Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “network solutions”

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 1,369 records · Page 76

Physics-informed neural networks for heterogeneous poroelastic media

This study presents a novel physics-informed neural network (PINN) framework for modeling poroelasticity in heterogeneous media with material interfaces. The approach introduces a composite neural network (CoNN) where separate neural networks predict displacement and pressure variables for each material. While sharing identical activation functions, these networks are independently trained for all other parameters. To address challenges posed by heterogeneous material interfaces, the CoNN is integrated with the Interface-PINNs (I-PINNs) framework (Sarma et al., Comput. Methods Appl. Mech. Eng. 429: 117135, 2024), allowing different activation functions across material interfaces. Further, this ensures accurate approximation of discontinuous solution fields and gradients. Performance and accuracy of this combined architecture were evaluated against the conventional PINNs approach, a single neural network (SNN) architecture, and the eXtended PINNs (XPINNs) framework through two one-dimensional benchmark examples with discontinuous material properties. The results show that the proposed CoNN with I-PINNs architecture achieves an RMSE that is two orders of magnitude better than the conventional PINNs approach and is at least 40 times faster than the SNN framework. Compared to XPINNs, the proposed method achieves an RMSE at least one order of magnitude better and is 40% faster.

42 ENGINEERING↗

A Scalable Solution Framework for Stochastic Transmission and Generation Planning Problems. Draft

Current commercial software tools for transmission and generation investment planning have limited stochastic modeling capabilities. Because of this limitation, electric power utilities generally rely on scenario planning heuristics to identify potentially robust and cost effective investment plans for a broad range of system, economic, and policy conditions. Several research studies have shown that stochastic models perform significantly better than deterministic or heuristic approaches, in terms of overall costs. However, there is a lack of practical solution approaches to solve such models. In this paper we propose a scalable decomposition algorithm to solve stochastic transmission and generation planning problems, respectively considering discrete and continuous decision variables for transmission and generation investments. Given stochasticity restricted to loads and wind, solar, and hydro power output, we develop a simple scenario reduction framework based on a clustering algorithm, to yield a more tractable model. The resulting stochastic optimization model is decomposed on a scenario basis and solved using a variant of the Progressive Hedging (PH) algorithm. We perform numerical experiments using a 240-bus network representation of the Western Electricity Coordinating Council in the US. Although convergence of PH to an optimal solution is not guaranteed for mixed-integer linear optimization models, we find that it is possible to obtain solutions with acceptable optimality gaps for practical applications. Our numerical simulations are performed both on a commodity workstation and on a high-performance cluster. The results indicate that large-scale problems can be solved to a high degree of accuracy in at most two hours of wall clock time.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Convolutional Neural Network Surrogate Models for the Mechanical Properties of Periodic Structures

The work in this paper describes neural network surrogate models for calculating the effective mechanical properties of a periodic composites. The models achieve good accuracy even when only provided with training data sampling a small portion of the design space. As an example, the surrogate models are applied to solving the inverse design problem of finding structures with optimal mechanical properties. The surrogate models are sufficiently accurate to recover optimal solutions in general agreement with established topology optimization methods. However, improvements will be required to develop robust, efficient neural network-based surrogate models and several directions for future research are highlighted here.

42 ENGINEERING↗

Decarbonizing the grid: Utilizing demand-side flexibility for carbon emission reduction through locational marginal emissions in distribution networks

Decarbonization of the electric grid has become an important world-wide priority and is actively happening in many ways by introducing innovations and new technologies from the generation sectors to the demand sectors. In particular, one promising pathway toward such net-zero carbon emissions is to utilize the demand-side flexibility with the increasing number of flexible loads in distribution networks. In this paper, we explore a load shifting strategy with the emerging concept of location marginal emissions (LMEs) to reduce carbon emissions. LMEs measure the impact of carbon emissions including the locational aspect in more granular way and thus provide a novel mechanism for the system operator and load aggregators to design the LME-based load shifting strategy, which can efficiently guide consumers and thus adjust their consumption behaviors. Simulation case studies on the IEEE test networks are performed to validate the capability of the proposed load shifting method to reduce carbon emissions. We also compare this with other relevant strategies to discuss multiple scenarios and corresponding results. Finally, while each provides a different level of flexibility, all the explored strategies tested have led to solutions that have lower carbon emissions, indicating the great potential of demand-side flexibility in reducing carbon emissions for future distribution networks.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

