Search NASA⌕ Search

SEARCH · Search NASA

Results for “regression models”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 163 records · Page 9

Optimization of artificial viscosity in production codes based on Gaussian Regression surrogate models

To accurately model flows with shock waves using staggered-grid Lagrangian hydrodynamics, artificial viscosity has to be introduced to convert kinetic energy into internal energy, thereby increasing the entropy across shocks. Determining the appropriate strength of the artificial viscosity is an art and strongly depends on the particular problem and experience of the researcher. The objective of this study is to pose the problem of finding the appropriate strength of artificial viscosity as an optimization problem and solve this problem using machine learning (ML) tools, specifically using surrogate models based on Gaussian Process regression and Bayesian analysis. We describe the optimization method and discuss various practical details of its implementation. The shock-containing problems for which we apply this method all have been implemented in the LANL code FLAG. First, we apply ML to find optimal values to isolated shock problems of different strengths. Second, we apply ML to optimize viscosity for a 1D propagating detonation problem based on Zel’dovich-von Neumann-Doring (ZND) detonation theory using a reactive burn model. We compare results for default (currently used values in FLAG) and optimized values of artificial viscosity for these problems demonstrating the potential for significant improvement in the accuracy of computations.

42 ENGINEERING↗

Aerobic Fitness Does Not Contribute to Prediction of Orthostatic Intolerance

Several investigations have suggested that orthostatic tolerance may be inversely related to aerobic fitness (VO (sub 2max)). To test this hypothesis, 18 males (age 29 to 51 yr) underwent both treadmill VO(sub 2max) determination and graded lower body negative pressures (LBNP) exposure to tolerance. VO(2max) was measured during the last minute of a Bruce treadmill protocol. LBNP was terminated based on pre-syncopal symptoms and LBNP tolerance (peak LBNP) was expressed as the cumulative product of LBNP and time (torr-min). Changes in heart rate, stroke volume cardiac output, blood pressure and impedance rheographic indices of mid-thigh-leg initial accumulation were measured at rest and during the final minute of LBNP. For all 18 subjects, mean (plus or minus SE) fluid accumulation index and leg venous compliance index at peak LBNP were 139 plus or minus 3.9 plus or minus 0.4 ml-torr-min(exp -2) x 10(exp 3), respectively. Pearson product-moment correlations and step-wise linear regression were used to investigate relationships with peak LBNP. Variables associated with endurance training, such as VO(sub 2max) and percent body fat were not found to correlate significantly (P is less than 0.05) with peak LBNP and did not add sufficiently to the prediction of peak LBNP to be included in the step-wise regression model. The step-wise regression model included only fluid accumulation index leg venous compliance index, and blood volume and resulted in a squared multiple correlation coefficient of 0.978. These data do not support the hypothesis that orthostatic tolerance as measured by LBNP is lower in individuals with high aerobic fitness.

Convertino, Victor A.↗

Application of laser induced breakdown spectroscopy for total carbon quantification in soil samples

The increase of greenhouse gas (i.e., CO{sub 2}) levels in the atmosphere has caused noticeable climate change. Many nations are currently looking into methods of permanent underground storage for CO{sub 2} in an attempt to mitigate this problem. The goal of this work is to develop a process for studying the total carbon content in soils before, during, and after CO{sub 2} injection to ensure that no leakage is occurring or to determine how much is leaking if it is occurring and what effect it will have on the ecosystem between the injection formation and the atmosphere. In this study, we quantitatively determine the total carbon concentration in soil using laser-induced breakdown spectroscopy (LIBS). A soil sample from Starkville, Mississippi, USA was mixed with different amounts of carbon powder, which was used as a calibration for additional carbon in soil. Test samples were prepared by adding different but known amounts of carbon powder to a soil sample and then mixing with polyvinyl alcohol binder before being pressed into pellets. LIBS spectra of the test samples were collected and analyzed to obtain optimized conditions for the measurement of total carbon in soil with LIBS. The total carbon content in the samples was also measured by a carbon analyzer, and the data (average of triplicates) were used as a reference in developing calibration curves for a modified version of the single linear regression model and the multiple linear regression model. The calibration data were then used to determine the total carbon concentration of an unknown sample. This work is intended to be used in the initial development of a miniaturized, field-portable LIBS analyzer for CO{sub 2} leak detection.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

