Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “regression”

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 1,315 records · Page 73
All sourcesNASA NTRSDOE OSTI
DOE OSTI2019

Prediction of liver remnant regeneration after living donor liver transplantation using preoperative CT texture analysis

Purpose: To predict the rate of liver regeneration after living donor liver transplantation (LDLT) using pre-operative computed tomography (CT) texture analysis. Materials and methods: 112 living donors who performed right hepatectomy for LDLT were included retrospectively. We measured the volume of future remnant liver (FLR) on pre-operative CT and the volume of remnant liver (LR) on follow-up CT, taken at a median of 123 days after transplantation. The regeneration index (RI) was calculated using the following equation: [(V{sub LR}−V{sub FLR})/V{sub FLR}] × 100. Computerized texture analysis of the semi-automatically segmented FLR was performed. We used a stepwise, multivariable linear regression to assess associations of clinical features and texture parameters in relation to RI and to make the best-fit predictive model. Results: The mean RI was 110.7 ± 37.8%, highly variable ranging from 22.4% to 247.0%. Among texture parameters, volume of FLR, standard deviation, variance, and gray level co-occurrence matrices (GLCM) contrast were found to have significant correlations between RI. In multivariable analysis, smaller volume of FLR (ß − 0.17, 95% CI − 0.22 to − 0.13) and lower GLCM contrast (ß − 1.87, 95% CI − 3.64 to − 0.10) were associated with higher RI. The regression equation predicting RI was following: RI = 203.82 + 10.42 × pre-operative serum total bilirubin (mg/dL) − 0.17 × V{sub FLR} (cm{sup 3}) − 1.87 × GLCM contrast (× 100). Conclusion: Volume of FLR and GLCM contrast were independent predictors of RI, showing significant negative correlations. Pre-operative CT with texture analysis can be useful for predicting the rate of liver regeneration in living donor of liver transplantation.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2019

Early response evaluation of neoadjuvant therapy with PET/MRI to predict resectability in patients with adenocarcinoma of the esophagogastric junction

Study design and purpose: Positron emission tomography (PET)/magnetic resonance imaging (MRI) is a new modality that has showed promising results for various clinical indications. Currently, evaluation of neoadjuvant therapy (NT) among patients with adenocarcinoma of the esophagogastric junction has primarily been reserved for PET/computed tomography. Our aim was to evaluate if early response evaluation by PET/MRI is a feasible method to predict resectability. Methods and materials: Patients with untreated adenocarcinoma of the esophagogastric junction (Siewert types I/II) and fit for NT with no contraindications for PET/MRI were considered eligible. A baseline scan was performed prior to NT induction and an evaluation scan 3 weeks later. For histopathological response evaluation, the Mandard tumor regression grade score was applied. Response on PET/MRI was evaluated with Response Evaluation Criteria in Solid Tumors (RECIST 1.1), and change in ADC and SUV{sub max} values. Results: Twenty-eight patients were enrolled, and 22 completed both scans and proceeded to final analyses. Seventeen patients were found resectable versus five who were found unresectable. PET/MRI response evaluation had a sensitivity 94%, specificity 80%, and AUC = 0.95 when predicting resectability in patients with adenocarcinoma of the esophagogastric junction. No association with histopathological response (tumor regression grade) was found nor was RECIST correlated with resectability. Conclusion: Response evaluation using PET/MRI was a feasible method to predict resectability in patients with adenocarcinoma of the esophagogastric junction in this pilot study. However, larger studies are warranted to justify the use of the modality for this indication.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2019

Analyzing the post-contrast attenuation of the esophageal wall on routine contrast-enhanced MDCT examination can improve the diagnostic accuracy in response evaluation of the squamous cell esophageal carcinoma to neoadjuvant chemoradiotherapy in comparison with the esophageal wall thickness

