ha for different vegetation types. There was not a significant effect of ownership over the standard industrial rotation length (~50 yrs) for the productive Douglas-fir/western hemlock zone, indicating that differences in stockability and productivity captured much of the variation attributed to management intensity. Our models suggest that doubling the rotation length on these intensively managed lands from 35 to 70 years would result in 2.35 times more live tree carbon stored on the landscape. These findings are at odds with some studies that have projected higher carbon densities with stand age for the same vegetation types, and have not found an increase in yields (on an annual basis) with longer rotations. We suspect that differences are primarily due to the application of yield curves developed from fully-stocked, undisturbed, single-species, “normal” stands without accounting for the substantial proportion of forests that don’t meet those assumptions. The carbon accumulation curves developed here can be applied directly in growth-and-yield style projection models, and used to validate the predictions of ecophysiological, cohort, or single-tree style models being used to project carbon futures for forests in the region. Our approach may prove useful for developing robust models in other forest types.

Chisholm, Paul J. (ORCID:0000000238784707)↗
DOE OSTI2024

Tuning crosslinking of hybrid preceramic polymers in vat photopolymerization toward controlled ceramic yields

Control of preceramic polymer crosslinking for UV-curable processing is essential for fine 3D printing with high ceramic conversion for sustainable polymer-derived ceramics (PDC) engineering. While various factors influencing ceramic yield have been studied, the systematic exploration of the relationship between crosslinking and ceramic yield, especially when crosslinking increases volatile elements, remains open for further investigation. This study addresses this gap by utilizing vat photopolymerization (VP) additive manufacturing (AM) as a versatile platform for controlling preceramic crosslinking and ceramic yield. By rationally designing and tuning the photochemical crosslinking through digital light processing (DLP), we demonstrate that the ceramic yield can be enhanced from 64% to over 86%, even with added volatile elements. We reveal that the post-pyrolysis ceramic yield can be closely correlated with the pre-pyrolysis crosslinking of the preceramic network represented by its stiffness. This correlation thereby suggests a fast, energy-efficient, non-destructive methodology to predict and improve ceramic yield. Combined with these findings, our one-pot thiol-ene hybridization of polycarbosilane and polycarbosiloxane via DLP offers an exemplary method to generate hybrid preceramic polymers with tailored material properties toward target applications, and potentially even higher ceramic yields. This study thus contributes to achieving better resource- and energy-efficient preceramic polymer and PDC processing routes toward sustainable, advanced organic–inorganic materials manufacturing.

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

A cross-study analysis of drug response prediction in cancer cell lines

Abstract To enable personalized cancer treatment, machine learning models have been developed to predict drug response as a function of tumor and drug features. However, most algorithm development efforts have relied on cross-validation within a single study to assess model accuracy. While an essential first step, cross-validation within a biological data set typically provides an overly optimistic estimate of the prediction performance on independent test sets. To provide a more rigorous assessment of model generalizability between different studies, we use machine learning to analyze five publicly available cell line-based data sets: National Cancer Institute 60, ancer Therapeutics Response Portal (CTRP), Genomics of Drug Sensitivity in Cancer, Cancer Cell Line Encyclopedia and Genentech Cell Line Screening Initiative (gCSI). Based on observed experimental variability across studies, we explore estimates of prediction upper bounds. We report performance results of a variety of machine learning models, with a multitasking deep neural network achieving the best cross-study generalizability. By multiple measures, models trained on CTRP yield the most accurate predictions on the remaining testing data, and gCSI is the most predictable among the cell line data sets included in this study. With these experiments and further simulations on partial data, two lessons emerge: (1) differences in viability assays can limit model generalizability across studies and (2) drug diversity, more than tumor diversity, is crucial for raising model generalizability in preclinical screening.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Effects of HCP/BCC element ratios on the room-temperature tensile properties of Ti-Zr-Hf-Nb-Ta refractory high-entropy alloys

Equiatomic and non-equiatomic Ti-Zr-Hf-Nb-Ta refractory high-entropy alloys (RHEAs) were arc melted, homogenized, cold rolled, and recrystallized to produce single-phase, body-centered cubic (BCC), microstructures with weak texture and equiaxed grains 76–199 μm in size. Here, the non-equiatomic alloys had either a 60:40 or 80:20 atomic ratio of hexagonal close-packed (HCP) elements (Ti + Zr + Hf) to BCC elements (Nb + Ta). Alloy compositions were measured after thermomechanical processing to determine the concentrations of the major (substitutional) and minor (interstitial) elements. We investigated how elastic constants and uniaxial tensile properties were affected by changes in the relative concentrations of the constituent elements at fixed HCP:BCC ratios. Yield strengths ranged from 801 to 922 MPa and ultimate tensile strengths from 815 to 933 MPa. Good agreement is obtained between the experimental yield strengths and those predicted by a strength theory based on edge dislocations indicating that the observed compositional effects are due to their effects on shear modulus and volume misfit. Fracture occurred by dimpled rupture with fracture strains of 19.4%–25.7%, but uniform strains were an order of magnitude lower at 1.1%–3.2%, calling into question the useable ductility (prior to necking) of RHEAs considered to be ductile based on their fracture strain. Contrary to predictions in the literature that HCP elements promote ductility, our present results show that increasing the HCP:BCC ratio decreases both the total strain and the uniform strain. Similar trends were not evident in the yield or ultimate strengths; rather, strengths were affected mainly by shear modulus and volume misfit.

BCC high-entropy alloys↗