Search NASA⌕ Search

SEARCH · Search NASA

Results for “Network Models”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 433 records · Page 24

Understanding and Verifying Neural Networks

Deep Neural Networks (DNNs) have gained immense popularity in recent times and have widespread use in applications such as image classification, sentiment analysis, speech recognition and also in safety-critical applications such as autonomous driving. However, they suffer limitations such as lack of explainability and robustness which raise safety and security concerns in their usage. Further, the complex structure and large input spaces of DNNs act as an impediment to thorough verification and testing. The SafeDNN project at the Robust Software Engineering (RSE) group at NASA aims at exploring techniques to ensure that systems that use deep neural networks are safe, robust and interpretable. In this talk, I will be presenting our technique Prophecy that automatically infers formal properties of deep neural network models. The tool extracts patterns based on neuron activations as preconditions that imply certain desirable output properties of the model. I would be highlighting case studies that use Prophecy in obtaining explanations for network decisions, understanding correct and incorrect behavior, providing formal guarantees wrt safety and robustness, and debugging neural network models. We have applied the tool on image classification networks, neural network controllers providing turn advisories in unmanned aircrafts, regression models used for autonomous center-line tracking in aircrafts and neural network object detectors

Deep Neural Networks↗

Neural Networks for Modeling and Control of Particle Accelerators

Particle accelerators are host to myriad nonlinear and complex physical phenomena. They often involve a multitude of interacting systems, are subject to tight performance demands, and should be able to run for extended periods of time with minimal interruptions. Often times, traditional control techniques cannot fully meet these requirements. One promising avenue is to introduce machine learning and sophisticated control techniques inspired by artificial intelligence, particularly in light of recent theoretical and practical advances in these fields. Within machine learning and artificial intelligence, neural networks are particularly well-suited to modeling, control, and diagnostic analysis of complex, nonlinear, and time-varying systems, as well as systems with large parameter spaces. Consequently, the use of neural network-based modeling and control techniques could be of significant benefit to particle accelerators. For the same reasons, particle accelerators are also ideal test-beds for these techniques. Many early attempts to apply neural networks to particle accelerators yielded mixed results due to the relative immaturity of the technology for such tasks. The purpose of this paper is to re-introduce neural networks to the particle accelerator community and report on some work in neural network control that is being conducted as part of a dedicated collaboration between Fermilab and Colorado State University (CSU). We describe some of the challenges of particle accelerator control, highlight recent advances in neural network techniques, discuss some promising avenues for incorporating neural networks into particle accelerator control systems, and describe a neural network-based control system that is being developed for resonance control of an RF electron gun at the Fermilab Accelerator Science and Technology (FAST) facility, including initial experimental results from a benchmark controller.

43 PARTICLE ACCELERATORS↗

Multiscale topology optimization using neural network surrogate models

In this study, we are concerned with optimization of macroscale elastic structures that are designed utilizing spatially varying microscale metamaterials. The macroscale optimization is accomplished using gradient-based nonlinear topological optimization. But instead of using density as the optimization decision variable, the decision variables are the multiple parameters that define the local microscale metamaterial. This is accomplished using single layer feedforward Gaussian basis function networks as a surrogate models of the elastic response of the microscale metamaterial. The surrogate models are trained using highly resolved continuum finite element simulations of the microscale metamaterials and hence are significantly more accurate than analytical models e.g. classical beam theory. Because the derivative of the surrogate model is important for sensitivity analysis of the macroscale topology optimization, a neural network training procedure based on the Sobolev norm is described. Since the SIMP method is not appropriate for spatially varying lattices, an alternative method is developed to enable creation of void regions. Lastly, the efficacy of this approach is demonstrated via several examples in which the optimal graded metamaterial outperforms a traditional solid structure.

42 ENGINEERING↗

Localization-length exponent in two models of quantum Hall plateau transitions

Motivated by the recent numerical studies on the Chalker-Coddington network model that found a larger-than-expected critical exponent of the localization length characterizing the integer quantum Hall plateau transitions, we revisited the exponent calculation in the continuum model and in the lattice model, both projected to the lowest Landau level or subband. Combining scaling results with or without the corrections of an irrelevant length scale, we obtain ν = 2.48 ± 0.02, which is larger but still consistent with the earlier results in the two models, unlike what was found recently in the network model. Here, the scaling of the total number of conducting states, as determined by the Chern number calculation, is accompanied by an effective irrelevant length scale exponent y = 4.3 in the lattice model, indicating that the irrelevant perturbations are insignificant in the topology number calculation.