Predicting rare events using neural networks and short-trajectory data

Estimating the likelihood, timing, and nature of events is a major goal of modeling stochastic dynamical systems. When the event is rare in comparison with the timescales of simulation and/or measurement needed to resolve the elemental dynamics, accurate prediction from direct observations becomes challenging. In such cases a more effective approach is to cast statistics of interest as solutions to Feynman-Kac equations (partial differential equations). Here, we develop an approach to solve Feynman-Kac equations by training neural networks on short-trajectory data. Our approach is based on a Markov approximation but otherwise avoids assumptions about the underlying model and dynamics. This makes it applicable to treating complex computational models and observational data. Additionally, we illustrate the advantages of our method using a low-dimensional model that facilitates visualization, and this analysis motivates an adaptive sampling strategy that allows on-the-fly identification of and addition of data to regions important for predicting the statistics of interest. Finally, we demonstrate that we can compute accurate statistics for a 75-dimensional model of sudden stratospheric warming. This system provides a stringent test bed for our method.

97 MATHEMATICS AND COMPUTING↗

On the role of shear in cosmological averaging

Using the spherically symmetric inhomogeneous Lemaitre-Tolman-Bondi dust solution, we study how the shear and the backreaction depend on the sharpness of the spatial transition between voids and walls and on the size of the voids. The voids considered here are regions with matter density Ω{sub 0} ≅ 0 and expansion rate H{sub 0}t{sub 0} ≅ 1, while the walls are regions with matter density Ω{sub 0} ≅ 1 and expansion rate H{sub 0}t{sub 0} ≅ 2/3. The results indicate that both the volume-average shear and the variance of the expansion rate grow proportional to the sharpness of the transition and diverge in the limit of a step function, but, for realistic-sized voids, are virtually independent of the size of the void. However, the backreaction, given by the difference of the variance and the shear, has a finite value in the step-function limit. By comparing the exact result for the backreaction to the case where the shear is neglected by treating the voids and walls as separate Friedmann-Robertson-Walker models, we find that the shear suppresses the backreaction by a factor of (r{sub 0}/t{sub 0}){sup 2}, the squared ratio of the void size to the horizon size. This exemplifies the importance of using the exact solution for the interface between the regions of different expansion rates and densities. The suppression is justified to hold also for a network of compensated voids, but may not hold if the universe is dominated by uncompensated voids.

79 ASTRONOMY AND ASTROPHYSICS↗

Learning the factors controlling mineral dissolution in three-dimensional fracture networks: applications in geologic carbon sequestration

We perform a set of high-fidelity simulations of geochemical reactions within three-dimensional discrete fracture networks (DFN) and use various machine learning techniques to determine the primary factors controlling mineral dissolution. The DFN are partially filled with quartz that gradually dissolves until quasi-steady state conditions are reached. At this point, we measure the quartz remaining in each fracture within the domain as our primary quantity of interest. We observe that a primary sub-network of fractures exists, where the quartz has been fully dissolved out. This reduction in resistance to flow leads to increased flow channelization and reduced solute travel times. However, depending on the DFN topology and the rate of dissolution, we observe substantial variability in the volume of quartz remaining within fractures outside of the primary subnetwork. This variability indicates an interplay between the fracture network structure and geochemical reactions. We characterize the features controlling these processes by developing a machine learning framework to extract their relevant impact. Specifically, we use a combination of high-fidelity simulations with a graph-based approach to study geochemical reactive transport in a complex fracture network to determine the key features that control dissolution. We consider topological, geometric and hydrological features of the fracture network to predict the remaining quartz in quasi-steady state. We found that the dissolution reaction rate constant of quartz and the distance to the primary sub-network in the fracture network are the two most important features controlling the amount of quartz remaining. This study is a first step towards characterizing the parameters that control carbon mineralization using an approach with integrates computational physics and machine learning.