Purpose: To evaluate the accuracy of the multidetector computed tomography (MDCT) in the response evaluation of the esophageal squamous cell carcinoma (ESCC) to neoadjuvant chemoradiotherapy (nCRT) by analyzing the thickness and post-contrast attenuation of the esophageal wall after the nCRT. Methods: Contrast-enhanced (CE)-MDCT examinations in portal venous phase of one hundred patients with locally advanced ESCC who received nCRT and underwent esophageal resection and histopathology assessment of tumor regression grade (TRG) were retrospectively analyzed by measuring the maximal thickness and mean density of the esophageal wall in the segment involved by tumor and visually searching for hyperdense foci within it. Diagnostic performance was evaluated using the ROC analysis. Results: Average attenuation of the esophageal wall had stronger diagnostic performance for predicting pathologic complete regression (pCR) (AUC = 0.994; p < 0.001) in relation to maximal esophageal wall thickness (AUC = 0.731; p < 0.001). Maximal esophageal wall thickness ≤ 9 mm and average attenuation of the esophageal wall ≤ 64 HU predicted pCR with the sensitivity, specificity, and overall accuracy of 62.5%, 77.9%, and 73%, and 96.9%, 98.5%, and 98%, respectively. Combination of both cutoff values enabled correct assessment of pCR with the 100% accuracy. Visual detection of the hyperdense focus within the esophageal wall predicted pCR with the sensitivity, specificity, and overall accuracy values of 100%, 94.1%, and 96%, respectively. Conclusion: Visual analysis and measurement of post-contrast attenuation of the esophageal wall after the nCRT can improve diagnostic accuracy of MDCT in the response evaluation of the ESCC to nCRT in comparison with measuring the esophageal wall thickness.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2019

Can MRI be used to diagnose histologic grade in T1a (< 4 cm) clear cell renal cell carcinomas?

Objective: To assess whether MRI can differentiate low-grade from high-grade T1a cc-RCC. Materials and methods: With IRB approval, 49 consecutive solid < 4 cm cc-RCC (low grade [Grade 1 or 2] N = 38, high grade [Grade 3] N = 11) with pre-operative MRI before nephrectomy were identified between 2013 and 2018. Tumor size, apparent diffusion coefficient (ADC) histogram analysis, enhancement wash-in and wash-out rates, and chemical shift signal intensity index (SI index) were assessed by a blinded radiologist. Subjectively, two blinded Radiologists also assessed for (1) microscopic fat, (2) homogeneity (5-point Likert scale), and (3) ADC signal (relative to renal cortex); discrepancies were resolved by consensus. Outcomes were studied using Chi square, multivariate analysis, logistic regression modeling, and ROC. Inter-observer agreement was assessed using Cohen’s kappa. Results: Tumor size was 24 ± 7 (13–39) mm with no association to grade (p = 0.45). Among quantitative features studied, corticomedullary phase wash-in index (p = 0.015), SI index (p = 0.137), and tenth-centile ADC (p = 0.049) were higher in low-grade tumors. 36.8% (14/38) low-grade tumors versus zero high-grade tumors demonstrated microscopic fat (p = 0.015; Kappa = 0.67). Microscopic fat was specific for low-grade disease (100.0% [71.5–100.0]) with low sensitivity (36.8% [21.8–54.6]). Other subjective features did not differ between groups (p > 0.05). A logistic regression model combining microscopic fat + wash-in index + tenth-centile-ADC yielded area under ROC curve 0.98 (Confidence Intervals 0.94–1.0) with sensitivity/specificity 87.5%/100%. Conclusion: The combination of microscopic fat, higher corticomedullary phase wash-in and higher tenth-centile ADC is highly accurate for diagnosis of low-grade disease among T1a clear cell RCC.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2019

A prospective feasibility study evaluating the role of multimodality imaging and liquid biopsy for response assessment in locally advanced rectal carcinoma