2-dimensional systems↗

DETECTING FIRE WITH MACHINE LEARNING-ENABLED VISUAL MONITORING FOR NUCLEAR POWER PLANT ENVIRONMENTS

Nuclear power plants are experiencing significant cost challenges to remain competitive with other energy-generation utilities. Unlike other industries, the cost of operation and maintenance activities is mostly attributed to workforce costs. To mitigate this, nuclear power plant stakeholders are increasingly interested in the development and deployment of machine learning methods to potentially automate or augment manually intensive tasks to reduce costs, especially for monitoring activities. One monitoring function that is visually demanding and that can occur frequently to meet the requirements of a fire protection program is visually monitoring an area for fire occurrence. Currently, fire watch activities consist of a worker physically stationed at a given location with the sole responsibility of observing a given area to ensure a fire is detected and mitigated promptly. This effort focused on the development and evaluation of a suitable deep convolutional neural network to classify individual video frames at a sub-second frequency for the occurrence of “fire” and “no fire” in varying industrial environments similar to nuclear power plants. It is believed that a trained neural network model could be integrated with existing facility video surveillance camera feeds to generate alerts when fire inferences occur in individual frames captured at sub-second temporal resolutions. Extensive effort was dedicated to identifying and curating suitable imagery training data representing varying environments and scene settings with and without flame features to maximize generalization in nuclear power plant environments. The data collection effort resulted in the aggregation of a large, labeled image library exceeding 12,000 images to support model training for diverse industrial environments. A deep neural network model incorporating parallel multi-scale capabilities was developed and trained to support accurate image-based detection of flame incidents of varying sizes and spectral feature properties within heterogeneous scenes. Analysis results show that the trained model can achieve high inference accuracy despite heterogeneous scene environments and components. Testing accuracy exceeded 95.0 percent with very low false positive and false negative inferences.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Neural networks for modeling and control of particle accelerators

Charged particle accelerators support a wide variety of scientific, industrial, and medical applications. They range in scale and complexity from systems with just a few components for beam acceleration and manipulation, to large scientific user facilities that span many kilometers and have hundreds-to-thousands of individually-controllable components. Specific operational requirements must be met by adjusting the many controllable variables of the accelerator. Meeting these requirements can be challenging, both in terms of the ability to achieve specific beam quality metrics in a reliable fashion and in terms of the time needed to set up and maintain the optimal operating conditions. One avenue toward addressing this challenge is to incorporate techniques from the fields of machine learning (ML) and artificial intelligence (AI) into the way particle accelerators are modeled and controlled. While many promising approaches within AI/ML could be used for particle accelerators, this dissertation focuses on approaches based on neural networks. Neural networks are particularly well-suited to modeling, control, and diagnostic analysis of nonlinear systems, as well as systems with large parameter spaces. They are also very appealing for their ability to process high-dimensional data types, such as images and time series (both of which are ubiquitous in particle accelerators). In this work, key studies that demonstrated the potential utility of modern neural network-based approaches to modeling and control of particle accelerators are presented. The context for this work is important: at the start of this work in 2012, there was little interest in AI/ML in the particle accelerator community, and many of the advances in neural networks and deep learning that enabled its present success had not yet been made at that time. As such, this work was both an exploration of possible application areas and a generator of proof-of-concept demonstrations in these areas.

43 PARTICLE ACCELERATORS↗

New Techniques and Data for Understanding the Global Soil Respiration Flux

Abstract Soil respiration ( R s ; the soil surface‐to‐atmosphere CO 2 flux) has been measured in the field for decades, but only recently have we begun to assemble and leverage these small‐scale but extensive data. Recently, Zhao et al. (2017, https://doi.org/10.1002/2016ef000480 ) applied a novel artificial neural network model to the problem of estimating the global R s flux and understanding its variations between regions and biomes. Their results point to a convergence in estimates of global R s , and the power of leveraging the long record of observed R s in global ecosystems, but also to uncertainties about soils' response to climate change. It will take a combination of long‐term studies, data syntheses, modeling intercomparisons, and probably a new generation of sampling networks and experiments to fully resolve these questions.

