Physics guided neural networks for modelling of non-linear dynamics
Not Available
SEARCH · Search NASA
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.
Not Available
Mathematical modeling of plant metabolism allows the plant science community to understand the organization of plant metabolism, obtain quantitative insights into metabolic functions, and derive engineering strategies for manipulation of metabolism. Among the various modeling approaches, metabolic pathway analysis can dissect the basic functional modes of subsections of core metabolism, such as photorespiration, and reveal how classical definitions of metabolic pathways have overlapping functionality. In current research using constraint-based modeling in plants, numerous computational tools are currently available to analyze large-scale and genome-scale metabolic networks. For 13 C-metabolic flux analysis, principles of isotopic steady state have been used to study heterotrophic plant tissues, while nonstationary isotope labeling approaches are amenable to the study of photoautotrophic and secondary metabolism. Enzyme kinetic models explore pathways in mechanistic detail, and we discuss different approaches to determine or estimate kinetic parameters. In this review, we describe recent advances and challenges in modeling plant metabolism.
The dynamics of powertrain control systems are complicated and involve both nonlinear plant model and control functionalities, albeit they are well defined and formulated using first principle approaches. This constitutes difficulties in exploring implementable optimal tuning rules for some selected control parameters using vehicle-to-vehicle (V2V) communications. This paper presents a way to use neural networks (NN) to represent the problem of parameter tuning for optimizing fuel consumption. For this purpose, physical modelling and validation have been firstly performed for the closed loop powertrain system of the concerned vehicle for some given driving cycles. This is then followed by the sensitivity analysis that selects most influential control parameters to optimize. Using the data generated from the obtained physical models, an equivalent NN formulation has finally been obtained that gives simple yet unified objectives and constraints ready to be used to solve the optimization problem that produces optimal tuning rules for the selected control parameters to minimize fuel consumption.
The axion insulator is a higher-order topological insulator protected by inversion symmetry. We show that, under quenched disorder respecting inversion symmetry on average, the topology of the axion insulator stays robust, and an intermediate metallic phase in which states are delocalized is unavoidable at the transition from an axion insulator to a trivial insulator. We derive this conclusion from general arguments, from classical percolation theory, and from the numerical study of a 3D quantum network model simulating a disordered axion insulator through a layer construction. We find the localization length critical exponent near the delocalization transition to be ν = 1.42 ± 0.12. We further show that this delocalization transition is stable even to weak breaking of the average inversion symmetry, up to a critical strength. Finally, we also quantitatively map our quantum network model to an effective Hamiltonian and we find its low-energy k ∙ p expansion.
Abstract The deformation behavior of Ti-6Al-4V titanium alloy is significantly influenced by slip localized within crystallographic slip bands. Experimental observations reveal that intense slip bands in Ti-6Al-4V form at strains well below the macroscopic yield strain and may serially propagate across grain boundaries, resulting in long-range localization that percolates through the microstructure. These connected, localized slip bands serve as potential sites for crack initiation. Although slip localization in Ti-6Al-4V is known to be influenced by various factors, an investigation of optimal microstructures that limit localization remains lacking. In this work, we develop a novel strategy that integrates an explicit slip band crystal plasticity technique, graph networks, and neural network models to identify Ti-6Al-4V microstructures that reduce the propensity for strain localization. Simulations are conducted on a dataset of 3D polycrystals, each represented as a graph to account for grain neighborhood and connectivity. The results are then used to train neural network surrogate models that accurately predict localization-based properties of a polycrystal, given its microstructure. These properties include the ratio of slip accumulated in the band to that in the matrix, fraction of total applied strain accommodated by slip bands, and spatial connectivity of slip bands throughout the microstructure. The initial dataset is enriched by synthetic data generated by the surrogate models, and a grid search optimization is subsequently performed to find optimal microstructures. Describing a 3D polycrystal with only a few features and a combination of graph and neural network models offer robustness compared to the alternative approaches without compromising accuracy. We show that while each material property is optimized through a unique microstructure solution, elongated grain shape emerges as a recurring feature among all optimal microstructures. This finding suggests that designing microstructures with elongated grains could potentially mitigate strain localization without compromising strength.