Associations between cadmium levels in blood and urine, blood pressure and hypertension among Canadian adults

Background: Cadmium has been inconsistently related to blood pressure and hypertension. The present study seeks to clarify the relationship between cadmium levels found in blood and urine, blood pressure and hypertension in a large sample of adults. Methods: The study sample included participants ages 20 through 79 from multiple cycles of the Canadian Health Measures Survey (2007 through 2013) with measured blood cadmium (n=10,099) and urinary cadmium (n=6988). Linear regression models examined the association between natural logarithm transformed cadmium levels and blood pressure (separate models for systolic and diastolic blood pressure) after controlling for known covariates. Logistic regression models were used to examine the association between cadmium and hypertension. Models were run separately by sex, smoking status, and body mass index category. Results: Men had higher mean systolic (114.8 vs. 110.8 mmHg, p<0.01) and diastolic (74.0 vs. 69.6 mmHg, p<0.01) blood pressure compared to women. Although, geometric mean blood (0.46 vs. 0.38 µg/L, p<0.01) and creatinine-adjusted standardized urinary cadmium levels (0.48 vs. 0.38 µg/L, p<0.01) were higher among those with hypertension, these differences were no longer significant after adjustment for age, sex and smoking status. In overall regression models, increases in blood cadmium were associated with increased systolic (0.70 mmHg, 95% confidence interval [CI]=0.25–1.16, p<0.01) and diastolic blood pressure (0.74 mmHg, 95% CI=0.30–1.19, p<0.01). The associations between urinary cadmium, blood pressure and hypertension were not significant in overall models. Model stratification revealed significant and negative associations between urinary cadmium and hypertension among current smokers (OR=0.61, 95% CI=0.44–0.85, p<0.01), particularly female current smokers (OR=0.52, 95% CI=0.32–0.85, p=0.01). Conclusion: This study provides evidence of a significant association between cadmium levels, blood pressure and hypertension. However, the significance and direction of this association differs by sex, smoking status, and body mass index category. - Highlights: • Blood and urinary cadmium levels higher among those with hypertension. • Evidence of association between cadmium levels, blood pressure and hypertension. • Significance and direction of association differs by sex, smoking status, and BMI. • Higher urinary cadmium levels lower hypertension risk for current (female) smokers.

54 ENVIRONMENTAL SCIENCES↗

Machine Learning–Augmented Laser-Induced Breakdown Spectroscopy for Spectral Discrimination of Iron Oxalates

Enhanced characterization and phase identification of post-PUREX Pu Oxalates (PuOXA) are pivotal for nonproliferation and pre-detonation nuclear forensics. Despite significant advances in the characterization of PuO 2 samples, little is known about the impact of both the chemical structure and oxidation states of PuOXA (i.e., Pu(III) and Pu(IV)) have on optical emission signatures. Here, we demonstrate the analytical capabilities of laser-induced breakdown spectroscopy (LIBS) applied to Fe(II) and Fe(III) oxalate samples as surrogates for PuOXA, highlighting the discriminating features in the LIBS emission spectra arising from differences in the oxidation states within mixed FeOXA samples. We report the enhancement of spectral feature selection using Principal Component Analysis (PCA), which enables the analytical superiority of machine learning algorithms such as Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA), Partial Least Squares Regression (PLSR), Support Vector Regression (SVR), and Random Forest Regression (RFR) over conventional univariate techniques for phase discrimination and chemometric analysis. Cluster analysis revealed how both matrix effects and laser ablation influence cluster separability by introducing spectral artifacts that misdirect the maximization of variance. PCA-selected emission lines were used in the regression models, demonstrating that both univariate and multivariate linear regression models (i.e., PLSR and SVR) can achieve acceptable performance, with machine learning models outperforming conventional calibration regressions. Furthermore, the application of non-linearly activated PCA-selected emission lines illustrates how simplifying the data while retaining captured variance enables the use of less complex and more computationally efficient models. Furthermore, this is particularly evident in the underperformance of RFR, which suffers from increased computational costs and overfitting owing to its high complexity.

