SEARCH · Search NASA

Results for “model accuracy”

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.

SourceResearch connectionFrom yearThrough year

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

At least 163 records · Page 9

All sources NASA NTRS DOE OSTI

DOE OSTI2021

AtomSets as a hierarchical transfer learning framework for small and large materials datasets

Abstract Predicting properties from a material’s composition or structure is of great interest for materials design. Deep learning has recently garnered considerable interest in materials predictive tasks with low model errors when dealing with large materials data. However, deep learning models suffer in the small data regime that is common in materials science. Here we develop the AtomSets framework, which utilizes universal compositional and structural descriptors extracted from pre-trained graph network deep learning models with standard multi-layer perceptrons to achieve consistently high model accuracy for both small compositional data (<400) and large structural data (>130,000). The AtomSets models show lower errors than the graph network models at small data limits and other non-deep-learning models at large data limits. They also transfer better in a simulated materials discovery process where the targeted materials have property values out of the training data limits. The models require minimal domain knowledge inputs and are free from feature engineering. The presented AtomSets model framework can potentially accelerate machine learning-assisted materials design and discovery with less data restriction.

Chen, Chi (ORCID:0000000180087043)↗

NASA NTRS2004

LISA Thermal Design

The Laser Interferometer Space Antenna (LISA) mission, a space based gravitational wave detector, uses laser metrology to measure distance fluctuations between proof masses aboard three spacecraft. The total acceleration disturbance to each proof mass is required to be below 3 x 10(exp -15) meters per second squared per the square root of Hertz. Optical path length variations on each optical bench must be kept below about 3 pm per the square root of Hertz. Noise due to spacecraft thermal distortions, temperature difference variations across proof mass housing, and other thermal effects are expected to be a significant contributors to these noise budgets. The LISA Integrated Modeling team developed a detailed thermal model that is currently being used to drive the design of LISA. Several new thermal analysis techniques are also being developed in order to achieve model accuracies to LISA levels. We present here an overview of the LISA thermal design and modeling efforts. The latest thermal results calculated using the current baseline design of LISA are also discussed.

Merkowitz, Stephen↗

NASA NTRS2009

Using Numerical Modeling to Simulate Space Capsule Ground Landings

Experimental work is being conducted at the National Aeronautics and Space Administration s (NASA) Langley Research Center (LaRC) to investigate ground landing capabilities of the Orion crew exploration vehicle (CEV). The Orion capsule is NASA s replacement for the Space Shuttle. The Orion capsule will service the International Space Station and be used for future space missions to the Moon and to Mars. To evaluate the feasibility of Orion ground landings, a series of capsule impact tests are being performed at the NASA Langley Landing and Impact Research Facility (LandIR). The experimental results derived at LandIR provide means to validate and calibrate nonlinear dynamic finite element models, which are also being developed during this study. Because of the high cost and time involvement intrinsic to full-scale testing, numerical simulations are favored over experimental work. Subsequent to a numerical model validated by actual test responses, impact simulations will be conducted to study multiple impact scenarios not practical to test. Twenty-one swing tests using the LandIR gantry were conducted during the June 07 through October 07 time period to evaluate the Orion s impact response. Results for two capsule initial pitch angles, 0deg and -15deg , along with their computer simulations using LS-DYNA are presented in this article. A soil-vehicle friction coefficient of 0.45 was determined by comparing the test stopping distance with computer simulations. In addition, soil modeling accuracy is presented by comparing vertical penetrometer impact tests with computer simulations for the soil model used during the swing tests.

Heymsfield, Ernie↗

NASA NTRS1999

Lunar Prospector Orbit Determination Uncertainties Using the High Resolution Lunar Gravity Models

The Lunar Prospector (LP) mission began on January 6, 1998, when the LP spacecraft was launched from Cape Canaveral, Florida. The objectives of the mission were to determine whether water ice exists at the lunar poles, generate a global compositional map of the lunar surface, detect lunar outgassing, and improve knowledge of the lunar magnetic and gravity fields. Orbit determination of LP performed at the Jet Propulsion Laboratory (JPL) is conducted as part of the principal science investigation of the lunar gravity field. This paper will describe the JPL effort in support of the LP Gravity Investigation. This support includes high precision orbit determination, gravity model validation, and data editing. A description of the mission and its trajectory will be provided first, followed by a discussion of the orbit determination estimation procedure and models. Accuracies will be examined in terms of orbit-to-orbit solution differences, as a function of oblateness model truncation, and inclination in the plane-of-sky. Long term predictions for several gravity fields will be compared to the reconstructed orbits to demonstrate the accuracy of the orbit determination and oblateness fields developed by the Principal Gravity Investigator.