A set of deep neural network models for rheumatoid arthritis (RA) classification using a highway network, a convolutional neural network and a residual network is proposed based on the data of diffuse optical tomography (DOT) utilising near-infrared light, which ensures early diagnosis of pathophysiological changes resulting from inflammation. A numerical model of the finger is used to generate images to overcome the inherent problem of insufficient clinical DOT images available. The proposed deep neural network models are applied to automatically classify simulated DOT images of inflamed and non-inflamed joints and transfer learning is also used to improve the performance of the classification. The results demonstrate that all three deep neural network methods improve the diagnostic accuracy as compared to the widely applied support vector machine (SVM), especially for high inter-subject variability databases. In cases of distinct modelled severity of disease, residual network achieved the highest accuracy (> 99 %), and both of highway and convolutional neural networks reached 99 %, respectively. However, as the severity of the modelled disease is reduced, this accuracy is reduced to 75.2 % for residual networks. The results indicate that transfer learning can improve the performance of deep neural network methods on RA classification from DOT data and highlight their potential as a computer aided tool in DOT diagnostic systems. (paper)
The overall goal of the project was to bridge the gap between our knowledge of small-scale geochemical reaction rates and reaction rates meaningful for modeling transport at core scales. The working hypothesis was that reaction rates, determined from laboratory measurements based upon reactions typically conducted in well mixed batch reactors using pulverized reactive media may be significantly changed in in situ porous media flow due to rock microstructure heterogeneity. Specifically we hypothesized that, generally, reactive mineral surfaces are not uniformly accessible to reactive fluids due to the random deposition of mineral grains and to the variation in flow rates within a pore network. Expected bulk reaction rates would therefore have to be correctly up-scaled to reflect such heterogeneity. The specific objective was to develop a computational tool that integrates existing measurement capabilities with pore-scale network models of fluid flow and reactive transport. The existing measurement capabilities to be integrated consisted of (a) pore space morphology, (b) rock mineralogy, and (c) geochemical reaction rates. The objective was accomplished by: (1) characterizing sedimentary sandstone rock morphology using X-ray computed microtomography, (2) mapping rock mineralogy using back-scattered electron microscopy (BSE), X-ray dispersive spectroscopy (EDX) and CMT, (3) characterizing pore-accessible reactive mineral surface area, and (4) creating network models to model acidic CO{sub 2} saturated brine injection into the sandstone rock samples.
Popular HPC network interconnection simulators such as SST/macro provide a variety of configurable parameters to explore the design space of hardware components such as network interface cards (NIC), switches, and links among them. While such knobs provide flexibility to explore design trade-offs for novel hardware, manually configuring simulations for matching configurations of the existing hardware to focus on topology exploration can be cumbersome and error-prone, leading to widely inaccurate simulations. This challenge is compounded when specifications of various (proprietary) technologies are not readily available or intentionally omitted. In this work, we propose a framework to autotune the multiple network models’ simulation configurations within SST/macro using Tree-structured Parzen Estimator-based Bayesian optimization to observe the effect on simulation accuracy across different message regimes. These regimes consist of small to large message sizes and latency to bandwidth-bound messages. We provide a detailed analysis of the simulation error for four representative HPC systems. Our Bayesian optimization based autotuning framework for network models achieves a maximum of 5x improvement in accuracy over best-effort manual configurations based on available hardware specifications.
Critical infrastructure resilience has become a national priority for the U. S. Department of Homeland Security. System resilience has been studied for several decades in many different disciplines, but no standards or unifying methods exist for critical infrastructure resilience analysis. This report documents the results of a late-start Laboratory Directed Research and Development (LDRD) project that investigated the identification of optimal recovery strategies that maximize resilience. To this goal, we formulate a bi-level optimization problem for infrastructure network models. In the 'inner' problem, we solve for network flows, and we use the 'outer' problem to identify the optimal recovery modes and sequences. We draw from the literature of multi-mode project scheduling problems to create an effective solution strategy for the resilience optimization model. We demonstrate the application of this approach to a set of network models, including a national railroad model and a supply chain for Army munitions production.
Traffic congestion leads to severe problems especially in urban traffic networks. It increases the chance of accidents, energy waste, and social costs. In order to address these problems, an adaptive linear quadratic regulator (LQR) approach is developed for traffic signal control at multiple intersections in an urban area. The proposed method controls the green time of the traffic signals to reduce traffic congestion and smooth traffic flow. Real-world data from vision-based traffic sensors are used to build the traffic network model, which mimics the real-world traffic behavior. In addition, the proposed control utilizes recursive least square parameter estimation, which is capable of tracking dynamic changes in traffic conditions. Simulation of Urban MObility (SUMO) is used to analyze the efficacy of the proposed method. Results of the simulation show that the proposed method outperforms pretimed control in various aspects.
