SEARCH · Engineering Papers
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.
A new learning algorithm termed cascade error projection (CEP) is presented. CEP is an adaption of a constructive architecture from cascade correlation and the dynamical stepsize of A/D conversion from the cascade back propagation algorithm.
Deep neural networks have been successfully employed in an extensive variety of research areas, including solving partial differential equations. Despite its significant success, there are some challenges in effectively training DNN, such as avoiding overfitting in over-parameterized DNNs and accelerating the optimization in DNNs with pathological curvature. Here, we propose a Bayesian type sparse deep learning algorithm. The algorithm utilizes a set of spike-and-slab priors for the parameters in the deep neural network. The hierarchical Bayesian mixture will be trained using an adaptive empirical method. That is, one will alternatively sample from the posterior using preconditioned stochastic gradient Langevin Dynamics (PSGLD), and optimize the latent variables via stochastic approximation. The sparsity of the network is achieved while optimizing the hyperparameters with adaptive searching and penalizing. A popular SG-MCMC approach is Stochastic gradient Langevin dynamics (SGLD). However, considering the complex geometry in the model parameter space in nonconvex learning, updating parameters using a universal step size in each component as in SGLD may cause slow mixing. To address this issue, we apply a computationally manageable preconditioner in the updating rule, which provides a step-size parameter to adapt to local geometric properties. Moreover, by smoothly optimizing the hyperparameter in the preconditioning matrix, our proposed algorithm ensures a decreasing bias, which is introduced by ignoring the correction term in the preconditioned SGLD. According to the existing theoretical framework, we show that the proposed algorithm can asymptotically converge to the correct distribution with a controllable bias under mild conditions. Numerical tests are performed on both synthetic regression problems and learning solutions of elliptic PDE, which demonstrate the accuracy and efficiency of the present work.
We study the temporal co-variation of network co-evolution via the cross-link structure of networks, for which we take advantage of the formalism of hypergraphs to map cross-link structures back to network nodes. We investigate two sets of temporal network data in detail. In a network of coupled nonlinear oscillators, hyperedges that consist of network edges with temporally co-varying weights uncover the driving co-evolution patterns of edge weight dynamics both within and between oscillator communities. In the human brain, networks that represent temporal changes in brain activity during learning exhibit early co-evolution that then settles down with practice. Subsequent decreases in hyperedge size are consistent with emergence of an autonomous subgraph whose dynamics no longer depends on other parts of the network. Our results on real and synthetic networks give a poignant demonstration of the ability of cross-link structure to uncover unexpected co-evolution attributes in both real and synthetic dynamical systems. This, in turn, illustrates the utility of analyzing cross-links for investigating the structure of temporal networks.
This report details the development of a machine learning (ML)-driven framework to coarse-grain inter-particle collision dynamics from high-fidelity Discrete Element Method (DEM) simulations to Particle-in-Cell (PIC) scales for gas-solid systems. Traditional PIC models, while computationally efficient, rely on empirical granular stress formulations that fail to capture the full complexity of collision physics, particularly the heterogeneity in particle dynamics. This study adopts a bottom-up approach, integrating insights from DEM simulations to improve the physical fidelity and interpretability of PIC-scale models.