Bond‐Lamberty, Ben↗

Interim Service ISDN Satellite (ISIS) simulator development for advanced satellite designs and experiments

The simulation development associated with the network models of both the Interim Service Integrated Services Digital Network (ISDN) Satellite (ISIS) and the Full Service ISDN Satellite (FSIS) architectures is documented. The ISIS Network Model design represents satellite systems like the Advanced Communications Technology Satellite (ACTS) orbiting switch. The FSIS architecture, the ultimate aim of this element of the Satellite Communications Applications Research (SCAR) Program, moves all control and switching functions on-board the next generation ISDN communications satellite. The technical and operational parameters for the advanced ISDN communications satellite design will be obtained from the simulation of ISIS and FSIS engineering software models for their major subsystems. Discrete event simulation experiments will be performed with these models using various traffic scenarios, design parameters, and operational procedures. The data from these simulations will be used to determine the engineering parameters for the advanced ISDN communications satellite.

Pepin, Gerard R.↗

Multi-Node Thermal System Model for Lithium-Ion Battery Packs: Preprint

Temperature is one of the main factors that controls the degradation in lithium ion batteries. Accurate knowledge and control of cell temperatures in a pack helps the battery management system (BMS) to maximize cell utilization and ensure pack safety and service life. In a pack with arrays of cells, a cells temperature is not only affected by its own thermal characteristics but also by its neighbors, the cooling system and pack configuration, which increase the noise level and the complexity of cell temperatures prediction. This work proposes to model lithium ion packs thermal behavior using a multi-node thermal network model, which predicts the cell temperatures by zones. The model was parametrized and validated using commercial lithium-ion battery packs. neighbors, the cooling system and pack configuration, which increase the noise level and the complexity of cell temperatures prediction. This work proposes to model lithium ion packs thermal behavior using a multi-node thermal network model, which predicts the cell temperatures by zones. The model was parametrized and validated using commercial lithium-ion battery packs.

25 ENERGY STORAGE↗

Hybrid interconnection structures for real-time parallel processing

The use of hybrid interconnection structures that combine link connections and bus connections for real-time parallel processing is discussed. Idealistic parallel computation models for two real-time computing applications are described with attention given to a tightly coupled network model for object tracking and a network model for image processing. Consideration is given to the following different interconnection structures: the crossbar, the hypercube, the circular linked array, and the bus array.

Kim, K. H.↗

Performance Evaluation of Vertical Federated Machine Learning Against Adversarial Threats on Wide-Area Control System: Preprint

Federated machine learning (FL) is gaining significant popularity to develop cybersecurity solutions in power grids because of its advanced capability to support decentralized data handing at local devices, its privacy preservation, and its low-bandwidth requirement. However, the evolving adversarial machine learning (AML) threats raise significant concerns for the cybersecurity of FL architectures. The FL-based split neural network (SplitNN) achieves high performance through the decentralized training of local neural network models while preserving data privacy across multiple entities. In this paper, we propose a methodology for evaluating the performance of a vertical FLbased anomaly detector against different types of AML attacks, including denial-of-service attacks, adversarial data injection attacks, and replay attacks on the trained local models deployed in the grid network. For a case study, we consider the modified IEEE 13-bus system, and we develop SplitNN-based binary and multiclass classification models to detect, locate, and identify different types of data integrity attacks on the volt-watt control with two pooling layers: maximum pooling and AvgPool. Our experimental results, computed through performance metrics, reveal that the severity of these AML attacks varies with the integrated pooling mechanism, the type of classification model, and the nature of the cyberattack. Further, the AML attacks negatively impacted the prediction time per sample for the pretrained SplitNN during the online testing.

adversarial threats↗

Fifth-Generation District Heating and Cooling Substations: Demand Response with Artificial Neural Network-Based Model Predictive Control