Purpose: Colorectal cancer is a commonly encountered disease that poses several diagnostic and therapeutic challenges. The inherent heterogeneity of tumor biology and propensity to relapse despite “curative” resection pose significant challenges with regard to response assessment. Although MR imaging already plays a key role in primary staging of patients with rectal carcinoma, its reliability in restaging after neoadjuvant therapy is debatable (Van der broek et al. in Dis Colon Rectum 60(3):274–283, 2017). Therefore, there is significant interest in developing additional methods which may improve diagnostic accuracy. This study aims to evaluate the role of multimodality imaging and liquid biopsy in therapeutic response assessment. Methods: Seventeen patients were enrolled into the study over a span of 24 months. All underwent hybrid PET-MRI and CT-perfusion (CT-P), prior to and following neoadjuvant therapy for locally advanced rectal carcinoma. Twelve of the 17 patients also underwent liquid biopsy, which consisted of blood sampling and analysis of circulating tumor cells (CTCs) and extracellular vesicles (EVs), including cell fragments and microparticles (MPs), using the Cell Search System (Menarini Silicon Biosystems). SUV, DWI, and ADC were calculated during PET-MRI, and several parameters were evaluated during CT-perfusion, including average perfusion, blood flow (BF), blood volume (BV), mean transit time (MTT), permeability-surface area product (PS), contrast extraction efficiency (E), and K-trans (K). Changes observed pre- and post-neoadjuvant therapy in each modality were compared to tumor response at histopathology using a modified Ryan tumor regression grading system. Results: Of the 17 patients included in the study, 14 were classified as non-responders, and 3 were classified as responders as determined by the modified Ryan Tumor Regression Grade (TRG) scoring system (Van der broek et al. in Dis Colon Rectum 60(3):274–283, 2017). When combined, blood markers and CT-P parameters (mean transit time (MTT), K-trans, and permeability-surface area product (PS)) produced the strongest models (p < 0.01). PET (SUV measurement) combined with CT-P-derived K-trans produced a marginally significant (p = 0.057) model for predicting response. MRI-derived ADC value did not provide a significant model for response prediction. Conclusion: A model of CT-P parameters plus liquid biopsy more accurately predicts tumor response than PET-MRI, CT-P alone, or liquid biopsy alone. These results suggest that in the evaluation of treatment response, liquid biopsy could provide additional information to functional imaging modalities such as CT-P and should therefore be explored further in a trial with larger sample size.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2014

Pauling bond strength, bond length and electron density distribution

A power law regression equation, = 1.46(<ρ(rc)>/r)-0.19, connecting the average experimental bond lengths, , with the average accumulation of the electron density at the bond critical point, <ρ(rc)>, between bonded metal M and oxygen atoms, determined at ambient conditions for oxide crystals, where r is the row number of the M atom, is similar to the regression equation R(M-O) = 1.39(ρ(rc)/r)-0.21 determined for three perovskite crystals for pressures as high as 80 GPa. The two equations are also comparable with those, = 1.43( /r)-0.21, determined for a large number of oxide crystals at ambient conditions and = 1.39( /r)-0.22, determined for geometry optimized hydroxyacid molecules, that connect the bond lengths to the average Pauling electrostatic bond strength, , for the M-O bonded interactions. On the basis of the correspondence between the two sets of equations connecting ρ(rc) and the Pauling bond strength s with bond length, it appears that Pauling’s simple definition of bond strength closely mimics the accumulation of the electron density between bonded pairs of atoms. The similarity of the expressions for the crystals and molecules is compelling evidence that the M-O bonded interactions for the crystals and molecules 2 containing the same bonded interactions are comparable. Similar expressions, connecting bond lengths and bond strength, have also been found to hold for fluoride, nitride and sulfide molecules and crystals. The Brown-Shannon bond valence, σ, power law expression σ = [R1/(R(M-O)]N that has found wide use in crystal chemistry, is shown to be connected to a more universal expression determined for oxides and the perovskites, <ρ(rc)> = r[(1.41)/ ]4.76, demonstrating that the bond valence for a bonded interaction is likewise closely connected to the accumulation of the electron density between the bonded atoms. Unlike the Brown-Shannon expression, it is universal in that it holds for the M-O bonded interactions for a relatively wide range of M atoms of the periodic table. The power law equation determined for the oxide crystals at ambient conditions is similar to the power law expression <ρ(rc)> = r[1.46/ ]5.26 determined for the perovskites at pressures as high as 80 GPa, indicating that the intrinsic connection between R(M-O) and ρ(rc) that holds at ambient conditions also holds, to a first approximation, at high pressures.

Environmental Molecular Sciences Laboratory↗
DOE OSTI2017

Evaluation of Visualization Using a 50/50 (Contrast Media/Glucose 5% Solution) Technique for Radioembolization as an Alternative to a Standard Sandwich Technique

