Search NASA⌕ Search

SEARCH · Search NASA

Results for “Network 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 1,531 records · Page 85

Efficient surrogate modeling methods for large-scale Earth system models based on machine-learning techniques

Abstract. Improving predictive understanding of Earth system variability and change requires data–model integration. Efficient data–model integration for complex models requires surrogate modeling to reduce model evaluation time. However, building a surrogate of a large-scale Earth system model (ESM) with many output variables is computationally intensive because it involves a large number of expensive ESM simulations. In this effort, we propose an efficient surrogate method capable of using a few ESM runs to build an accurate and fast-to-evaluate surrogate system of model outputs over large spatial and temporal domains. We first use singular value decomposition to reduce the output dimensions and then use Bayesian optimization techniques to generate an accurate neural network surrogate model based on limited ESM simulation samples. Our machine-learning-based surrogate methods can build and evaluate a large surrogate system of many variables quickly. Thus, whenever the quantities of interest change, such as a different objective function, a new site, and a longer simulation time, we can simply extract the information of interest from the surrogate system without rebuilding new surrogates, which significantly reduces computational efforts. We apply the proposed method to a regional ecosystem model to approximate the relationship between eight model parameters and 42 660 carbon flux outputs. Results indicate that using only 20 model simulations, we can build an accurate surrogate system of the 42 660 variables, wherein the consistency between the surrogate prediction and actual model simulation is 0.93 and the mean squared error is 0.02. This highly accurate and fast-to-evaluate surrogate system will greatly enhance the computational efficiency of data–model integration to improve predictions and advance our understanding of the Earth system.

58 GEOSCIENCES↗

Thermodynamic Control of Activity Patterns in Cytoskeletal Networks

Biological materials, such as the actin cytoskeleton, exhibit remarkable structural adaptability to various external stimuli by consuming different amounts of energy. In this Letter, we use methods from large deviation theory to identify a thermodynamic control principle for structural transitions in a model cytoskeletal network. Specifically, we demonstrate that biasing the dynamics with respect to the work done by nonequilibrium components effectively renormalizes the interaction strength between such components, which can eventually result in a morphological transition. Further, our work demonstrates how a thermodynamic quantity can be used to renormalize effective interactions, which in turn can tune structure in a predictable manner, suggesting a thermodynamic principle for the control of cytoskeletal structure and dynamics.

59 BASIC BIOLOGICAL SCIENCES↗

Projection of Future Fire Emissions Over the Contiguous US Using Explainable Artificial Intelligence and CMIP6 Models

Abstract Increasing temperature and water cycle changes due to warming climate may increase the frequency and intensity of wildfires. Fire emission projections are useful for informing strategies for adaptation and mitigation of fire impacts on societies and ecosystems. Here, we construct a neural network (NN) model explained by the Shapley Additive explanation to predict fire PM 2.5 emissions change and understand their drivers over the contiguous US (CONUS) in the mid‐21st century under a high greenhouse gas emissions scenario (SSP5‐8.5). Using future meteorology and leaf area index (LAI) simulated by eight global climate models from the Coupled Model Intercomparison Project Phase 6, future population density, and present‐day land use and land cover (LULC) as input to the NN model, the total fire PM 2.5 emissions over CONUS are projected to increase by 4%–75% (model spread). Among different regions, fire emissions in the western US are projected to increase more significantly in June‐July‐August than in other seasons and regions, with the median ratios of future to present‐day fire emissions ranging from 1.67 to 2.86. The increases in fire emissions are mainly driven by increasing normalized temperature (23%–29%) and decreasing soil moisture (2%–10%) in the future. When future LULC change is considered, the projected fire emissions further increase by 58%–83% over the western US compared to projections without LULC change because of future increases in vegetation fraction. The results highlight the important role of warmer temperature, decreasing soil moisture, and LULC change in increasing fire emissions in the future.