In the headwater catchments of the Rocky Mountains, plant productivity and its dynamics are largely dependent upon water availability, which is influenced by changing snowmelt dynamics associated with climate change. Understanding and quantifying the interactions between snow, plants and soil moisture is challenging, since these interactions are highly heterogeneous in mountainous terrain, particularly as they are influenced by microtopography within a hillslope. Recent advances in satellite remote sensing have created an opportunity for monitoring snow and plant dynamics at high spatiotemporal resolutions that can capture microtopographic effects. In this study, we investigate the relationships among topography, snowmelt, soil moisture and plant dynamics in the East River watershed, Crested Butte, Colorado, based on a time series of 3-meter resolution PlanetScope normalized difference vegetation index (NDVI) images. To make use of a large volume of high-resolution time-lapse images (17 images total), we use unsupervised machine learning methods to reduce the dimensionality of the time lapse images by identifying spatial zones that have characteristic NDVI time series. We hypothesize that each zone represents a set of similar snowmelt and plant dynamics that differ from other identified zones and that these zones are associated with key topographic features, plant species and soil moisture. We compare different distance measures (Ward and complete linkage) to understand the effects of their influence on the zonation map. Results show that the identified zones are associated with particular microtopographic features; highly productive zones are associated with low slopes and high topographic wetness index, in contrast with zones of low productivity, which are associated with high slopes and low topographic wetness index. The zones also correspond to particular plant species distributions; higher forb coverage is associated with zones characterized by higher peak productivity combined with rapid senescence in low moisture conditions, while higher sagebrush coverage is associated with low productivity and similar senescence patterns between high and low moisture conditions. In addition, soil moisture probe and sensor data confirm that each zone has a unique soil moisture distribution. This cluster-based analysis can tractably analyze high-resolution time-lapse images to examine plant-soil-snow interactions, guide sampling and sensor placements and identify areas likely vulnerable to ecological change in the future.
This article proposes a new traffic signal control algorithm to deal with unknown-traffic-system uncertainties and reduce delays in vehicle travel time. Unknown-traffic-system dynamics are approximated using a recurrent neural network (NN). To accurately identify the traffic system model, an online-learning scheme is developed to switch among a set of candidate NNs (i.e., multiple-model NNs) based on their estimation errors. Then, a bank of optimal signal-timing controllers is designed based on the online identification of the traffic system. Simulation studies have been carried out for the obtained control strategies using multiple-model NNs, and the desired results have been obtained. Moreover, compared with the widely used actuated traffic signal control schemes, it is shown that the proposed method can reduce vehicle travel delays and improve traffic system robustness.
This data package is associated with the publication “Prediction of Distributed River Sediment Respiration Rates using Community-Generated Data and Machine Learning’’ submitted to the Journal of Geophysical Research: Machine Learning and Computation (Scheibe et al. 2024). River sediment respiration observations are expensive and labor intensive to obtain and there is no physical model for predicting this quantity. The Worldwide Hydrobiogeochemisty Observation Network for Dynamic River Systems (WHONDRS) observational data set (Goldman et al.; 2020) is used to train machine learning (ML) models to predict respiration rates at unsampled sites. This repository archives training data, ML models, predictions, and model evaluation results for the purposes of reproducibility of the results in the associated manuscript and community reuse of the ML models trained in this project. One of the key challenges in this work was to find an optimum configuration for machine learning models to work with this feature-rich (i.e. 100+ possible input variables) data set. Here, we used a two-tiered approach to managing the analysis of this complex data set: 1) a stacked ensemble of ML models that can automatically optimize hyperparameters to accelerate the process of model selection and tuning and 2) feature permutation importance to iteratively select the most important features (i.e. inputs) to the ML models. The major elements of this ML workflow are modular, portable, open, and cloud-based, thus making this implementation a potential template for other applications. This data package is associated with the GitHub repository found at Please see the file level metadata (flmd; “sl-archive-whondrs_flmd.csv”) for a list of all files contained in this data package and descriptions for each. Please see the data dictionary (dd; “sl-archive-whondrs_dd.csv”) for a list of all column headers contained within comma separated value (csv) files in this data package and descriptions for each. The GitHub repository is organized into five top-level directories: (1) “input_data” holds the training data for the ML models; (2) “ml_models” holds machine learning models trained on the data in “input_data”; (3) “scripts” contains data preprocessing and postprocessing scripts and intermediate results specific to this data set that bookend the ML workflow; (4) “examples” contains the visualization of the results in this repository including plotting scripts for the manuscript (e.g., model evaluation, FPI results) and scripts for running predictions with the ML models (i.e., reusing the trained ML models); (5) “output_data” holds the overall results of the ML model on that branch. Each trained ML model resides on its own branch in the repository; this means that inputs and outputs can be different branch-to-branch. Furthermore, depending on the number of features used to train the ML models, the preprocessing and postprocessing scripts, and their intermediate results, can also be different branch-to-branch. The “main-*” branches are meant to be starting points (i.e. trunks) for each model branch (i.e. sprouts). Please see the Branch Navigation section in the top-level README.md in the GitHub repository for more details. There is also one hidden directory “.github/workflows”. This hidden directory contains information for how to run the ML workflow as an end-to-end automated GitHub Action but it is not needed for reusing the ML models archived here. Please the top-level README.md in the GitHub repository for more details on the automation.