Carranza, Eric↗

DOE OSTI2022

Predicting Elastic Constants of Refractory Complex Concentrated Alloys Using Machine Learning Approach

Refractory complex concentrated alloys (RCCAs) have drawn increasing attention recently owing to their balanced mechanical properties, including excellent creep resistance, ductility, and oxidation resistance. The mechanical and thermal properties of RCCAs are directly linked with the elastic constants. However, it is time consuming and expensive to obtain the elastic constants of RCCAs with conventional trial-and-error experiments. The elastic constants of RCCAs are predicted using a combination of density functional theory simulation data and machine learning (ML) algorithms in this study. The elastic constants of several RCCAs are predicted using the random forest regressor, gradient boosting regressor (GBR), and XGBoost regression models. Based on performance metrics R-squared, mean average error and root mean square error, the GBR model was found to be most promising in predicting the elastic constant of RCCAs among the three ML models. Additionally, GBR model accuracy was verified using the other four RHEAs dataset which was never seen by the GBR model, and reasonable agreements between ML prediction and available results were found. The present findings show that the GBR model can be used to predict the elastic constant of new RHEAs more accurately without performing any expensive computational and experimental work.

36 MATERIALS SCIENCE↗

DOE OSTI2024

Machine learning for the redox potential prediction of molecules in organic redox flow battery

Here, organic redox flow batteries (ORFB) are recognized as an innovative technology for the large-scale storage of renewable energy. The redox potential of organic redox-active molecules plays a vital role in their performance. Advanced screening techniques like high-throughput experiment and machine learning (ML) have significantly enhanced organic material performance and transformed the field of ORFB. However, the scarcity of experimental data poses a considerable challenge for ML model development in this domain. In our study, we developed lightweight graph-based Gaussian process regression (GPR) models with GPU-accelerated marginalized graph kernel and hybrid kernel to predict the redox potentials of organic redox-active molecules for ORFBs, specifically focusing on small datasets. To evaluate model accuracy, we created a new experimental database of organic redox-active molecules by the data from hundreds of published papers and assembled previous computational datasets. We also considered some key parameters, such as pH conditions and solvent type, to assess their impact on redox potential prediction. Our GPR model predicted redox potentials with high accuracy across all datasets using minimal training data. The study provides powerful tools for molecule screening and design and delivers valuable guidance on designing training datasets for costly experiments.

25 ENERGY STORAGE↗

NASA NTRS2023

Efficient Calibration of Expensive Computational Models

Accounting for uncertainty when calibrating expensive computational models is a common challenge faced by scientists and engineers. Often Bayesian techniques are adopted to estimate a probability density function over the model parameters given noisy empirical data. The methods used to perform this type of probabilistic calibration are computationally prohibitive in that they require a large number of evaluations of the expensive model. In these cases, surrogate modeling -- that is, using a fast-to-evaluate, lower fidelity stand-in for the original computational model -- may be the only option to alleviate this computational burden. However, the upfront cost of generating training data to build a surrogate model can itself be expensive. As such, it is important to be judicious when selecting training points at which the full-fidelity model is evaluated. Here, an active learning approach is proposed that enables efficient selection of training points using approximate samples of the calibrated parameter probability density function. In this way, the training points can be concentrated in regions where the calibration algorithm requires high model accuracy.

active learning↗

NASA NTRS2009

Identifying Optimal Temporal Scale for the Correlation of AOD and Ground Measurements of PM2.5 to Improve the Model Performance in a Real-time Air Quality Estimation System

Aerosol optical depth (AOD), an indirect estimate of particle matter using satellite observations, has shown great promise in improving estimates of PM 2.5 air quality surface. Currently, few studies have been conducted to explore the optimal way to apply AOD data to improve the model accuracy of PM 2.5 surface estimation in a real-time air quality system. We believe that two major aspects may be worthy of consideration in that area: 1) the approach to integrate satellite measurements with ground measurements in the pollution estimation, and 2) identification of an optimal temporal scale to calculate the correlation of AOD and ground measurements. This paper is focused on the second aspect on the identifying the optimal temporal scale to correlate AOD with PM2.5. Five following different temporal scales were chosen to evaluate their impact on the model performance: 1) within the last 3 days, 2) within the last 10 days, 3) within the last 30 days, 4) within the last 90 days, and 5) the time period with the highest correlation in a year. The model performance is evaluated for its accuracy, bias, and errors based on the following selected statistics: the Mean Bias, the Normalized Mean Bias, the Root Mean Square Error, Normalized Mean Error, and the Index of Agreement. This research shows that the model with the temporal scale of within the last 30 days displays the best model performance in this study area using 2004 and 2005 data sets.