54 ENVIRONMENTAL SCIENCES↗

Impurity gas monitoring using ultrasonic sensing and neural networks: forward and inverse problems

Ultrasonic sensing is a non-invasive technique for monitoring impurity gas composition in various industrial applications where safety and regulatory compliance are crucial. In this study, ultrasonic sensing and neural networks were used to analyze impurity gases (i.e., air and argon) in helium. An experimental platform was established to acquire ultrasonic data. In the forward problem, an artificial neural network (ANN) model was used to forecast the response and time-of-flight (TOF) based on the excitation, and argon and air concentrations. The inverse problem was solved using a convolutional neural network (CNN) to predict the argon and air concentrations given the ultrasonic response and excitation. The results showed that the ANN accurately predicted the ultrasonic response and the change in TOF with concentration. As the air concentration was increased from 0 to 9.8%, the TOF sensitivity to detect argon decreased by 39.8% and 16.1% from ANN and sound speed theory, respectively. The CNN demonstrated high accuracy in predicting concentrations for inputs in the testing dataset. The application of the trained CNN indicated that it over-predicts air concentration while under-predicting the argon concentration. To improve accuracy, the predicted air and argon concentrations should be corrected by -0.992% and 1.027% bias, respectively.

47 OTHER INSTRUMENTATION↗

Dominant Controls on Preferential Flow and Their Implications for Future Soil Water Fluxes

Abstract Soil water flow, particularly preferential flow (PF), is a critical control on hydrological and biogeochemical processes, including groundwater recharge, contaminant transport, and carbon cycling. However, it remains challenging to predict PF occurrence across large environmental gradients. Here, we developed a deep learning (DL) model to estimate event‐scale soil water flow velocity and the probability of PF occurrence using high‐frequency soil moisture and precipitation data from 33 sites across the National Ecological Observatory Network. The model demonstrated high skill in predicting the binary occurrence of PF (91% F1‐score; 85% accuracy) but the performance was limited in predicting soil water velocity ( R 2 = 0.31). We found that precipitation characteristics (duration, volume, and intensity) were the most important predictors for soil water velocity. Among the non‐precipitation event variables, sand content showed relatively high predictive skill, though differences among non‐event climate variables were generally modest. Lower sand content was associated with increased predicted soil water velocity, a finding that highlights the role of soil structure in producing more non‐uniform flow, which contrasts with traditional uniform flow models. Projecting a reduced DL model under both moderate and high‐emissions future climate scenarios (2060–2099 Representative Concentration Pathways 4.5 and 8.5), we found ∼7.3% increase under RCP4.5 and ∼15% under RCP8.5 of soil water velocities compared to the historical simulation, while modeled likelihood of PF changed little. These findings suggest climate change is not making PF more frequent, but it is making existing PF pathways more efficient with important consequences for associated nutrient and contaminant transport under climate change. Plain Language Summary Water movement in soil is critical for water quality. While often modeled as a uniform flow process, in reality water moves rapidly through cracks and burrows in what is called “preferential flow” (PF), which limits natural filtration and can transport pollutants. We developed a deep learning model, trained on data from 33 U.S. sites, to predict when and how fast this PF occurs based on precipitation, soil, and climate data. The model showed that precipitation characteristics (duration, intensity, volume) were the most important predictors of PF. Lower soil sand content/higher clay content was associated with faster water flow, likely due to clay soils forming aggregates and cracks that water moves through rather than infiltrating uniformly. Further analyses based on climate projections suggest that the speed at which PF occurs will become more rapid under future climate scenarios compared to historical simulation. This highlights the need to represent PF in soil water models when assessing future water quality. Key Points The effect of precipitation peak intensity on soil water velocities declined with increasing precipitation intensity Antecedent soil moisture failed to predict preferential flow (PF), contrasting the high predictive power of sand content Climate predictions suggest that soil water velocities through PF paths will increase ∼15% by 2099