54 ENVIRONMENTAL SCIENCES↗

Differentiable programming for online training of a neural artificial viscosity function within a staggered grid Lagrangian hydrodynamics scheme

Lagrangian methods to solve the inviscid Euler equations produce numerical oscillations near shock waves. A common approach to reducing these oscillations is to add artificial viscosity (AV) to the discrete equations. The AV term acts as a dissipative mechanism that attenuates oscillations by smearing the shock across a finite number of computational cells. However, AV introduces several control parameters that are not determined by the underlying physical model, and hence, in practice are tuned to the characteristics of a given problem. We seek to improve the standard quadratic-linear AV form by replacing it with a learned neural function that reduces oscillations relative to exact solutions of the Euler equations, resulting in a hybrid numerical-neural hydrodynamic solver. Because AV is an artificial construct that exists solely to improve the numerical properties of a hydrodynamic code, there is no offline ‘viscosity data’ against which a neural network can be trained before inserting into a numerical simulation, thus requiring online training. We achieve this via differentiable programming, i.e. end-to-end backpropagation or adjoint solution through both the neural and differential equation code, using automatic differentiation of the hybrid code in the Julia programming language to calculate the necessary loss function gradients. A novel offline pre-training step accelerates training by initializing the neural network to the default numerical AV scheme, which can be learned rapidly by space-filling sampling over the AV input space. We find that online training over early time steps of simulation is sufficient to learn a neural AV function that reduces numerical oscillations in long-term hydrodynamic shock simulations. These results offer an early proof-of-principle that online differentiable training of hybrid numerical schemes with novel neural network components can improve certain performance aspects existing in purely numerical schemes.

97 MATHEMATICS AND COMPUTING↗

Mechanical Properties of Self-Assembled Fmoc-Diphenylalanine Molecular Gels

We explore the phase diagram and mechanical properties of molecular gels produced from mixing water with a dimethyl sulfoxide (DMSO) solution of the aromatic dipeptide derivative fluorenylmethoxycarbonyl-diphenylalanine (Fmoc-FF). Highly soluble in DMSO, Fmoc-FF assembles into fibrous networks that form gels upon addition of water. At high water concentrations, rigid gels can be formed at Fmoc-FF concentrations as low as 0.01 wt %. Here, the conditions are established defining the Fmoc-FF and water concentrations at which gels are formed. Below the gel boundary, the solutions are clear and colorless and have long-term stability. Above the gel boundary, gels are formed with increasing rapidity with increasing water or Fmoc-FF concentrations. A systematic characterization of the effect of Fmoc-FF and water concentrations on the mechanical properties of the gels is presented, demonstrating that the elastic behavior of the gels follows a specific, robust scaling with Fmoc-FF volume fraction. Furthermore, we characterize the kinetics of gelation and demonstrate that these gels are reversible in the sense that they can be disrupted mechanically and rebuild strength over time.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Sampling and noise in vision networks

This research is part of the Human Interface Research Branch-Vision Group's program to develop computable models of biological solutions to general vision system problems. Two problem areas are addressed: effects of discrete sampling by receptors, and effects of visual system noise.

Ahumada, Albert J., Jr.↗

Dynamic model-based sensor network design algorithm for system efficiency maximization

Here, a dynamic model-based sensor network design (DMSND) algorithm has been developed for maximizing system efficiency for an estimator-based control system. The algorithm synthesizes the optimal sensor network in the face of disturbances or set point changes. Computational expense of the large-scale combinatorial optimization problem is significantly reduced by parallel computing and by using combination of three novel strategies: multi-rate sampling frequency, model order reduction, and use of an incumbent solution that enables early termination of evaluation of infeasible sensor sets. The developed algorithm is applied to an acid gas removal unit as part of an integrated gasification combined cycle power plant with carbon capture. Even though there are more than thousand process states and more than hundred candidate sensor locations, the optimal sensor network design problem for maximizing process efficiency could be solved within couple of hours for a given budget.