Systems with different interactions could develop the same critical behavior due to the underlying symmetry and universality. Using this principle of universality, we can embed critical correlations modeled on the 3D Ising model into the simulated data of heavy-ion collisions, hiding weak signals of a few inter-particle correlations within a large particle cloud. Employing a point cloud network with dynamical edge convolution, we are able to identify events with critical fluctuations through supervised learning, and pick out a large fraction of signal particles used for decision-making in each single event.
Charge transport in polymers, such as polyacrylonitrile (PAN), is crucial for electronics and energy storage. For instance, PAN can transport cations e.g., Li + , by facilitating dynamic cation-nitrile coordination in batteries. However, little is known regarding the underlying role of complex reactive polymer configurations. Herein, we develop a deep-learning potential, trained on ab initio energies and forces of nonequilibrium reactive PAN configurations, to unravel the kinetics of PAN cyclization initiated by a nucleophile (OH – dissociated from LiOH) attacking the terminal nitrile carbon. We find, based on the reaction free-energetics, rates, and charge analysis, that the nucleophile attack producing the first ring is the rate-limiting step, which subsequently triggers Li + -coupled electron transfer along the PAN backbone, causing ∼10 4 times faster sequential ring-formation of the remaining nitriles. PAN’s extended configurations, where dipolar and H-bonding interactions are minimal, enable such rapid kinetics. By validating our computational findings with IR and NMR experiments, we establish a pathway for designing reactive polymers with enhanced charge transport for energy applications.
Human mobility is a primary driver of infectious disease spread. However, existing data is limited in availability, coverage, granularity, and timeliness. Data-driven forecasts of disease dynamics are crucial for decision-making by health officials and private citizens alike. In this work, we focus on a machine-learned anonymized mobility map (hereon referred to as AMM) aggregated over hundreds of millions of smartphones and evaluate its utility in forecasting epidemics. We factor AMM into a metapopulation model to retrospectively forecast influenza in the USA and Australia. We show that the AMM model performs on-par with those based on commuter surveys, which are sparsely available and expensive. We also compare it with gravity and radiation based models of mobility, and find that the radiation model’s performance is quite similar to AMM and commuter flows. Additionally, we demonstrate our model’s ability to predict disease spread even across state boundaries. Our work contributes towards developing timely infectious disease forecasting at a global scale using human mobility datasets expanding their applications in the area of infectious disease epidemiology.
Molecular Dynamics (MD) simulations are used to understand the effects of corrosion on metallic materials in salt brine. Reactive force fields in classical MD enable accurate modeling of bond formation and breakage in the aqueous medium and at the metal-electrolyte interface, while also facilitating dynamic partial charge equilibration. However, MD simulations are computationally intensive and unsuitable for modeling the long time scales characteristic of corrosive phenomena. To address this, we develop reduced-order machine learning models that provide accurate and efficient predictions of charge density in corrosive environments. Specifically, we use Long Short-Term Memory (LSTM) networks to forecast charge density evolution based on atomic environments represented by Smooth Overlap of Atomic Positions (SOAP) descriptors. A physics-informed loss function enforces charge neutrality and electronegativity equivalence. The atomic charges predicted by the deep learning model trained on this work were obtained two orders of magnitude faster than those from molecular dynamics (MD) simulations, with an error of less than 3% compared to the MD-obtained charges, even in extrapolative scenarios, while adhering to physical constraints. This demonstrates the excellent accuracy, computational efficiency, and validity of the developed model. Lastly, even though developed for corrosion, these protocols are formulated in a phenomenon-agnostic manner, allowing application to various variable-charge interatomic potentials and related fields.