BackgroundRadioembolization (RE) with {sup 90}yttrium ({sup 90}Y) resin microspheres generally employs a sandwich technique with separate sequential administration of contrast medium (CM), followed by vehicle (e.g., glucose 5% [G5] solution), then {sup 90}Y resin microspheres (in G5), then G5, and then CM again to avoid contact of CM and microspheres under fluoroscopic guidance. This study evaluates the visualization quality and safety of a modified sandwich technique with a 50/50-mixture of CM (Imeron 300) and G5 for administration of {sup 90}Y resin microspheres.Materials and MethodsA retrospective analysis of 81 RE procedures in patients with primary or secondary liver tumors was performed. The quality of angiographic visualization of the hepatic vessels was assessed before the first injection and immediately before the whole dose has been injected. Visualization and flow rate were graded on a 5-point scale: 1 = very good to 5 = not visible/no antegrade flow. Univariate logistic regression models and multiple linear regression models were used to evaluate the prognostic variables associated with visualization and flow scores.ResultsVisualization quality was inversely related to flow rate, the lower the flow rate the better the grade of the visualization. Visualization quality was also inversely related to body-mass-index (BMI). Performing RE with the 50/50-CM/G5 mixture resulted in a mean injection time for 1 GBq of 15 min. No clinically significant adverse events, including radiation-induced liver disease were reported.ConclusionRE with a 50/50-mixture of CM and G5 for administration of {sup 90}Y resin microspheres in a modified sandwich technique is a safe administration alternative and provides good visualization of hepatic vessels, which is inversely dependent on flow rate and BMI. Injection time was reduced compared with our experience with the standard sandwich technique.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2018

Radiation Dose and Risk Estimates of CT-Guided Percutaneous Liver Ablations and Factors Associated with Dose Reduction

PurposeTo determine the radiation dose associated with CT-guided percutaneous liver ablations and identify potential risk factors that result in higher radiation doses.Materials and MethodsBetween June 2011 and June 2015, 245 consecutive patients underwent 304 CT-guided liver ablation treatments. Patient demographics, tumor characteristics and procedural parameters were identified and analyzed. The peak skin dose and effective dose were assessed for each procedure. Excess relative risk related to radiation effects was calculated. A logistic regression model was prepared by means of stepwise logistic regression to identify variables predictive of increased radiation exposure.ResultsTumor ablations were performed with microwave (n = 220), radiofrequency (n = 74) or irreversible electroporation (IRE) (n = 10). The mean peak skin dose for ablations was 239.2 ± 136.4 mGy, and the mean effective dose was 36.6 ± 22.3 mSv. Of the patient and procedural parameters that were analyzed, increasing weight, use of intravenous contrast and/or hydrodissection during the procedure, together with treatment of multiple lesions in the same sitting were all associated with higher radiation exposure. The mean increase in the absolute risk of fatal malignancy from a single procedure was 0.18% (range 0.02–0.9%). No deterministic skin changes were identified in the patient cohort.ConclusionThe overall risk of stochastic and deterministic effects from radiation associated with CT-guided ablations is low compared with other inherent procedural complications. This study identifies several factors that are associated with higher radiation dose in percutaneous liver ablation procedures.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2019

Two-Year Mortality After Angioplasty of the Femoro-Popliteal Artery with Uncoated Balloons and Paclitaxel-Coated Balloons—A Pooled Analysis of Four Randomized Controlled Multicenter Trials

Purpose: In view of a recent meta-analysis reporting increased mortality following angioplasty with paclitaxel-coated devices in peripheral arteries, we performed a patient-level 2-year mortality analysis based on pooled original data of four randomized controlled trials (THUNDER, FEMPAC, PACIFIER and CONSEQUENT). Methods and Results: Clinical data of four randomized controlled trial were pooled to assess 2-year mortality following paclitaxel-coated balloon (PCB) angioplasty compared to angioplasty without paclitaxel (control group). A logistic regression model was applied to identify potential predictors of mortality. At two years, 13 of 185 (7.0%) patients had died in the control group and 16/184 (8.7%) in the PBC group, p = 0.55. Kaplan–Meier analysis revealed no significant difference from all-cause death at 2 years (log rank p = 0.54). Causes of death were well balanced between the groups with no pattern or trend in favour of any specific causes in the PBC group. Logistic regression revealed that treatment groups (controls or PBC) were not a predictor of 2-year mortality. The only predictor for mortality was patient age ≥ 75 years. The delivered paclitaxel doses per patient were not significantly different in patients that died and those who did not die during the 24-month follow-up (5.300 ± 4.224 μg vs. 6.248 ± 4.629 μg, p = 0.433). Conclusions: Based on original patient-level data of four pooled randomized controlled trials, we found no increase in 2-year mortality in patients treated with PCB compared to control patients treated with uncoated balloons. Causes of death were well balanced between PCB and control patients.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2016

