Probing Atomic Distributions in Mono- and Bimetallic Nanoparticles by Supervised Machine Learning
Abstract not provided
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.
Abstract not provided
To understand how redox processes influence carbon, nitrogen, and iron cycling within the intrameander hyporheic zone, we developed a biotic and abiotic reaction network and incorporated it into the reactive transport simulator PFLOTRAN. Additionally, two-dimensional reactive flow and transport simulations were performed (1) to evaluate how transient hydrological conditions control the lateral redox zonation within an intrameander region of the East River in Colorado and (2) to quantify the impact of a single meander on subsurface exports of carbon and other geochemical species to the river. The meander's overall contribution to the river was quantified by integrating geochemical outfluxes along the outside of the meander bend. The model was able to capture the field-observed trends of dissolved oxygen, nitrate, iron, pH, and total inorganic carbon along a 2-D transect. Consistent with field observations, simulated dissolved oxygen and nitrate decreased along the intrameander flow paths while iron (Fe 2+ ) concentration increased. The simulation results further demonstrated that the reductive potential of the lateral redox zonation was controlled by groundwater velocities resulting from river stage fluctuations, with low-water conditions promoting reducing conditions. The sensitivity analysis results showed that permeability had a more significant impact on biogeochemical zonation compared to the reaction pathways under transient hydrologic conditions. Lastly, the simulation results further indicated that the meander acted as a sink for organic and inorganic carbon as well as iron during the extended baseflow and high-water conditions; however, these geochemical species were released into the river during the falling limb of the hydrograph.
Abstract To bolster the accuracy of existing methods for automated phase identification from X-ray diffraction (XRD) patterns, we introduce a machine learning approach that uses a dual representation whereby XRD patterns are augmented with simulated pair distribution functions (PDFs). A convolutional neural network is trained directly on XRD patterns calculated using physics-informed data augmentation, which accounts for experimental artifacts such as lattice strain and crystallographic texture. A second network is trained on PDFs generated via Fourier transform of the augmented XRD patterns. At inference, these networks classify unknown samples by aggregating their predictions in a confidence-weighted sum. We show that such an integrated approach to phase identification provides enhanced accuracy by leveraging the benefits of each model’s input representation. Whereas networks trained on XRD patterns provide a reciprocal space representation and can effectively distinguish large diffraction peaks in multi-phase samples, networks trained on PDFs provide a real space representation and perform better when peaks with low intensity become important. These findings underscore the importance of using diverse input representations for machine learning models in materials science and point to new avenues for automating multi-modal characterization.
Abstract Advances in the field of goal-directed molecular optimization offer the promise of finding feasible candidates for even the most challenging molecular design applications. One example of a fundamental design challenge is the search for novel stable radical scaffolds for an aqueous redox flow battery that simultaneously satisfy redox requirements at the anode and cathode, as relatively few stable organic radicals are known to exist. To meet this challenge, we develop a new open-source molecular optimization framework based on AlphaZero coupled with a fast, machine-learning-derived surrogate objective trained with nearly 100,000 quantum chemistry simulations. The objective function comprises two graph neural networks: one that predicts adiabatic oxidation and reduction potentials and a second that predicts electron density and local three-dimensional environment, previously shown to be correlated with radical persistence and stability. With no hard-coded knowledge of organic chemistry, the reinforcement learning agent finds molecule candidates that satisfy a precise combination of redox, stability and synthesizability requirements defined at the quantum chemistry level, many of which have reasonable predicted retrosynthetic pathways. The optimized molecules show that alternative stable radical scaffolds may offer a unique profile of stability and redox potentials to enable low-cost symmetric aqueous redox flow batteries.
In this paper, we propose swarm intelligence algorithms to deal with dynamical and spatial organization emergence. The goal is to model and simulate the developement of spatial centers using multi-criteria. We combine a decentralized approach based on emergent clustering mixed with spatial constraints or attractions. We propose an extension of the ant nest building algorithm with multi-center and adaptive process. Typically, this model is suitable to analyse and simulate urban dynamics like gentrification or the dynamics of the cultural equipment in urban area.
Symbolic regression (SR) with a multi-gene genetic program has been used to elucidate new empirical equations describing diffusion in Lennard-Jones (LJ) fluids. Some examples include equations to predict self-diffusion in pure LJ fluids and equations describing the finite-size correction for self-diffusion in binary LJ fluids. The performance of the SR-obtained equations was compared to that of both the existing empirical equations in the literature and to the results from artificial neural net (ANN) models recently reported. It is found that the SR equations have improved predictive performance in comparison to the existing empirical equations, even though employing a smaller number of adjustable parameters, but show an overall reduced performance in comparison to more extensive ANNs.