Highlights: • X-ray μ-CT was used to characterize 3D structures of the pores in SiCp preforms. • 3D pore network models of the pore structure were extracted and analyzed. • With the increase of particle size, the homogeneity of pore distribution decreased. • Small particles in large spaces of coarse particles decreased preform connectivity. - Abstract: Silicon-carbide particle (SiC{sub p}) preforms with particle sizes of 20, 50, 100 and 150 μm were prepared by compression mold forming and heat treatment. A high resolution (~ 1.0 μm) three-dimensional (3D) X-ray micro-computed tomography (X-ray μ-CT) combined with 3D pore-network models were used to study the effect of particle size on 3D characteristics and spatial structures of pores in the preforms. The results demonstrated that as the particle size increased, the gap expansion caused by burning of starch reduced. When the particle size increased from 20 to 100 μm, the average areal porosity, homogeneity of pore distribution, connectivity, the numbers of small pores and throats, and average pore coordination number decreased; the average effective sizes of pores and throats increased. When the particle size further increased to 150 μm, the large gaps were filled by small particles resulted from the broken coarse particles, which decreased the homogeneity and connectivity of pores. As a result, the pore-network analysis shows that the numbers of small pores and throats and average pore coordination number increased.
Improved understanding of structural and chemical properties through local experimental probes, such as X-ray absorption near-edge structure (XANES) spectroscopy, is crucial for the understanding and design of functional materials. In recent years, significant advancements have been made in the development of data science approaches for the automated interpretation of XANES structure–spectrum relationships. However, existing studies have primarily focused on crystalline solids and small molecules, while fewer efforts have been devoted to disordered systems. Thus, in this work, we demonstrate the development of neural network models for predicting and interpreting XANES spectra of amorphous carbon (a-C) from local structural descriptors. Comparison between different structural descriptors expectedly shows that the inclusion of both bond length and bond angle information is necessary for an accurate prediction of the spectra. Among the descriptors considered in this work, we find that the local many-body tensor representation yields the highest accuracy and greatest interpretability so that it can be leveraged to understand the importance of structural motifs in determining XANES spectra. Furthermore, we also discuss performance of neural network models for predicting both local structure features, such as bond lengths and bond angles, and global chemical composition, such as the sp:sp 2 :sp 3 ratio.
Abstract. Rivers and river habitats around the world are under sustained pressure from human activities and the changing global environment. Our ability to quantify and manage the river states in a timely manner is critical for protecting the public safety and natural resources. In recent years, vector-based river network models have enabled modeling of large river basins at increasingly fine resolutions, but are computationally demanding. This work presents a multistage, physics-guided, graph neural network (GNN) approach for basin-scale river network learning and streamflow forecasting. During training, we train a GNN model to approximate outputs of a high-resolution vector-based river network model; we then fine-tune the pretrained GNN model with streamflow observations. We further apply a graph-based, data-fusion step to correct prediction biases. The GNN-based framework is first demonstrated over a snow-dominated watershed in the western United States. A series of experiments are performed to test different training and imputation strategies. Results show that the trained GNN model can effectively serve as a surrogate of the process-based model with high accuracy, with median Kling–Gupta efficiency (KGE) greater than 0.97. Application of the graph-based data fusion further reduces mismatch between the GNN model and observations, with as much as 50 % KGE improvement over some cross-validation gages. To improve scalability, a graph-coarsening procedure is introduced and is demonstrated over a much larger basin. Results show that graph coarsening achieves comparable prediction skills at only a fraction of training cost, thus providing important insights into the degree of physical realism needed for developing large-scale GNN-based river network models.
As repositories for CO₂ and radioactive waste, as oil and gas reservoirs, and as contaminated sites needing remediation, rock formations play a central role in energy and environmental management. The connectivity of the rock's porespace strongly affects fluid flow and solute transport. This work examines pore connectivity and its implications for fluid flow and chemical transport. Three experimental approaches (imbibition, tracer concentration profiles, and imaging) were used in combination with network modeling. In the imbibition results, three types of imbibition slope [log (cumulative imbibition) vs. log (imbibition time)] were found: the classical 0.5, plus 0.26, and 0.26 transitioning to 0.5. The imbibition slope of 0.26 seen in Indiana sandstone, metagraywacke, and Barnett shale indicates low pore connectivity, in contrast to the slope of 0.5 seen in the well-connected Berea sandstone. In the tracer profile work, rocks exhibited different distances to the plateau porosity, consistent with the pore connectivity from the imbibition tests. Injection of a molten metal into connected pore spaces, followed by 2-D imaging of the solidified alloy in polished thin sections, allowed direct assessment of pore structure and lateral connection in the rock samples. Pore-scale network modeling gave results consistent with measurements, confirming pore connectivity as the underlying cause of both anomalous behaviors: imbibition slope not having the classical value of 0.5, and accessible porosity being a function of distance from the edge. A poorly connected porespace will exhibit anomalous behavior in fluid flow and chemical transport, such as a lower imbibition slope (in air–water system) and diffusion rate than expected from classical behavior.