47 OTHER INSTRUMENTATION↗

Through-thickness microstructural evolution during grain boundary engineering type thermomechanical processing and its implication on sensitization behavior in austenitic stainless steel

Highlights: • GBE-treated 304L steel exhibits thickness-dependent microstructural evolution. • Degree of sensitization increases from surface to interior in GBE specimen. • Low fractal dimension and large TRD size lead to lesser sensitization at surface. • As-received specimen shows high albeit thickness-independent sensitization behavior. • Higher residual strain leads to higher sensitization at surface of deformed sample. - Abstract: Through-thickness microstructural variations and their synergistic influence on degree of sensitization in grain boundary engineering (GBE) treated 304L stainless steel was compared with solution-annealed and marginally cold-worked (~ 3% thickness reduction) specimens. The solution-annealed specimen has shown higher degree of sensitization due to the presence of lower fraction of Σ3 (as well as other low Σ CSL) boundaries and dominant random high-angle boundaries (RHAGBs) connectivity. However, it has shown thickness-independent sensitization behavior as grain boundary character distribution, residual strain and grain boundary network topology towards thickness direction are similar. Owing to higher residual strain, as revealed by local misorientation analysis, the cold-worked specimen showed maximum degree of sensitization (higher than solution-annealed condition) at surface. The GBE specimen has shown much lower degree of sensitization at surface due to higher-fraction of Σ3 boundaries and discontinuous RHAGBs connectivity. However, increasing fraction of RHAGBs and their connectivity, as confirmed by twinning-related domain and fractal analysis, cause more sensitization at interior of this specimen. In spite of this, the degree of sensitization at the interior of GBE specimen is lower than the solution-annealed or cold-worked specimen. Even though the residual strain in GBE specimen is higher than the solution-annealed condition, the lower degree of sensitization indicates that the grain boundary character distribution and grain boundary network topology have higher implication than residual strain in controlling sensitization.

36 MATERIALS SCIENCE↗

MHD shocks in the ISM

Researchers survey shock solutions of a partially ionized gas with a magnetic field. The gas is modeled by interacting neutral, ion, electron and charged grain components. They employ a small neutral-ion chemical network to follow the dissociation and ionization of the major species. Cooling by molecular hydrogen (rotational, vibrational and dissociation), grains and dipole molecules is included. There are three basic types of solutions (C, C asterisk, and J) and some more complicated flows involving combinations of the basic types. The initial preshock conditions cover hydrogen nuclei densities of 1 less than n less than 10(exp 10) cm(-3) and shock velocities of 5 less than v(sub s) less than 60 km/s. The magnetic field is varied over 5 decades and the sensitivity of the results to grain parameters, UV and cosmic ray fluxes is ascertained. The parameter space is quite complicated, but there exist some simple divisions. When the initial ionization fraction is small (chi sub i less than 10(-5)), there is a sharp transition between fully C solutions at low velocity and strong J solutions at high velocity. When the initial ionization fraction is larger, C asterisk and/or very weak J shocks are present at low velocities in addition to the C solutions. The flow again changes to strong J shocks at high velocities. When the ionization fraction is large and the flow is only slightly greater than the bulk Alfven velocity, there is a complicated mixture of C, C asterisk and J solutions.