Li, Bonan↗

Demand Response for Thermal Fairness in District Heating Networks

District Heating (DH) networks are complex thermal grids wherein a centrally heated fluid is circulated through a network of pipes and heat exchangers to meet the heating needs of residential and commercial buildings. Several factors can lead to inefficient and unfair energy distribution among consumers in these networks. These include varying levels of building insulation, distance of individual buildings from the central energy source, and thermal losses in network pipes. Moreover, shortage of energy at the central energy source and extreme weather conditions can exacerbate these issues, leading to differing levels of thermal comfort and customer disgruntlement in the long run. In this paper, we propose and study a demand response scheme that attempts to ensure thermal fairness among energy consumers in modern thermal grids. We develop optimization formulations based on thermodynamic models of DH networks, which determine optimal energy flow for individual buildings in order to achieve thermal fairness across the network. Our experimental results using physics based models for DH networks show that it is possible to achieve network level thermal fairness based social welfare objectives by controlling network parameters such as mass flow rates of water to the consumer premises and the supply water temperature, even when some network parameters like building thermal characteristics remain unknown.

Bhattacharya, Saptarshi↗

A viscoelastic model for seismic attenuation using fractal mechanical networks

SUMMARY Seismic attenuation (quantified by the quality factor Q) has a significant impact on the seismic waveforms, especially in the fluid-saturated rocks. This dissipative process can be phenomenologically represented by viscoelastic models. Previous seismological studies show that the Q value of Earth media exhibits a nearly frequency-independent behaviour (often referred to as constant-Q in literature) in the seismic frequency range. Such attenuation can be described by the mathematical Kjartansson constant-Q model, which lacks of a physical representation in the viscoelastic sense. Inspired by the fractal nature of the pore fluid distribution in patchy-saturated rocks, here we propose two fractal mechanical network (FMN) models, that is, a fractal tree model and a quasi-fractal ladder model, to phenomenologically represent the frequency-independent Q behaviour. As with the classic viscoelastic models, the FMN models are composed of mechanical elements (spring and dashpots) arranged in different hierarchical patterns. A particular parametrization of each model can produce the same complex modulus as in the Kjartansson model, which leads to the constant-Q. Applying the theory to several typical rock samples, we find that the seismic attenuation signature of these rocks can be accurately represented by either one of the FMN models. Besides, we demonstrate that the ladder model in particular exhibits the realistic multiscale fractal structure of the saturated rocks. Therefore, the FMN models as a proxy could provide a new way to estimate the microscopic rock structure property from macroscopic seismic attenuation observation.

Geochemistry & Geophysics↗

Computational Analysis of Interactions of the Oral Microbiota

The human oral microbiome is a complex site containing many hundreds of species with the community response of this microbiome having a large effect on human health. The complexity of this site has spoken to the need for applying –omics techniques to better understand which microbial species are present, their contribution to the microbiome and how this may change as a function of disease. The kinds of datasets that emerge from these –omics studies require computational analysis to gain robust results. New research has also been focused on applying these methods to build computational networks and models that better describe the myriad of interactions between microbial species of the oral cavity and how this changes as a result of infection. In this work, I describe the state of the field at this point, what kinds of computational networks have been inferred and what they show. I also describe where the field is heading in the next few years and how more in depth –omics analysis can lead to more detailed models of interactions at this site and improve human health.

-omics↗

A comprehensive strategy for stable, high productivity cultivation of microalgae with controllable biomass composition

This project was performed primarily as a partnership between three Universities – The University of Toledo (UToledo), Montana State University (MSU) and The University of North Carolina at Chapel Hill (UNC). North Carolina State University also contributed to the project through a sub-subcontract. The project developed fundamental and applied knowhow for cultivation systems that are not restricted by (i) proximate availability of flue gas or other high concentration CO 2 sources and (ii) the energy and infrastructure burden to deliver CO 2 over long distances. To overcome these constraints, this project focused on cultivation of alkaliphilic algal cultures in high pH, high alkalinity media (pH>10, alkalinity >100mEq) and demonstrated high biomass productivity, improvement of biomass composition for fuel production, and culture stability. The high media alkalinity ensured availability of sufficient HCO 3 - in solution for high rates of carbon fixation while the high pH of the media facilitated rapid scavenging of atmospheric CO 2 . Overall, the direct air capture rates (DAC) in our system were sufficient to achieve high biomass productivities using atmospheric CO 2 alone. The project also developed and validated toolkits related to metabolic flux modeling and microbiome characterization.