Photosynthesis achieves near unity light-harvesting quantum efficiency yet it remains unknown whether there exists a fundamental organizing principle giving rise to robust light harvesting in the presence of dynamic light conditions and noisy physiological environments. Here, we present a noise-canceling network model that relates noisy physiological conditions, power conversion efficiency, and the resulting absorption spectra of photosynthetic organisms. Using light conditions in full solar exposure, light filtered by oxygenic phototrophs, and light filtered under seawater, we derived optimal absorption characteristics for efficient solar power conversion. We show how light-harvesting antennae can be tuned to maximize power conversion efficiency by minimizing excitation noise, thus providing a unified theoretical basis for the observed wavelength dependence of absorption in green plants, purple bacteria, and green sulfur bacteria.
Unit Commitment (UC) problems that consider the Alternating Current (AC) model of the transmission network have long been considered intractable to solve at scale by the power system community. Recently, the Grid-Optimization (GO) Competition held by the Advanced Research Project Agency-Energy (ARPA-E) has facilitated the development of the first algorithms to solve large-scale ACUC problems. This new capability opens a path towards the explicit consideration of the AC transmission network model in UC problems used to clear day-ahead electricity markets. This calls for the analysis of electricity market structures that accommodate both the continuous non-linearity of the AC transmission network and the discrete non-linearity of the UC problem simultaneously. This paper serves as an initial effort to do so by proposing an Average Incremental Cost (AIC) pricing structure that is designed around the ACUC problem. In particular, an AIC one-pass pricing problem is proposed that represents a continuously constrained variant of the ACUC problem and allows for the computation of Locational Incremental Prices (LIPs) for both real and reactive power as the local optimal Lagrange multipliers of the power balance constraints. To avoid degeneracy, the pricing problem includes a small parameter ϵ > 0. Under certain assumptions market participants are shown to realize profit that converges to a non-negative value as ϵ approaches zero, practically ensuring profitability for small values of ϵ. We additionally provide many simple and important examples that provide intuition and insights into the proposed prices. Examples illustrate the basic concept of AIC pricing, the derived profitability results, the existence of multiple LIPs, the importance of including reactive power in the dispatch and pricing problems, the need for reactive power prices, and the improved incentives exhibited by LIPs as compared to traditional Locational Marginal Prices (LMPs). We additionally indicate many directions for future work including analysis of larger test cases.
Electromagnetic interactions between colliding heavy ions at the Large Hadron Collider (LHC) at CERN will give rise to localized beam losses that may quench superconducting magnets, apart from contributing significantly to the luminosity decay. To quantify their impact on the operation of the collider, we have used a three-step simulation approach, which consists of optical tracking, a Monte-Carlo shower simulation and a thermal network model of the heat flow inside a magnet. We present simulation results for the case of {sup 208}Pb{sup 82+} ion operation in the LHC, with focus on the alice interaction region, and show that the expected heat load during nominal {sup 208}Pb{sup 82+} operation is 40% above the quench level. This limits the maximum achievable luminosity. Furthermore, we discuss methods of monitoring the losses and possible ways to alleviate their effect.
The purpose of this study is to examine the importance of the isotopic composition of plutonium on PWR MOX fuel fabrication as part of a multi-tiered fuel cycle, and the effect of its evolution during a transition. Two main effects can modify the isotopic composition of the plutonium used to build the PWR-MOX fuel: decay and depletion. This analysis is based on a transition of the nuclear fuel cycle to a configuration identified as 'Evaluation Group 29' or 'EG29', in the recently-published Evaluation and Screening (E and S) report. The EG29 fuel cycle corresponds to a double strata fleet. The first stratum is composed of sodium-cooled fast reactors (SFRs) multi-recycling plutonium. Surplus plutonium is sent to the second stratum composed of pressurized water reactors with full mixed oxide (U/Pu)O{sub 2} cores (MOX-PWRs). The PWR stratum is also multi-recycling its plutonium. The target energy generation ratio for the final EG29 system is 70:30 (SFR:PWR), with a 1% annual growth in nuclear energy demand. All the calculation of this study have been performed using the Cyclus fuel cycle simulator framework. The facility models are from the Cycamore package, developed by the Cyclus team, and containing a Pu-equivalent fuel fabrication model as well as a simple fixed mixing ratio fuel fabrication model. A new neural network model has also been contributed to the Cyclus environment using the cyCLASS tool, allowing the use of the fuel fabrication and depletion models from the CLASS tool in Cyclus. (authors)