Chernoff, D. F.↗

Analytical transport network theory to guide the design of 3-D microstructural networks in energy materials: Part 2. Flow with reactions

Here we extend the fully analytical, heuristic “Analytical Transport Network Model” for steady-state, diffusive flow in a 3-D network to account for surface reactions. In the limit of negligible reactions, the model reduces to the conserved flow solution. The extension does not increase the time required to run the model, which in Part 1 was shown to be 0.5–1.5 and 5–6 orders of magnitude faster than electrochemical fin (ECF) theory and finite element analysis (FEA) for conserved flow, respectively. The model is compared to reacting-flow ECF and FEA as well as to experiments and is demonstrated as a potentially useful heuristic for understanding the influence of morphology and topology on reactive-diffusive flow through a 3-D microstructural network.

25 ENERGY STORAGE↗

A Novel Multi-Agent Deep Reinforcement Learning-enabled Distributed Power Allocation Scheme for mmWave Cellular Networks

We consider the power allocation problem over shared spectrum for millimeter-Wave (mmWave) cellular downlink. Existing approaches usually find sub-optimal solutions by solving a non-convex optimization which leads to scalability issues due to centralized control. Therefore, distributed and adaptive approaches are desirable. Recently, model-free Deep Reinforcement Learning (DRL) has achieved success in such wireless resource management tasks. By modeling the radio environment as a Markov Decision Process (MDP) with the base stations (BSs) being the agents, power allocation can be automated at the agent level with comparable throughput performance to conventional centralized schemes. The multi-agent setting presents new challenges as the radio environment is impacted by the joint actions of the agents and is no longer stationary from any individual agent’s perspective. Existing literature bypasses this non-stationarity violation by ignoring it which may cause performance degradation. To tackle this issue, we propose a distributed continuous power allocation scheme based on a modified version of multi-agent Deep Deterministic Policy Gradient (MADDPG) that is tailored for the distributed multiple-agent setting. The proposed scheme employs a centralized-training distributed-execution framework where Q-functions are trained over subsets of BSs while each BS determines its transmit power based only on its own local observation. It admits constant per-BS communication and computation complexity and is thus scalable to large networks. Numerical evaluation shows that the proposed scheme adapts well to a wide range of interference conditions and can achieve comparable or better performance than several state-of-the-art non-learning approaches.

99 GENERAL AND MISCELLANEOUS↗

A demonstration of sub-meter GPS orbit determination and high precision user positioning

It was demonstrated that the submeter GPS (Global Positioning System) orbits can be determined using multiday arc solutions with the current GPS constellation subset visible for about 8 h each day from North America. Submeter orbit accuracy was shown through orbit repeatability and orbit prediction. North American baselines of 1000-2000 km length can be estimated simultaneously with the GPS orbits to an accuracy of better than 1.5 parts in 108 (3 cm over 2000 km distance) with a daily precision of two parts in 108 or better. The most reliable baseline solutions are obtained using the same type of receivers and antennas at each end of the baseline. Baselines greater than 1000 km distance from Florida to sites in the Caribbean region have also been determined with daily precision of 1-4 parts in 108. The Caribbean sites are located well outside the fiducial tracking network and the region of optimal GPS common visibility. Thus, these results further demonstrate the robustness of the multiday arc GPS orbit solutions.

Bertiger, Willy I.↗

A Hierarchical Optimization Method for Electric Vertical Takeoff and Landing Aircraft Network Design

Electric vertical takeoff and landing aircraft (eVTOLs) are expected to serve urban air mobility in a station-to-station configuration, which makes the optimal network design of eVTOL stations a critical question to explore. Existing approaches often face limitations, such as the inability to interact station locations with demand or difficulty in finding the optimal solution for large study regions. Here, this paper first proposes a mathematical model to generate optimal eVTOL station locations while considering associated potential eVTOL demand, and then proposes a heuristic algorithm, Hierarchical Optimization MEthod (HOME), to efficiently solve the model. With a case study of Southern California, HOME was compared to 1) directly solving the original integer linear programming-based network design problem, and 2) employing the widely used genetic algorithm. Results suggest that HOME can find optimal solutions with limited computational resources. The proposed framework powered by HOME provides a computationally efficient way to support urban air mobility planning.

97 MATHEMATICS AND COMPUTING↗