09 BIOMASS FUELS↗

Mitigation of spatial nonstationarity with vision transformers

Spatial nonstationarity, the location variance of features’ statistical distributions, is ubiquitous in many natural settings. For example, in geological reservoirs rock matrix porosity varies vertically due to geomechanical compaction trends, in mineral deposits grades vary due to sedimentation and concentration processes, in hydrology rainfall varies due to the atmosphere and topography interactions, and in metallurgy crystalline structures vary due to differential cooling. Conventional geostatistical modeling workflows rely on the assumption of stationarity to be able to model spatial features for geostatistical inference. Nevertheless, this is often not a realistic assumption when dealing with nonstationary spatial data and this has motivated a variety of nonstationary spatial modeling workflows such as trend and residual decomposition, cosimulation with secondary features, and spatial segmentation and independent modeling over stationary subdomains. The advent of deep learning technologies has enabled new workflows for modeling spatial relationships. However, there is a paucity of demonstrated best practice and general guidance on mitigation of spatial nonstationarity with deep learning in the geospatial context. We demonstrate the impact of two common types of geostatistical spatial nonstationarity on deep learning model prediction performance and propose the mitigation of such impacts using self-attention (vision transformer) models. We demonstrate the utility of vision transformers for the mitigation of nonstationarity with relative errors as low as 10%, exceeding the performance of alternative deep learning methods such as convolutional neural networks. We establish best practice by demonstrating the ability of self-attention networks for modeling large-scale spatial relationships in the presence of commonly observed geospatial nonstationarity.

58 GEOSCIENCES↗

SU-E-J-163: A Biomechanical Lung Model for Respiratory Motion Study

Purpose: This work presents a biomechanical model to investigate the complex respiratory motion for the lung tumor tracking in radiosurgery by computer simulation. Methods: The models include networked massspring-dampers to describe the tumor motion, different types of surrogate signals, and the force generated by the diaphragm. Each mass-springdamper has the same mechanical structure and each model can have different numbers of mass-spring-dampers. Both linear and nonlinear stiffness parameters were considered, and the damping ratio was tuned in a range so that the tumor motion was over-damped (no natural tumor oscillation occurs without force from the diaphragm). The simulation was run by using ODE45 (ordinary differential equations by Runge-Kutta method) in MATLAB, and all time courses of motions and inputs (force) were generated and compared. Results: The curvature of the motion time courses around their peaks was sensitive to the damping ratio. Therefore, the damping ratio can be determined based on the clinical data of a high sampling rate. The peak values of different signals and the time the peaks occurred were compared, and it was found that the diaphragm force had a time lead over the tumor motion, and the lead time (0.1–0.4 seconds) depended on the distance between the tumor and the diaphragm. Conclusion: We reported a model based analysis approach for the spatial and temporal relation between the motion of the lung tumor and the surrogate signals. Due to the phase lead of the diaphragm in comparing with the lung tumor motion, the measurement of diaphragm motion (or its electromyography signal) can be used as a beam gating signal in radiosurgery, and it can also be an additional surrogate signal for better tumor motion tracking. The research is funded by the American Cancer Society (ACS) grant. The grant name is: Frameless SRS Based on Robotic Head Motion Cancellation. The grant number is: RSG-13-313-01-CCE.