We investigate financial market dynamics by introducing a heterogeneous agent-based opinion formation model. In this work, we organize individuals in a financial market according to their trading strategy, namely, whether they are noise traders or fundamentalists. The opinion of a local majority compels the market exchanging behavior of noise traders, whereas the global behavior of the market influences the decisions of fundamentalist agents. We introduce a noise parameter, q , to represent the level of anxiety and perceived uncertainty regarding market behavior, enabling the possibility of adrift financial action. We place individuals as nodes in an Erdös-Rényi random graph, where the links represent their social interactions. At any given time, individuals assume one of two possible opinion states ±1 regarding buying or selling an asset. The model exhibits fundamental qualitative and quantitative real-world market features such as the distribution of logarithmic returns with fat tails, clustered volatility, and the long-term correlation of returns. We use Student’s t distributions to fit the histograms of logarithmic returns, showing a gradual shift from a leptokurtic to a mesokurtic regime depending on the fraction of fundamentalist agents. Furthermore, we compare our results with those concerning the distribution of the logarithmic returns of several real-world financial indices.
We present calculations showing that upcoming Cosmic Microwave Background (CMB) experiments will have the power to improve on current constraints on neutrino masses and provide new limits on neutrino degeneracy parameters. The latter could surpass those derived from Big Bang Nucleosynthesis (BBN) and the observationally-inferred primordial helium abundance. These conclusions derive from our Monte Carlo Markov Chain (MCMC) simulations which incorporate a full BBN nuclear reaction network. This provides a self-consistent treatment of the helium abundance, the baryon number, the three individual neutrino degeneracy parameters and other cosmological parameters. Our analysis focuses on the effects of gravitational lensing on CMB constraints on neutrino rest mass and degeneracy parameter. We find for the PLANCK experiment that total (summed) neutrino mass M{sub ν} > 0.29 eV could be ruled out at 2σ or better. Likewise neutrino degeneracy parameters ξ{sub ν{sub e}} > 0.11 and |ξ{sub ν{sub μ{sub /{sub τ}}}}| > 0.49 could be detected or ruled out at 2σ confidence, or better. For POLARBEAR we find that the corresponding detectable values are M{sub ν} > 0.75 eV, ξ{sub ν{sub e}} > 0.62, and |ξ{sub ν{sub μ{sub /{sub τ}}}}| > 1.1, while for EPIC we obtain M{sub ν} > 0.20 eV, ξ{sub ν{sub e}} > 0.045, and |ξ{sub ν{sub μ{sub /{sub τ}}}}| > 0.29. Our forcast for EPIC demonstrates that CMB observations have the potential to set constraints on neutrino degeneracy parameters which are better than BBN-derived limits and an order of magnitude better than current WMAP-derived limits.
Optics tuning in transfer lines and LINACs can be challenging due to the fact that multiple combinations of machine settings can lead to the same diagnostic output. Moreover, the lack of a periodic solution can limit the ability to infer optics in the same way as rings from BPM signals. Model based approaches are often used to assist with the optics tuning in combination with optimization or parameter estimation. Here we have developed a novel approach using machine learning inverse models trained on a known configuration to detect variations in quadrupole settings without explicitly including them in the model. This paper shows a comparison of neural network models and linear models on both a simulation based study and experimental studies conducted at the AGS to RHIC transfer line at Brookhaven National Lab.
We present DES14X2fna, a high-luminosity, fast-declining Type IIb supernova (SN IIb) at redshift z = 0.0453, detected by the Dark Energy Survey (DES). DES14X2fna is an unusual member of its class, with a light curve showing a broad, luminous peak reaching M r ≃ –19.3 mag 20 d after explosion. This object does not show a linear decline tail in the light curve until ≃60 d after explosion, after which it declines very rapidly (4.30 ± 0.10 mag 100 d –1 in the r band). By fitting semi-analytic models to the photometry of DES14X2fna, we find that its light curve cannot be explained by a standard 56 Ni decay model as this is unable to fit the peak and fast tail decline observed. Inclusion of either interaction with surrounding circumstellar material or a rapidly-rotating neutron star (magnetar) significantly increases the quality of the model fit. Furthermore, we also investigate the possibility for an object similar to DES14X2fna to act as a contaminant in photometric samples of SNe Ia for cosmology, finding that a similar simulated object is misclassified by a recurrent neural network (RNN)-based photometric classifier as an SN Ia in ~1.1–2.4 per cent of cases in DES, depending on the probability threshold used for a positive classification.
Abstract For over 10 years, ModelSEED has been a primary resource for the construction of draft genome-scale metabolic models based on annotated microbial or plant genomes. Now being released, the biochemistry database serves as the foundation of biochemical data underlying ModelSEED and KBase. The biochemistry database embodies several properties that, taken together, distinguish it from other published biochemistry resources by: (i) including compartmentalization, transport reactions, charged molecules and proton balancing on reactions; (ii) being extensible by the user community, with all data stored in GitHub; and (iii) design as a biochemical ‘Rosetta Stone’ to facilitate comparison and integration of annotations from many different tools and databases. The database was constructed by combining chemical data from many resources, applying standard transformations, identifying redundancies and computing thermodynamic properties. The ModelSEED biochemistry is continually tested using flux balance analysis to ensure the biochemical network is modeling-ready and capable of simulating diverse phenotypes. Ontologies can be designed to aid in comparing and reconciling metabolic reconstructions that differ in how they represent various metabolic pathways. ModelSEED now includes 33,978 compounds and 36,645 reactions, available as a set of extensible files on GitHub, and available to search at https://modelseed.org and KBase.