The role of Atlantic Multi-decadal Oscillation in the global mean temperature variability

We simulated the global mean 1900–2015 warming by 42 Coupled Models Inter-comparison Project, phase 5 (CMIP5) climate models varies between 0.58 and 1.70 °C. The observed warming according to the NASA GISS temperature analysis is 0.95 °C with a 1200 km smoothing radius, or 0.86 °C with a 250 km smoothing radius. The projection of the future 2015–2100 global warming under a moderate increase of anthropogenic radiative forcing (RCP4.5 scenario) by individual models is between 0.7 and 2.3 °C. The CMIP5 climate models agree that the future climate will be warmer; however, there is little consensus as to how large the warming will be (reflected by an uncertainty of over a factor of three). Moreover, a parsimonious statistical regression model with just three explanatory variables [anthropogenic radiative forcing due to greenhouse gases and aerosols (GHGA), solar variability, and the Atlantic Multi-decadal Oscillation (AMO) index] accounts for over 95 % of the observed 1900–2015 temperature variance. This statistical regression model reproduces very accurately the past warming (0.96 °C compared to the observed 0.95 °C) and projects the future 2015–2100 warming to be around 0.95 °C (with the IPCC 2013 suggested RCP4.5 radiative forcing and an assumed cyclic AMO behavior). The AMO contribution to the 1970–2005 warming was between 0.13 and 0.20 °C (depending on which AMO index is used) compared to the GHGA contribution of 0.49–0.58 °C. During the twenty-first century AMO cycle the AMO contribution is projected to remain the same (0.13–0.20 °C), while the GHGA contribution is expected to decrease to 0.21–0.25 °C due to the levelling off of the GHGA radiative forcing that is assumed according to the RCP4.5 scenario. Therefore, the anthropogenic contribution and natural variability are expected to contribute about equally to the anticipated global warming during the second half of the twenty-first century for the RCP4.5 trajectory.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2016

Drivers and potential predictability of summer time North Atlantic polar front jet variability

The variability of the North Atlantic polar front jet stream is crucial in determining summer weather around the North Atlantic basin. Recent extreme summers in western Europe and North America have highlighted the need for greater understanding of this variability, in order to aid seasonal forecasting and mitigate societal, environmental and economic impacts. Here we find that simple linear regression and composite models based on a few predictable factors are able to explain up to 35 % of summertime jet stream speed and latitude variability from 1955 onwards. Sea surface temperature forcings impact predominantly on jet speed, whereas solar and cryospheric forcings appear to influence jet latitude. The cryospheric associations come from the previous autumn, suggesting the survival of an ice-induced signal through the winter season, whereas solar influences lead jet variability by a few years. Regression models covering the earlier part of the twentieth century are much less effective, presumably due to decreased availability of data, and increased uncertainty in observational reanalyses. Wavelet coherence analysis identifies that associations fluctuate over the study period but it is not clear whether this is just internal variability or genuine non-stationarity. Finally we identify areas for future research.

Meteorology & Atmospheric Sciences↗
DOE OSTI2018

Inferring the source of evaporated waters using stable H and O isotopes

Stable isotope ratios of H and O are widely used to identify the source of water, e.g., in aquifers, river runoff, soils, plant xylem, and plant-based beverages. In situations where the sampled water is partially evaporated, its isotope values will have evolved along an evaporation line (EL) in δ 2 H/δ 18 O space, and back-correction along the EL to its intersection with a meteoric water line (MWL) has been used to estimate the source water’s isotope ratios. Here, we review the theory underlying isotopic estimation of source water for evaporated samples (iSW E ). We note potential for bias from a commonly used regression-based approach for EL slope estimation and suggest that a model-based approach may be preferable if assumptions of the regression approach are not valid. We then introduce a mathematical framework that eliminates the need to explicitly estimate the EL–MWL intersection, simplifying iSW E analysis and facilitating more rigorous uncertainty estimation. We apply this approach to data from the US EPA's 2007 National Lakes Assessment. We find that data for most lakes are consistent with a water source similar to annual runoff, estimated from monthly precipitation and evaporation within the lake basin. Strong evidence for both summer- and winter-biased sources exists, however, with winter bias pervasive in most snow-prone regions. Furthermore, the new analytical framework should improve the rigor of iSW E in ecohydrology and related sciences, and our initial results from US lakes suggest that previous interpretations of lakes as unbiased isotope integrators may only be valid in certain climate regimes.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