60 APPLIED LIFE SCIENCES↗

Data-Driven Learning of Nonautonomous Systems

In this work, we present a numerical framework for recovering unknown nonautonomous dynamical systems with time-dependent inputs. To circumvent the difficulty presented by the nonautonomous nature of the system, our method transforms the solution state into piecewise integration of the system over a discrete set of time instances. The time-dependent inputs are then locally parameterized by using a proper model, for example, polynomial regression, in the pieces determined by the time instances. This transforms the original system into a piecewise parametric system that is locally time invariant. We then design a deep neural network structure to learn the local models. Once the network model is constructed, it can be iteratively used over time to conduct global system prediction. We provide theoretical analysis of our algorithm and present a number of numerical examples to demonstrate the effectiveness of the method.

97 MATHEMATICS AND COMPUTING↗

Data-Driven Learning of Non-Autonomous Systems

We present a numerical framework for recovering unknown non-autonomous dynamical systems with time-dependent inputs. To circumvent the difficulty presented by the non-autonomous nature of the system, our method transforms the solution state into piecewise integration of the system over a discrete set of time instances. The time-dependent inputs are then locally parameterized by using a proper model, for example, polynomial regression, in the pieces determined by the time instances. This transforms the original system into a piecewise parametric system that is locally time invariant. We then design a deep neural network structure to learn the local models. Once the network model is constructed, it can be iteratively used over time to conduct global system prediction. We provide theoretical analysis of our algorithm and present a number of numerical examples to demonstrate the effectiveness of the method.

97 MATHEMATICS AND COMPUTING↗

Physics-informed machine-learning model of temperature evolution under solid phase processes

We model temperature dynamics during Shear Assisted Proccess Extrusion (ShAPE), a solid phase process that plasticizes feedstock with a rotating tool and subsequently extrudes it into a consolidated tube, rod, or wire. Control of temperature is critical during ShAPE processing to avoid liquefaction, ensure smooth extrusion, and develop desired material properties in the extruded products. Accurate modeling of the complicated thermo-mechanical feedbacks between process inputs, material temperature, and heat generation presents a significant barrier to predictive modeling and process design. In particular, connecting micro-structural scale mechanisms of heat generation to macro-scale predictions of temperature can become computationally intractable. In this work we use a neural network (NN) model of heat generation to bridge this gap, by combining it with a simplified model of the temperature dynamics due to conduction and convection to capture the macro scale evolution of temperature. We inform the construction of the NN heat generation model using crystal plasticity simulations at the micro-structural scale to model the effects of process inputs on generation of heat. We achieved close fits of the temperature dynamics model to a diverse experimental data-set. Further, the relationships learned by the NN model between process inputs and heat generation showed qualitative agreement with those predicted by crystal plasticity simulations.

36 MATERIALS SCIENCE↗

Predicting Student Success using Analytics in Course Learning Management Systems

Educational data analytics is an emerging discipline, concerned with developing methods for exploring the unique types of data that come from the educational context. For example, predicting college student performance is crucial for both the student and educational institutions. It can support timely intervention to prevent students from failing a course, increasing efficacy of advising functions, and improving course completion rate. In this paper, we present the efforts carried out at Oak Ridge National Laboratory (ORNL) toward conducting predictive analytics to academic data collected from 2009 through 2013 and available in one of the most commonly used learning management systems, called Moodle. First, we have identified the data features useful for predicting student outcomes such as students scores in homework assignments, quizzes, exams, in addition to their activities in discussion forums and their total GPA at the same term they enrolled in the course. Then, Logistic Regression and Neural Network predictive models are used to identify students as early as possible that are in danger of failing the course they are currently enrolled in. These models compute the likelihood of any given student failing (or passing) the current course. Numerical results are presented to evaluate and compare the performance of the developed models and their predictive accuracy.