Li, Hui↗

DOE OSTI2020

Improvements to CTF Closure Models for Modelling of Two-Phase Flow

This report documents the efforts to improve the CTF prediction of void fraction and two-phase pressure drop by improving two-phase closure models. Previous validation activities have revealed that CTF tends to overpredict void fraction and two-phase pressure drop. In response to this, two approaches were taken to improve CTF’s predictive capabilities. Based on findings that the interfacial drag and subcooled boiling models significantly impact void prediction, alternative closure models for these physical effects were found and implemented into the code. An extensive assessment was performed by using the existing and newly added two-phase experimental data, which showed that void prediction, wall temperature, and two-phase pressure-drop results are improved by using the newly implemented models. A second approach for improving CTF modeling accuracy involved the use of a Bayesian calibration process that uses CTF validation data to optimize selected modeling coefficients and closure model multipliers to achieve a more accurate prediction of experimental results. A similar assessment was performed with calibrated models that improved the void and pressure-drop prediction for test cases that were both included and not included in the calibration dataset. Results of this study will be used to change the models used in CTF to achieve more accurate BWR analyses moving forward. This study also identified opportunities for improving additional closure models and other opportunities to use calibration techniques to improve CTF.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

DOE OSTI2024

Streamlining Ocean Dynamics Modeling with Fourier Neural Operators: A Multiobjective Hyperparameter and Architecture Optimization Approach

Training an effective deep learning model to learn ocean processes involves careful choices of various hyperparameters. We leverage DeepHyper’s advanced search algorithms for multiobjective optimization, streamlining the development of neural networks tailored for ocean modeling. The focus is on optimizing Fourier neural operators (FNOs), a data-driven model capable of simulating complex ocean behaviors. Selecting the correct model and tuning the hyperparameters are challenging tasks, requiring much effort to ensure model accuracy. DeepHyper allows efficient exploration of hyperparameters associated with data preprocessing, FNO architecture-related hyperparameters, and various model training strategies. We aim to obtain an optimal set of hyperparameters leading to the most performant model. Moreover, on top of the commonly used mean squared error for model training, we propose adopting the negative anomaly correlation coefficient as the additional loss term to improve model performance and investigate the potential trade-off between the two terms. The numerical experiments show that the optimal set of hyperparameters enhanced model performance in single timestepping forecasting and greatly exceeded the baseline configuration in the autoregressive rollout for long-horizon forecasting up to 30 days. Utilizing DeepHyper, we demonstrate an approach to enhance the use of FNO in ocean dynamics forecasting, offering a scalable solution with improved precision.

97 MATHEMATICS AND COMPUTING↗

NASA NTRS1989

Gravitational model improvement at the Goddard Space Flight Center

Major new computations of terrestrial gravitational field models were performed by the Geodynamics Branch of Goddard Space Flight Center (GSFC). This development has incorporated the present state of the art results in satellite geodesy and have relied upon a more consistent set of reference constants than was heretofore utilized in GSFC's GEM models. The solutions are complete in spherical harmonic coefficients out to degree 50 for the gravity field parameters. These models include adjustment for a subset of 66 ocean tidal coefficients for the long wavelength components of 12 major ocean tides. This tidal adjustment was made in the presence of 550 other fixed ocean tidal terms representing 32 major and minor ocean tides and the Wahr frequency dependent solid earth tidal model. In addition 5-day averaged values for Earth rotation and polar motion were derived for the time period of 1980 onward. Two types of models were computed. These are satellite only models relying exclusively on tracking data and combination models which have incorporated satellite altimetry and surface gravity data. The satellite observational data base consists of over 1100 orbital arcs of data on 31 satellites. A large percentage of these observations were provided by third generation laser stations (less than 5 cm). A calibration of the model accuracy of the GEM-T2 satellite only solution indicated that it was a significant improvement over previous models based solely upon tracking data. The rms geoid error for this field is 110 cm to degree and order 36. This is a major advancement over GEM-T1 whose errors were estimated to be 160 cm. An error propagation using the covariances of the GEM-T2 model for the TOPEX radial orbit component indicates that the rms radial errors are expected to be 12 cm. The combination solution, PGS-3337, is a preliminary effort leading to the development of GEM-T3. PGS-3337 has incorporated global sets of surface gravity data and the Seasat altimetry to produce a model complete to (50,50). A solution for the dynamic ocean topography to degree and order 10 was included as part of this adjustment.