Language model-accelerated deep symbolic optimization

Symbolic optimization methods have been used to solve varied challenging and relevant problems such as symbolic regression and neural architecture search. However, the current state of the art typically learns each problem from scratch and is unable to leverage pre-existing knowledge and datasets that are available for many applications. Here, inspired by the similarity between sequence representations learned in natural language processing and the formulation of symbolic optimization as a discrete sequence optimization problem, we propose language model-accelerated deep symbolic optimization (LA-DSO), a method that leverages language models to learn symbolic optimization solutions more efficiently. We demonstrate LA-DSO in two tasks: symbolic regression, which allows us to perform extensive experimentation due to its low computation requirements, and computational antibody optimization, which shows that our proposal accelerates learning in challenging real-world problems.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Estimation of the Surface Fluxes for Heat and Momentum in Unstable Conditions with Machine Learning and Similarity Approaches for the LAFE Data Set

Abstract Measurements of three flux towers operated during the land atmosphere feedback experiment (LAFE) are used to investigate relationships between surface fluxes and variables of the land–atmosphere system. We study these relations by means of two machine learning (ML) techniques: multilayer perceptrons (MLP) and extreme gradient boosting (XGB). We compare their flux derivation performance with Monin–Obukhov similarity theory (MOST) and a similarity relationship using the bulk Richardson number (BRN). The ML approaches outperform MOST and BRN. Best agreement with the observations is achieved for the friction velocity. For the sensible heat flux and even more so for the latent heat flux, MOST and BRN deviate from the observations while MLP and XGB yield more accurate predictions. Using MOST and BRN for latent heat flux, the root mean square errors (RMSE) are 107 Wm $$^{-2}$$ - 2 and 121 Wm $$^{-2}$$ - 2 , respectively, as well as the intercepts of the regression lines are $$\approx 110$$ ≈ 110 Wm $$^{-2}$$ - 2 . For the ML methods, the RMSEs reduce to 31 Wm $$^{-2}$$ - 2 for MLP and 33 Wm $$^{-2}$$ - 2 for XGB as well as the intercepts to just 4 Wm $$^{-2}$$ - 2 for MLP and $$-1$$ - 1 Wm $$^{-2}$$ - 2 for XGB with slopes of the regression lines close to 1, respectively. These results indicate significant deficiencies of MOST and BRN, particularly for the derivation of the latent heat flux. In fact, in contrast to the established theories, feature importance weighting demonstrates that the ML methods base their improved derivations on net radiation, the incoming and outgoing shortwave radiations, the air temperature gradient, and the available water contents, but not on the water vapor gradient. The results imply that further studies of surface fluxes and other turbulent variables with ML techniques provide great promise for deriving advanced flux parameterizations and their implementation in land–atmosphere system models.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2019

Data-enabled structure–property mappings for lanthanide-activated inorganic scintillators

The use of machine learning (ML) models toward development of validated structure–property relationships for two fundamental properties of activated inorganic scintillators for high energy radiation detection, namely the light yield (LY) and the decay time constant, is explored. The ML models are built on easily accessible proxies of materials—interchangeably referred to as features, descriptors or fingerprints—that are carefully selected on the basis of a physical understanding of the scintillation mechanism. Our study indicates that the developed physics-based ML models employing kernel ridge regression (KRR) and AdaBoost algorithm applied on top of a decision tree-based regression are able to “learn” the underlying design rules in a multi-dimensional feature space and thereby enable reasonably accurate predictions of the two target properties on unseen compounds (i.e., on a held-out test set). For instance, within a set of twenty-five cerium- or europium-doped scintillator materials, our analysis reveals a strong correlation between the average ionic part of the dielectric constant and the LY, irrespective of the specific chemistry of the compounds, indicating that the average ionic part of the dielectric constant is a particularly relevant descriptor toward prediction of the LY. Our results also demonstrate that, despite the use of small training datasets, the developed models are able to quickly distinguish high performing chemistries from those with relatively poor performance and therefore can play a crucial role in screening of new compounds with an attractive combination of targeted properties. Finally, the present study provides necessary motivation for future efforts involving ML models with relatively large training datasets, vast feature space explorations, and experimental design in search of promising novel scintillator chemistries.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Accelerated screening of functional atomic impurities in halide perovskites using high-throughput computations and machine learning