Oxalates↗

Predicting lettuce canopy photosynthesis with statistical and neural network models

An artificial neural network (NN) and a statistical regression model were developed to predict canopy photosynthetic rates (Pn) for 'Waldman's Green' leaf lettuce (Latuca sativa L.). All data used to develop and test the models were collected for crop stands grown hydroponically and under controlled-environment conditions. In the NN and regression models, canopy Pn was predicted as a function of three independent variables: shootzone CO2 concentration (600 to 1500 micromoles mol-1), photosynthetic photon flux (PPF) (600 to 1100 micromoles m-2 s-1), and canopy age (10 to 20 days after planting). The models were used to determine the combinations of CO2 and PPF setpoints required each day to maintain maximum canopy Pn. The statistical model (a third-order polynomial) predicted Pn more accurately than the simple NN (a three-layer, fully connected net). Over an 11-day validation period, average percent difference between predicted and actual Pn was 12.3% and 24.6% for the statistical and NN models, respectively. Both models lost considerable accuracy when used to determine relatively long-range Pn predictions (> or = 6 days into the future).

Non-NASA Center↗

Multi-Variate LSTM Prediction of Alaska Magnetometer Chain Utilizing a Coupled Model Approach

During periods of rapidly changing geomagnetic conditions electric fields form within the Earth’s surface and induce currents known as geomagnetically induced currents(GICs), which interact with unprotected electrical systems our society relies on. In this study, we train multi-variate Long-Short Term Memory neural networks to predict magnitude of north-south component of the geomagnetic field (|BN|) at multiple ground magnetometer stations across Alaska provided by the SuperMAG database with a future goal of predicting geomagnetic field disturbances. Each neural network is driven by solar wind and interplanetary magnetic field inputs from the NASA OMNI database spanning from 2000–2015 and is fine tuned for each station to maximize the effectiveness in predicting |BN|. The neural networks are then compared against multivariate linear regression models driven with the same inputs at each station using Heidke skill scores with thresholds at the 50, 75, 85, and 99 percentiles for |BN|. The neural network models show significant increases over the linear regression models for |BN| thresholds. We also calculate the Heidke skill scores for d|BN|/dt by deriving d|BN|/dt from |BN| predictions. However, neural network models do not show clear outperformance compared to the linear regression models. To retain the sign information and thus predict BN instead of |BN|, a secondary so-called polarity model is utilized. The polarity model is run in tandem with the neural networks predicting geomagnetic field in a coupled model approach and results in a high correlation between predicted and observed values for all stations. We find this model a promising starting point for a machine learned geomagnetic field model to be expanded upon through increased output time history and fast turnaround times.

Matthew Blandin↗

Modeling and Analysis of The Pressure Die Casting Using Response Surface Methodology

Pressure die casting is successfully used in the manufacture of Aluminum alloys components for automobile and many other industries. Die casting is a process involving many process parameters having complex relationship with the quality of the cast product. Though various process parameters have influence on the quality of die cast component, major influence is seen by the die casting machine parameters and their proper settings. In the present work, non-linear regression models have been developed for making predictions and analyzing the effect of die casting machine parameters on the performance characteristics of die casting process. Design of Experiments (DOE) with Response Surface Methodology (RSM) has been used to analyze the effect of effect of input parameters and their interaction on the response and further used to develop nonlinear input-output relationships. Die casting machine parameters, namely, fast shot velocity, slow shot to fast shot change over point, intensification pressure and holding time have been considered as the input variables. The quality characteristics of the cast product were determined by porosity, hardness and surface rough roughness (output/responses). Design of experiments has been used to plan the experiments and analyze the impact of variables on the quality of casting. On the other-hand Response Surface Methodology (Central Composite Design) is utilized to develop non-linear input-output relationships (regression models). The developed regression models have been tested for their statistical adequacy through ANOVA test. The practical usefulness of these models has been tested with some test cases. These models can be used to make the predictions about different quality characteristics, for the known set of die casting machine parameters, without conducting the experiments.

36 MATERIALS SCIENCE↗

Field-scale investigation of pulverized coal mill power consumption

Twenty field-scale tests were conducted in a 28 MW pulverized coal power plant in Healy, Alaska, to examine mill power consumption in relation to coal grind size. The intent in this field-scale study was to verify if grind size truly impacted power consumption by a detectable amount. The regression model developed from the data indicates that grind size does impact mill power consumption, with finer grinds consuming significantly more power than coarser grinds. However, other factors such as coal hardness (i.e. the lower the Hardgrove Grindability Index, or the harder the coal, the higher the power consumption) and mill throughput (i.e., the higher the throughput, the higher the power consumption) had to be included before the impact of grind size could be isolated. It was also observed that combining amperage and flow rate into a single parameter, i.e., specific amperage, hurt modeling. Cost analysis based on the regression model indicate a power savings of $19,972 per year if the coal were ground to 50% passing 76 {mu}m rather than the industry standard of 70% passing 76 {mu}m. The study also demonstrated that size reduction constituted a significant portion of the power consumption.