A high‐performance cathode is necessary to realize the great potentials of solid‐state batteries such as high energy density and long cycle life. It is also needed to validate electrolyte performance, which is lacking. Currently, cathodes for solid‐state batteries are thinner and have lower cathode active material content than their lithium‐ion battery counterpart, resulting from insufficient conductivity and limiting the battery energy density. This review article provides an overview of recent development in cathode microstructures and their impact on battery properties, including compatibility between cathode and electrolyte, cathode architecture design, interface engineering, correlation between material properties, cathode processing approaches, and performance, as well as the advanced characterization methods used to understand the above correlations . Some perspectives on future development are shared including utilizing in situ and operando characterization tools to better understand dynamic evolution of the cathode/electrolyte interface, adapting artificial intelligence and machine learning to design and optimize cathode structures. The article is aimed to promote research interest on cathode development and advance solid‐state battery technologies.
Here we introduce a denoising diffusion algorithm to discover microstructures with nonlinear fine-tuned properties. Denoising diffusion probabilistic models are generative models that use diffusion-based dynamics to gradually denoise images and generate realistic synthetic samples. By learning the reverse of a Markov diffusion process, we design an artificial intelligence to efficiently manipulate the topology of microstructures to generate a massive number of prototypes that exhibit constitutive responses sufficiently close to designated nonlinear constitutive behaviors. To identify the subset of microcstructures with sufficiently precise fine-tuned properties, a convolutional neural network surrogate is trained to replace high-fidelity finite element simulations to filter out prototypes outside the admissible range. Results of this study indicate that the denoising diffusion process is capable of creating microstructures of fine-tuned nonlinear material properties within the latent space of the training data. More importantly, this denoising diffusion algorithm can be easily extended to incorporate additional topological and geometric modifications by introducing high-dimensional structures embedded in the latent space. Numerical experiments are conducted on the open-source mechanical MNIST data set (Lejeune, 2020). Consequently, this algorithm is not only capable of performing inverse design of nonlinear effective media, but also learns the nonlinear structure–property map to quantitatively understand the multiscale interplay among the geometry, topology, and their effective macroscopic properties.
Data-driven machine learning force fields (MLF) are more and more popular in atomistic simulations, and exploit machine learning methods to predict energies and forces for unknown structures based on the knowledge learned from an existing reference database. The latter usually comes from density functional theory calculations. One main drawback of MLFs is that physical laws are not incorporated in the machine learning models and instead, MLFs are designed to be very flexible to simulate complex quantum chemistry potential energy surface (PES). In general, MLFs have poor transferability, and hence a very large trainset is required to span all the target feature space to get a reliable MLF. This procedure becomes more troublesome when the PES is complicated, with a large number of degrees of freedom, in which building a large database is inevitable and very expensive, especially when accurate but costly exchange-correlation functionals have to be used. In this manuscript, we exploit a high dimensional neural network potential (HDNNP) on Pt clusters of size 6 to 20 as one example. Furthermore, our standard level of energy calculation is DFT GGA (PBE) using a plane wave basis set. We introduce an approximate but fast level with the PBE functional and a minimal atomic orbital basis set, then a more accurate but expensive level, using a hybrid functional or non-local vdw functional and a plane wave basis set, is reliably predicted by learning the difference with HDNNP. Findings show that such a differential approach (named ΔHDNNP) can deliver very accurate predictions (error < 10 meV/atom) in reference to converged basis set energies as well as more accurate but expensive xc functional. The overall speedup can be as large as 900 for 20 atom Pt cluster. More importantly, ΔHDNNP shows much better transferability due to the intrinsic smoothness of delta potential energy surface, and accordingly one can use much smaller trainset data to obtain better accuracy than the conventional HDNNP. A multi-layer ΔHDNNP is thus proposed to obtain very accurate predictions versus expensive non-local vdW functional calculations in which the required trainset is further reduced. The approach can be easily generalized to any other machine learning methods and opens a path to study the structure and dynamics of Pt clusters and nanoparticles.