In this paper, a novel three-phase converter is proposed for ac to dc wireless power transfer (WPT) systems for electric vehicle (EV) charging applications. The proposed innovative solution, called as Oak Ridge Converter, reduces the design complexity and cost by eliminating the front-end converter stage compared to the conventional systems. Additionally, grid side requirements can be met with the proposed creative concept. In this concept, the three-phase single-stage Oak Ridge Converter directly converts the 60 Hz grid frequency into high-frequency voltage and utilizes hybrid grid-source and high-frequency to realize power transfer from AC source through resonant network and coupling coils to the battery load. Simulation validation of the proposed three-phase system is currently being carried on and the results will be will be provided to validate the theoretical studies with the input of 277 VAC,RMS and output of 675 VDC at 35 kW power. The system current total harmonic distortion (THD) is measured 5% with a power factor (PF) of 0.98 and overall hardware development of the system that will be used for experiments is presented.
Koopman operator theory provides a model-free technique for studying nonlinear dynamical systems purely from data. Since the Koopman operator is infinite-dimensional, researchers have developed several methods that provide a finite-dimensional approximation of the Koopman operator so that it can be applied for practical use cases. One common thing with most of the methods is that their solutions are obtained by solving a centralized minimization problem. In this work, we treat the dynamical system to be a multi-agent system and propose an algorithm to compute the finite-dimensional approximation of the Koopman operator in a distributed manner using the knowledge of the topology of the underlying multi-agent system. The proposed distributed approach is shown to be equivalent to the centralized learning problem and results in a sparse Koopman whose block structure mimics the Laplacian of the multi-agent system. Extensive simulation studies illustrate the proposed framework on the network of oscillators and the IEEE 68 bus system.
This paper presents the design, analyses and implementation of an optimized single stage high power wireless charging system capable of transferring 100 kW at an operating frequency of 22 kHz and a coil-to-coil distance of 5 inches.The detailed design and implementation of the power electronics including the high frequency inverter and rectifier and wireless power transfer coils along with the resonant stage are described. FEA simulation results of the coils and the resonant network analysis are validated by using a Venable frequency response analyzer. Experimental results corresponding to 50 kW operation are presented to validate the design process.
When an unintentional island is formed on the electric power system, distributed energy resources (DERs) are typically required to detect and de-energize the island. This requirement may become more challenging as the number of DERs in an island rises. Thus, it is of interest to experimentally verify whether DERs can successfully detect and de-energize islands containing many DERs connected at different points. This paper presents a power hardware-in-the-loop (PHIL) platform for testing the duration of islands containing multiple inverters connected at multiple points in a network. The PHIL platform uses real-time simulation to represent islanded distribution circuits, with DER inverters connected in hardware. This allows efficient testing of a large number of island configurations simply by changing the distribution circuit model in the real-time simulator. A method for calculating the quality factor of an arbitrary distribution circuit in real time, designed for anti-islanding tests, is also presented. Experimental results are included demonstrating the use of the PHIL method to test a variety of three-inverter islands.
Here, we introduce a data-driven and physics-informed framework for propagating uncertainty in stiff, multiscale random ordinary differential equations (RODEs) driven by correlated (colored) noise. Unlike systems subjected to Gaussian white noise, a deterministic equation for the joint probability density function (PDF) of RODE state variables does not exist in closed form. Moreover, such an equation would require as many phase-space variables as there are states in the RODE system. To alleviate this curse of dimensionality, we instead derive exact, albeit unclosed, reduced-order PDF (RoPDF) equations for low-dimensional observables/quantities of interest. The unclosed terms take the form of state-dependent conditional expectations, which are directly estimated from data at sparse observation times. However, for systems exhibiting stiff, multiscale dynamics, data sparsity introduces regression discrepancies that compound during RoPDF evolution. This is overcome by introducing a kinetic-like defect term to the RoPDF equation, which is learned by assimilating in sparse, low-fidelity RoPDF estimates. Two assimilation methods are considered, namely nudging and deep neural networks, which are successfully tested against Monte Carlo simulations.
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.
The software, called the Urban Deployment Model (UDM), is an integrated set of tools for the simulation and analysis of static and mobile radiation detection systems in complex urban environments. The UDM includes a graphical user interface through which the workflow associated with setting up and executing a simulation may be established. The modular toolsets combine aspects of radiation transport, navigation, graph theory, optimization, and visualization to allow the complete end-to-end simulation of large scale, highly-distributed sensor networks.