01 COAL, LIGNITE, AND PEAT↗

Benefit Analysis of CO 2 Delivery Options for Offshore Storage or Enhanced Oil Recovery

The analysis presented in this report evaluates the benefits of CO₂ offshore transport via pipeline or ship within the GOM. It takes a top-down framework to estimate the costs. First, this analysis designed a reduced-order model (ROM) based on the cash flows in the FECM/NETL CO₂ Transport Cost Model (also known as CO2_T_COM). The ROM takes capital expenses (CAPEX) and operating expenses (OPEX) to calculate the CO₂ breakeven price based on the cash flows. Second, this analysis developed regression models utilizing published data from other analyses to estimate CAPEX and OPEX. Since the ROM is a simplified cash flow calculation, it is easy to exchange the core regression models to estimate various costs. The ROM and regression models provided a framework that can be easily used by other researchers, decision-makers, operators, and regulators. The objective of this analysis is to assess the CO₂ breakeven cost range for pipeline and ship transport of captured CO₂ given the CO₂ source and storage reservoir located in the GOM.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

Distribution Grid Modeling Using Smart Meter Data

The knowledge of distribution grid models, including topologies and line impedances, is essential for grid monitoring, control and protection. However, such information is often unavailable, incomplete or outdated. The increasing deployment of smart meters (SMs) provides a unique opportunity to tackle this issue. This paper proposes a two-stage framework for distribution grid modeling using SM data. In the first stage, the network topology is identified by reconstructing a weighted Laplacian matrix of distribution networks. In the second stage, a least absolute deviations (LAD) regression model is developed for estimating line impedance of a single branch based on the nonlinear (inverse) power flow model, wherein a conductor library is leveraged to narrow down the solution space. The LAD regression model is originally a mixed-integer nonlinear program whose continuous relaxation is still non-convex. Furthermore, we specially address its convex relaxation and discuss the exactness. The modified regression model is then embedded within a bottom-up sweep algorithm to achieve the identification across the network in a branch-wise manner. Numerical results on the IEEE 13-bus, 37-bus and 69-bus test feeders validate the effectiveness of the proposed methods.

24 POWER TRANSMISSION AND DISTRIBUTION↗

A machine learning approach for efficient multi-dimensional integration

Many physics problems involve integration in multi-dimensional space whose analytic solution is not available. The integrals can be evaluated using numerical integration methods, but it requires a large computational cost in some cases, so an efficient algorithm plays an important role in solving the physics problems. We propose a novel numerical multi-dimensional integration algorithm using machine learning (ML). After training a ML regression model to mimic a target integrand, the regression model is used to evaluate an approximation of the integral. Then, the difference between the approximation and the true answer is calculated to correct the bias in the approximation of the integral induced by ML prediction errors. Because of the bias correction, the final estimate of the integral is unbiased and has a statistically correct error estimation. Three ML models of multi-layer perceptron, gradient boosting decision tree, and Gaussian process regression algorithms are investigated. The performance of the proposed algorithm is demonstrated on six different families of integrands that typically appear in physics problems at various dimensions and integrand difficulties. The results show that, for the same total number of integrand evaluations, the new algorithm provides integral estimates with more than an order of magnitude smaller uncertainties than those of the VEGAS algorithm in most of the test cases.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Calibration and Data Analysis of the MC-130 Air Balance