Educational data mining↗

Enhanced Boundary Layer Height Detection Using Ceilometer, Surface Meteorology, and Radiation Products With a Random Forest Ensemble Method

This study develops and evaluates a Random Forest (RF) model for estimating planetary boundary layer height (PBLH) using 9 years of data from the Atmospheric Radiation Measurement Southern Great Plains (ARM SGP) user facility, with potential application in the NOAA Surface Radiation (SURFRAD) Network. The model integrates ceilometer, surface meteorology, and radiation measurements, and is trained using thermodynamic PBLH estimates derived from radiosondes. This approach aims to bridge gaps between aerosol-based and thermodynamic-based PBLH estimates. The RF model outperformed traditional methods during daytime and better captured transition periods, demonstrating improved accuracy and robustness. At ARM SGP, it showed a substantial reduction in both bias and RMSE, with a bias near zero (−4.9 m) compared with traditional Haar Wavelet (HW) (70.9 m) and Vaisala BL-View software (124.1 m), and an RMSE of 303.2 m, lower than both BL-View (566.9 m) and HW (404.6 m). During daytime hours, RF consistently outperformed both alternatives, maintaining lower bias and RMSE across all periods. At a second evaluation site, RF achieved the lowest overall RMSE (323.7 m), similar to HW (326.4 m) and significantly better than BL-View (738.3 m). However, all models showed reduced accuracy under stable nighttime conditions, limiting the reliability of PBLH estimates. Key predictors for the model included the lifting condensation level height (LCLH), aerosol gradients, and month for seasonal variability. The study underscores the potential of integrating machine learning with multiple data sets such as surface energy and thermodynamic data to advance PBLH estimation.

boundary layer height↗

Surrogate-driven design optimization with uncertainty constraints in Monte Carlo simulations

In multi-objective design tasks, the computational cost increases rapidly when high-fidelity simulations are used to evaluate objective functions. Surrogate models help mitigate this cost by approximating the simulation output, simplifying the design process. However, under high uncertainty, surrogate models trained on noisy data can produce inaccurate predictions, as their performance depends heavily on the quality of training data. This study investigates the impact of data uncertainty on two multi-objective design problems modelled using Monte Carlo transport simulations: a neutron moderator and an ion-to-neutron converter. For each, a grid search was performed using five different tally uncertainty levels to generate training data for neural network surrogate models. These models were then optimized using NSGA-III. The recovered Pareto-fronts were analyzed across uncertainty levels: in the moderator problem, normalized hypervolume dropped from 0.886 at 1.0% uncertainty to 0.748 at 10% uncertainty, while in the converter problem it remained near 0.50 for all cases. Average simulation times were also compared to evaluate the trade-off between accuracy and computational cost. Results show that the influence of simulation uncertainty is strongly problem-dependent. In the neutron moderator case, higher uncertainties led to exaggerated objective sensitivities and distorted Pareto-fronts, reducing normalized hypervolume. In contrast, the ion-to-neutron converter task was less affected—low-fidelity simulations produced results similar to those from high-fidelity data. These findings suggest that a fixed-fidelity approach is not optimal. Surrogate models can recover the Pareto-front under noisy conditions, and multi-fidelity studies help identify suitable uncertainty levels for each problem to balance efficiency and accuracy.

07 ISOTOPE AND RADIATION SOURCES↗

Physics-Informed Machine Learning for Epidemiological Models

One challenge of using compartmental SEIR models for public health planning is the difficulty in manually tuning parameters to capture behavior reflected in the real-world data. This team conducted initial, exploratory analysis of a novel technique to use physics-informed machine learning tools to rapidly develop data-driven models for physical systems. This machine learning approach may be used to perform data assimilation of compartment models which account for unknown interactions between geospatial domains (i.e. diffusion processes coupling across neighborhoods/counties/states/etc.). Results presented here are early, proof-of-concept ideas that demonstrate initial success in using a physically informed neural network (PINN) model to assimilate data in a compartmental epidemiology model. The results demonstrate initial success and warrant further research and development.

60 APPLIED LIFE SCIENCES↗