Abstract Atomic simulations provide an effective means to understand the underlying physics of structural phase transformations. However, this remains a challenge for certain allotropic metals due to the failure of classical interatomic potentials to represent the multitude of bonding. Based on machine-learning (ML) techniques, we develop a hybrid method in which interatomic potentials describing martensitic transformations can be learned with a high degree of fidelity from ab initio molecular dynamics simulations (AIMD). Using zirconium as a model system, for which an adequate semiempirical potential describing the phase transformation process is lacking, we demonstrate the feasibility and effectiveness of our approach. Specifically, the ML-AIMD interatomic potential correctly captures the energetics and structural transformation properties of zirconium as compared to experimental and density-functional data for phonons, elastic constants, as well as stacking fault energies. Molecular dynamics simulations successfully reproduce the transformation mechanisms and reasonably map out the pressure–temperature phase diagram of zirconium.
Incorporating Connected and Automated Vehicles (CAVs) into urban traffic networks presents opportunities and challenges for traffic management systems. This paper aims to develop an integrated routing and traffic signal control system designed explicitly for CAVs, utilizing a Reinforcement Learning (RL) approach. The objective is to enhance traffic flow and improve overall transportation efficiency in the controlled areas. We propose an innovative framework that employs the Deep Reinforcement Learning (DRL) algorithm, especially the Deep Q-network (DQN), to dynamically adjust the number of vehicles in the routes and the duration of traffic signals. Our simulation results demonstrate that a DQN agent successfully optimizes the number of vehicles in the routes and traffic signal timings of traffic signal controllers, eventually reducing total travel time. The study illustrates the potential usage of RL-based systems in managing routing and traffic signals for CAVs, offering a promising opportunity for future urban traffic management strategies.
The goal of this project was to study the effects of DNA-based, force-generating motor proteins on the structure and dynamics of a DNA hydrogel. Motor proteins are nanoscale transducers, converting chemical energy embedded in the solution into local mechanical work on hydrogel strands, and thus potentially driving structural changes and/or non-equilibrium dynamics within the gel material. To explore this, we used self-assembly to create condensed DNA phases, and activated the phases with proteins. The specific aim was to study the deformations (strain fields) generated within a DNA gel by motor forces. We succeeded in this aim, developing methods to experimentally create gel/motor systems, and measure strain with high spatio-temporal resolution. A key finding was that simple continuum strain-field models fail to describe the data. A second major outcome was the development of novel models of hydrogel elasticity incorporating solvent effects that can be used to model dynamic motor-driven strains. A third major outcome was to learn how to control the phase and structure of condensed DNA particles, including both liquid-crystalline behavior, and the formation of DNA liquids.
The software Fire Risk Investigation in 3D (FRI3D) has been developed over the last 2 years to integrate 3D spatial modeling with existing fire probabilistic risk assessment (PRA) models and fire simulation codes. The goal of this research and development is to automate many of the fire analysis manual tasks to reduce industry efforts in the initial fire modeling and operational costs for the model maintenance and evaluations required during normal plant operations. The tasks for Fiscal Year (FY) 2021 include first testing the FRI3D modeling capabilities by importing an industry fire model into FRI3D and making a 3D model of a complex/high-risk significant area. (For this work, the switchgear room was chosen.) Then, the second task of FY 2021 is to develop a dynamic fire PRA process that can help optimize traditional fire PRA models. The switchgear room model will be used for the dynamic fire PRA work. This report describes the work and insights learned when using FRI3D software to model both a Nuclear Regulatory Report (NUREG) example models and a full industry switchgear room.