Design, calibration, calibration analysis, and intended use of the MC-130 air balance are discussed. The MC-130 balance is an 8.0 inch diameter force balance that has two separate internal air flow systems and one external bellows system. The manual calibration of the balance consisted of a total of 1854 data points with both unpressurized and pressurized air flowing through the balance. A subset of 1160 data points was chosen for the calibration data analysis. The regression analysis of the subset was performed using two fundamentally different analysis approaches. First, the data analysis was performed using a recently developed extension of the Iterative Method. This approach fits gage outputs as a function of both applied balance loads and bellows pressures while still allowing the application of the iteration scheme that is used with the Iterative Method. Then, for comparison, the axial force was also analyzed using the Non-Iterative Method. This alternate approach directly fits loads as a function of measured gage outputs and bellows pressures and does not require a load iteration. The regression models used by both the extended Iterative and Non-Iterative Method were constructed such that they met a set of widely accepted statistical quality requirements. These requirements lead to reliable regression models and prevent overfitting of data because they ensure that no hidden near-linear dependencies between regression model terms exist and that only statistically significant terms are included. Finally, a comparison of the axial force residuals was performed. Overall, axial force estimates obtained from both methods show excellent agreement as the differences of the standard deviation of the axial force residuals are on the order of 0.001 % of the axial force capacity.

Booth, Dennis↗

Home Performance with ENERGY STAR: Utility Bill Analysis on Homes Participating in Austin Energy's Program

Home Performance with ENERGY STAR (HPwES) is a jointly managed program of the U.S. Department of Energy (DOE) and the U.S. Environmental Protection Agency (EPA). This program focuses on improving energy efficiency in existing homes via a whole-house approach to assessing and improving a home's energy performance, and helping to protect the environment. As a local sponsor for HPwES, Austin Energy's HPwES program offers a complete home energy assessment and a list of recommendations for efficiency improvements, along with cost estimates. The owner can choose to implement only one or the complete set of energy conservation measures. Austin Energy facilitates the process by providing economic incentives to the homeowner through its HPwES Loan program and its HPwES Rebate program. In 2005, the total number of participants in both programs was approximately 1,400. Both programs are only available for improvements made by a participating HPwES contractor. The individual household billing data - encompassing more than 7,000 households - provided by Austin Energy provides a rich data set to estimate the impacts of its HPwES program. The length of the billing histories is sufficient to develop PRISM-type models of electricity use based on several years of monthly bills before and after the installation of the conservation measures. Individual household savings were estimated from a restricted version of a PRISM-type regression model where the reference temperature to define cooling (or heating degree days) was estimated along with other parameters. Because the statistical quality of the regression models varies across individual households, three separate samples were used to measure the aggregate results. The samples were distinguished on the basis of the statistical significance of the estimated (normalized) cooling consumption. A normalized measure of cooling consumption was based on average temperatures observed over the most recent nine-year period ending in 2006. This study provided a statistically rigorous approach to incorporating the variability of expected savings across the households in the sample together with the uncertainty inherent in the regression models used to estimate those savings. While the impact of the regression errors was found to be relatively small in these particular samples, this approach may be useful in future studies using individual household billing data. The median percentage savings for the largest sample of 6,000 households in the analysis was 32%, while the mean savings was 28%. Because the number of households in the sample is very large, the standard error associated with the mean percentage savings are very small, less than 1%. A conservative statement of the average savings is that is falls in the range of 25% to 30% with a high level of certainty. This preliminary analysis provides robust estimates of average program savings, but offers no insight into how savings may vary by type of conservation measure or whether savings vary by the amount of cooling electricity used prior to undertaking the measure. Follow-up researchers may want to analyze the impacts of specific ECMs. Households that use electricity for heating might also be separately analyzed. In potential future work several methodological improvements could also be explored. As mentioned in Section 2, there was no formal attempt to clean the data set of outliers and other abnormal patterns of billing data prior to the statistical analysis. The restriction of a constant reference temperature might also be relaxed. This approach may provide evidence as to whether any 'take-back' efforts are present, whereby thermostat settings are lowered during the summer months after the measures are undertaken (reflected in lower reference temperatures in the post-ECM period). A more extended analysis may also justify the investment in and use of the PRISM software package, which may provide more diagnostic measures with respect to the reference temperature. PRISM also appears to contain some built-in capability to detect outliers and other anomalous data points.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Prediction of Creep-Induced Strain Using a Symbolic Regression-Based Model

Material creep under high-temperature conditions limits the lifetime and safety of structural systems such as advanced nuclear reactors. Conventional creep testing is slow and often produces inconsistent results across nominally identical experiments, making lifetime prediction uncertain. Here, to address these challenges, this work develops a data-driven symbolic regression (SR) model that consolidates results from duplicate creep tests and predicts the remaining strain-time curve of an ongoing experiment. The method uses piece-wise multi-objective SR with physical constraints to generate analytic, interpretable functions describing transient creep strain. Applied to Inconel Alloy 617 data, the approach achieved relative mean absolute errors of 1.0–9.5%, providing closed-form predictions of strain evolution. These results demonstrate a first step toward reducing the duration and cost of long-term creep testing while retaining physically interpretable model forms.