Marsh, J. G.↗

NASA NTRS2006

Metrics for Linear Kinematic Features in Sea Ice

The treatment of leads as cracks or discontinuities (see Coon et al. presentation) requires some shift in the procedure of evaluation and comparison of lead-resolving models and their validation against observations. Common metrics used to evaluate ice model skills are by and large an adaptation of a least square "metric" adopted from operational numerical weather prediction data assimilation systems and are most appropriate for continuous fields and Eilerian systems where the observations and predictions are commensurate. However, this class of metrics suffers from some flaws in areas of sharp gradients and discontinuities (e.g., leads) and when Lagrangian treatments are more natural. After a brief review of these metrics and their performance in areas of sharp gradients, we present two new metrics specifically designed to measure model accuracy in representing linear features (e.g., leads). The indices developed circumvent the requirement that both the observations and model variables be commensurate (i.e., measured with the same units) by considering the frequencies of the features of interest/importance. We illustrate the metrics by scoring several hypothetical "simulated" discontinuity fields against the lead interpreted from RGPS observations.

Levy, G.↗

DOE OSTI2023

Integrating AlphaFold and deep learning for atomistic interpretation of cryo-EM maps

Abstract Interpretation of cryo-electron microscopy (cryo-EM) maps requires building and fitting 3D atomic models of biological molecules. AlphaFold-predicted models generate initial 3D coordinates; however, model inaccuracy and conformational heterogeneity often necessitate labor-intensive manual model building and fitting into cryo-EM maps. In this work, we designed a protein model-building workflow, which combines a deep-learning cryo-EM map feature enhancement tool, CryoFEM (Cryo-EM Feature Enhancement Model) and AlphaFold. A benchmark test using 36 cryo-EM maps shows that CryoFEM achieves state-of-the-art performance in optimizing the Fourier Shell Correlations between the maps and the ground truth models. Furthermore, in a subset of 17 datasets where the initial AlphaFold predictions are less accurate, the workflow significantly improves their model accuracy. Our work demonstrates that the integration of modern deep learning image enhancement and AlphaFold may lead to automated model building and fitting for the atomistic interpretation of cryo-EM maps.

59 BASIC BIOLOGICAL SCIENCES↗

NASA NTRS2018

Greedy Sampling and Incremental Surrogate Model-Based Tailoring of Aeroservoelastic Model Database for Flexible Aircraft

This paper presents a data analysis and modeling framework to tailor and develop linear parameter-varying (LPV) aeroservoelastic (ASE) model database for flexible aircrafts in broad 2D flight parameter space. The Kriging surrogate model is constructed using ASE models at a fraction of grid points within the original model database, and then the ASE model at any flight condition can be obtained simply through surrogate model interpolation. The greedy sampling algorithm is developed to select the next sample point that carries the worst relative error between the surrogate model prediction and the benchmark model in the frequency domain among all input-output channels. The process is iterated to incrementally improve surrogate model accuracy till a pre-determined tolerance or iteration budget is met. The methodology is applied to the ASE model database of a flexible aircraft currently being tested at NASA/AFRC for flutter suppression and gust load alleviation. Our studies indicate that the proposed method can reduce the number of models in the original database by 67%. Even so the ASE models obtained through Kriging interpolation match the model in the original database constructed directly from the physics-based tool with the worst relative error far below 1%. The interpolated ASE model exhibits continuously-varying gains along a set of prescribed flight conditions. More importantly, the selected grid points are distributed non-uniformly in the parameter space, a) capturing the distinctly different dynamic behavior and its dependence on flight parameters, and b) reiterating the need and utility for adaptive space sampling techniques for ASE model database compaction. The present framework is directly extendible to high-dimensional flight parameter space, and can be used to guide the ASE model development, model order reduction, robust control synthesis and novel vehicle design of flexible aircraft.

numerical analysi↗

DOE OSTI2021

Application of Artificial Neural Network Model for Optimized Control of Condenser Water Temperature Set-Point in a Chilled Water System

Here, in this study, real-time predictive control and optimization model based on an ANN (artificial neural network) was developed to evaluate the cooling energy saving performance of the optimized control of CndWT (condenser water temperature). For this purpose, the difference in TCEC (total cooling energy consumption) between the conventional control strategy when the CndWT produced by the cooling tower is fixed and the optimized control strategy when real-time control of the CndWT through the optimal ANN model is applied was compared and analyzed. For the modeling of the building to be simulated, the co-simulation of EnergyPlus and MATLAB was built through the middleware Building Controls Virtual Test Bed. For the prediction of TCEC, an ANN model was developed through MATLAB's neural network toolbox. The model accuracy of the ANN was examined through Cv(RMSE) index and as a result, Cv(RMSE) of the optimized ANN model turned out to be approximately 25 %. More importantly, the predictive control technique was able to save TCEC by 5.6 % compared to the conventional control method constantly fixing CndWT set-point to 30 °C. These results showed that the CndWT needs to be dynamically controlled using artificial intelligence technique such as ANN model and that significant energy savings were achievable compared to the conventional fixed control.