The pressing need for novel materials that can serve rising demands in solar cell and optoelectronic technologies makes the nexus of halide perovskites, high-throughput computations, and machine learning, very promising. Ever increasing amounts of data on the structure, fundamental properties, and device performance of halide perovskites provide opportunities for learning chemical rules and design principles that make these materials attractive, and applying them across wide chemical spaces. In this work, we show that impurity properties of halide perovskites computed using density functional theory (DFT) can be combined with machine learning (ML) to deliver predictive models and quick identification of optoelectronically active impurity atoms. Our computation lead to the largest reported dataset of the formation energies and charge transition levels of Pb-site impurities in methylammonium lead halide (MAPbX 3 ) perovskites. Descriptors are defined to uniquely represent any impurity atom in any MAPbX 3 compound and mapped to the computed impurity properties using regression techniques such as Gaussian process regression, neural networks, and random forests. We use the best optimized predictive models to make predictions for hundreds of impurities across 9 MAPbX 3 compounds and create lists of dominating impurities, that is, impurities that can shift the equilibrium Fermi level in the perovskite as determined by native point defects. Finally, this accelerated screening powered by computations and machine learning can guide the identification of problematic impurities that may cause undesired recombination of charge carriers, as well as impurities that can be deliberately introduced to tune the perovskite conductivity and resulting photovoltaic absorption.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Data-driven search for promising intercalating ions and layered materials for metal-ion batteries

The rise in demand for lithium-ion batteries has led to a large-scale search for electrode materials and intercalating ion species to meet the demands of next-generation energy technologies. Recent efforts largely focus on searching for cathodes that can accommodate large amounts of intercalating ions, but similar work on anodes is relatively limited. This study utilizes machine learning methods to find alternative two-dimensional (2D) materials and intercalating ions beyond Li for metal-ion batteries with high-power efficiencies. The approach first uses density functional theory (DFT) calculations to estimate the theoretical capacities and voltages of various metal ions on 2D materials. The DFT-generated data also provide insights into the local structural accommodation upon ion intercalation on various 2D materials. Significant changes to the lattice can result in irreversible changes to the bonding environments in the anode material, resulting in poor cycling stability. Next, this study develops a binding energy and structural accommodation-based classification model to screen anode materials for next-generation batteries. The classification model selects intercalating ions and 2D material pairs suitable for batteries based on the calculated voltage and volumetric changes in the 2D material upon intercalation. Finally, this study builds a regression model to accurately predict the binding energies of the various intercalating ions on 2D materials. The approach highlights the importance of different elemental and structural features for classification and regression tasks. In conclusion, the insights gained from this study on the role of involved features, such as electronegativities of the constituent ions and the presence of unfilled electronic levels, will help to streamline further studies towards the search for future layered battery materials.

36 MATERIALS SCIENCE↗
DOE OSTI2026

A Data-Driven Method for Modeling Creep-Fatigue Stress- Strain Behavior Using Neural ODEs

In this paper, we introduce a data-driven machine learning approach for modeling one-dimensional stress–strain behavior under cyclic loading, utilizing experimental data from the nickel-based Alloy 617. The study employs uniaxial creep–fatigue test data acquired under various loading histories and compares two distinct neural network-based ODE models. The first model, known as the black-box model, comprehensively describes the strain–stress relationship using a Neural ODE equation. To interpret this black-box model, we apply the Sparse Identification of Nonlinear Dynamical Systems (SINDy) technique, transforming the black-box model into an equation-based model using symbolic regression. The second model, the Neural flow rule model, incorporates Hooke’s Law for the linear elastic component, with the nonlinear part characterized by a Neural ODE. Both models are trained with experimental data to accurately reflect the observed stress–strain behavior. We conduct a detailed comparison with the standard Chaboche model, which includes three back stresses. Our results demonstrate that the neural network-based ODE models precisely capture the experimental creep–fatigue mechanical behavior, exceeding the standard Chaboche model’s accuracy. Furthermore, an interpretable model derived from the black-box neural ODE model through symbolic regression achieves accuracy comparable to the Chaboche model, enhancing its interpretability. The results highlight the potential of neural network-based ODE models to depict complex creep–fatigue behavior, eliminating the necessity for experts to define a specific, material-focused model form.

creep-fatigue↗