11 - NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Quantifying mean, variability, and uncertainty in indoor radon exposure in Pennsylvania using random forest and quantile regression forest models

Radon is a naturally occurring radioactive gas that poses a serious health risk as the primary cause of lung cancer in non-smokers. Despite the well-known adverse association with health outcomes, current radon exposure assessments are limited to county-level or average-level estimates, which fail to capture regional variability. This study uses Machine Learning models, including Random Forest (RF) and Quantile Regression Forest (QRF), to estimate the indoor radon concentrations at the ZCTA (Zip code tabulation area)-level and characterize uncertainties in model estimates. Incorporating geological, meteorological, and building-specific data, the models aim to improve radon risk assessment by capturing mean exposure, variability, and extreme concentration levels. Processed radon test data (n = 718,111) were analyzed using average, variability, and quantile prediction methods. Models that estimate the average radon exposure at the ZCTA-level can yield promising model-fit results, but they do not capture the underlying variability of indoor radon exposure within a ZCTA. We utilize volatility analyses to identify characteristics indicative of high variability of indoor radon exposure. We also show that a QRF model can be used to estimate upper quantiles of residential radon exposure, thereby uncovering localized areas of elevated exposure that were not apparent in mean estimates. The results highlighted the need for a deep characterization of exposure risk and show that regions with moderate average exposure levels could still harbor extreme outliers with implications for evaluating health risks. Utilizing multiple radon exposure models allows for a deeper characterization of radon risk within a geographic area and can better identify high-risk areas. The results from this study provide a foundation for developing mitigation strategies and examining associations between radon exposure and health outcomes at fine scales. Future research should extend the geographic scope and incorporate additional environmental risk factors to establish a comprehensive framework for risk assessment.

Lee, Heechan [ORNL]↗

Methodology for Constructing Reduced-Order Power Block Performance Models for CSP Applications: Preprint

The inherent variability of the solar resource presents a unique challenge for CSP systems. Incident solar irradiation can fluctuate widely over a short time scale, but plant performance must be assessed for long time periods. As a result, annual simulations with hourly (or sub-hourly) timesteps are the norm in CSP analysis. A highly detailed power cycle model provides accuracy but tends tosuffer from prohibitively long run-times; alternatively, simplified empirical models can run quickly but don?t always provide enough information, accuracy, or flexibility for the modeler. The ideal model for feasibility-level analysis incorporates both the detail and accuracy of a first-principle model with the low computational load of a regression model. The work presented in this paperproposes a methodology for organizing and extracting information from the performance output of a detailed model, then using it to develop a flexible reduced-order regression model in a systematic and structured way. A similar but less generalized approach for characterizing power cycle performance and a reduced-order modeling methodology for CFD analysis of heat transfer from electronic deviceshave been presented. This paper builds on these publications and the non-dimensional approach originally described.

14 SOLAR ENERGY↗

Fundamental microscopic properties as predictors of large-scale quantities of interest: Validation through grain boundary energy trends

Correlations between fundamental microscopic properties computable from first principles, which we term canonical properties, and complex large-scale quantities of interest (QoIs) provide an avenue to predictive materials discovery. Here, we propose that such correlations can be efficiently discovered through simulations utilizing approximate interatomic potentials (IPs), which serve as an ensemble of “synthetic materials”. As a proof of principle we build a regression model relating canonical properties to the symmetric tilt grain boundary (GB) energy curves in face-centered cubic crystals, characterized by the scaling factor in the universal lattice matching model of Runnels et al. (2016), which we take to be our QoI. Our analysis recovers known correlations of GB energy to other properties and discovers new ones. We also demonstrate, using available density functional theory (DFT) GB energy data, that the regression model constructed from IP data is consistent with DFT results, confirming the assumption that the IPs and DFT belong to same statistical pool and thereby validating the approach. Regression models constructed in this fashion can be used to predict large-scale QoIs based on first-principles data and provide a general method for training IPs for QoIs beyond the scope of first-principles calculations.

36 MATERIALS SCIENCE↗