42 ENGINEERING↗

DOE OSTI2022

GCoD: Graph Convolutional Network Acceleration via Dedicated Algorithm and Accelerator Co-Design

Graph Convolutional Networks (GCNs) have emerged as the state-of-the-art graph learning model. However, it remains notoriously challenging to inference GCNs over large graph datasets, limiting their application to large real-world graphs and hindering the exploration of deeper and more sophisticated GCN graphs. This is because real-world graphs can be extremely large and sparse. Furthermore, the node degree of GCNs tends to follow the power-law distribution and therefore have highly irregular adjacency matrices, resulting in prohibitive inefficiencies in both data processing and movement and thus substantially limiting the achievable GCN acceleration efficiency. To this end, this paper proposes the first GCN algorithm and accelerator Co-Design framework dubbed GCoD which can largely alleviate the aforementioned GCN irregularity and boost GCNs' inference efficiency. Specifically, on the algorithm level, GCoD integrates a divide and conquer GCN training strategy that polarizes the graphs to be either denser or sparser in local neighborhoods without compromising the model accuracy, resulting in graph adjacency matrices that (mostly) have merely two levels of workload and enjoys largely enhanced regularity and thus ease of acceleration. On the hardware level, we further develop a dedicated two-pronged accelerator with a separated engine to process each of the aforementioned workloads, further boosting the overall utilization and acceleration efficiency. Extensive experiments and ablation studies validate that our GCoD consistently outperforms state-of-the-art designs in terms of accelerator efficiency while maintaining or even improving the task accuracy. Additionally, we visualize GCoD trained graph adjacency matrices to better understand its advantages. All codes and pre-trained models will be released upon acceptance.

You, Haoran↗

DOE OSTI2025

Model Residuals as Shields: A Two-Level Formulation to Defend Smart Grids From Poisoning Attacks

The advancement of smart grids presents both vast opportunities and heightened cybersecurity risks. Data-driven defense mechanisms, though designed as a shield against these threats, can fall prey to poisoning attacks. We delve into regression settings, underscoring the imperative to fortify defenses against a spectrum of poison ratios, notably those above 0.5—an issue scarcely addressed in prior studies. Recognizing the susceptibilities of smart grids and their manipulable sensors, we exploit the very intent of poisoning attacks, compromising model accuracy, as our defense mechanism. Our proposed two-level optimization framework discerns between poisoned and authentic data based on model residuals, outperforming or matching existing methods in 72% to 77% of precision and 75% to 80% of recalls across various poisoning attacks, poison ratios, and datasets. Once the authentic data are identified, the trained model is adaptable for a variety of applications. Comprehensive evaluations on different smart grid datasets, pitted against myriad poisoning schemes, validate our methodology’s edge over existing methods. Here, we also shed light on the implications of model misspecification originating from temporal auto-correlation, a common feature in Internet of Things and smart grid data.

Adversarial machine learning (ML)↗

DOE OSTI2023

Multiobjective Hyperparameter Optimization for Deep Learning Interatomic Potential Training Using NSGA-II

Deep neural network (DNN) potentials are an emerging tool for simulation of dynamical atomistic systems, with the promise of quantum mechanical accuracy at speedups of 10000$\times$. As with other DNN methods, hyperparameters used during training can make a substantial difference in model accuracy, and optimal settings vary with dataset. To enable rapid tuning of hyperparameters for DNN potential training, we developed a scalable multiobjective optimization evolutionary algorithm for supercomputers and tested it on the Summit system at the Oak Ridge Leadership Computing Facility (OLCF). The multiobjective approach is required due to the coupling of two learned values defining the potential: the energy and force. Using a large-scale implementation of the NSGA-II algorithm adapted for training DNN potentials, we discovered several optimal multiobjective combinations, including best choices of activation functions, learning rate scaling scheme, and pairing of the two radial cutoffs used in the three dimensional descriptor function.

Coletti, Mark↗