District heating and cooling (DHC) is considered one of the most sustainable technologies to meet the heating and cooling demands of buildings in urban areas. The fifth-generation district heating and cooling (5GDHC) concept, often referred to as ambient loops, is a novel solution emerging in Europe and has become a widely discussed topic in current energy system research. 5GDHC systems operate at a temperature close to the ground and include electrically driven heat pumps and associated thermal energy storage in a building-sited energy transfer station (ETS) to satisfy user comfort. This work presents new strategies for improving the operation of these energy transfer stations by means of a model predictive control (MPC) method based on recurrent artificial neural networks. The results show that, under simple time-of-use utility rates, the advanced controller outperforms a rule-based controller for smart charging of the domestic hot water (DHW) thermal energy storage under specific boundary conditions. By exploiting the available thermal energy storage capacity, the MPC controller is capable of shifting up to 14% of the electricity consumption of the ETS from on-peak to off-peak hours. Therefore, the advanced control implemented in 5GDHC networks promotes coupling between the thermal and the electric sector, producing flexibility on the electric grid.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Simulator design for advanced ISDN satellite design and experiments

This simulation design task completion report documents the simulation techniques associated with the network models of both the Interim Service ISDN (integrated services digital network) Satellite (ISIS) and the Full Service ISDN Satellite (FSIS) architectures. The ISIS network model design represents satellite systems like the Advanced Communication Technology Satellite (ACTS) orbiting switch. The FSIS architecture, the ultimate aim of this element of the Satellite Communications Applications Research (SCAR) program, moves all control and switching functions on-board the next generation ISDN communication satellite. The technical and operational parameters for the advanced ISDN communications satellite design will be obtained from the simulation of ISIS and FSIS engineering software models for their major subsystems. Discrete events simulation experiments will be performed with these models using various traffic scenarios, design parameters and operational procedures. The data from these simulations will be used to determine the engineering parameters for the advanced ISDN communications satellite.

Pepin, Gerald R.↗

Spiral waves in externally excited neuronal network: Solvable model with a monotonically differentiable magnetic flux

Spiral waves are particular spatiotemporal patterns connected to specific phase singularities representing topological wave dislocations or nodes of zero amplitude, witnessed in a wide range of complex systems such as neuronal networks. The appearance of these waves is linked to the network structure as well as the diffusion dynamics of its blocks. We report a novel form of the Hindmarsh-Rose neuron model utilized as a square neuronal network, showing the remarkable multistructure of dynamical patterns ranging from characteristic spiral wave domains of spatiotemporal phase coherence to regions of hyperchaos. The proposed model comprises a hyperbolic memductance function as the monotone differentiable magnetic flux. Hindmarsh-Rose neurons with an external electromagnetic excitation are considered in three different cases: no excitation, periodic excitation, and quasiperiodic excitation. We performed an extensive study of the neuronal dynamics including calculation of equilibrium points, bifurcation analysis, and Lyapunov spectrum. We have found the property of antimonotonicity in bifurcation scenarios with no excitation or periodic excitation and identified wide regions of hyperchaos in the case of quasiperiodic excitation. Furthermore, the formation and elimination of the spiral waves in each case of external excitation with respect to stimuli parameters are investigated. We have identified novel forms of Hindmarsh-Rose bursting dynamics. Our findings reveal multipartite spiral wave formations and symmetry breaking spatiotemporal dynamics of the neuronal model that may find broad practical applications. © 2019 Author(s).

60 APPLIED LIFE SCIENCES↗

CANA v1.0.0: efficient quantification of canalization in automata networks

The biomolecular networks underpinning cell function exhibit canalization, or the buffering of fluctuations required to function in a noisy environment. We present a new major release of $\tt{CANA}$, v1.0.0, an open-source Python package for understanding canalization in automata network models, discrete dynamical systems in which activation of biomolecular entities (e.g. transcription of genes) is modeled as the activity of coupled automata. One understudied putative mechanism for canalization is the functional equivalence of biomolecular regulators (e.g. among the transcription factors for a gene). We study this mechanism using the theory of symmetry in discrete functions. We present a new exact method, $\tt{schematodes}$, for finding maximal symmetry groups among the inputs to discrete functions, and integrate it into $\tt{CANA}$. The $\tt{schematodes}$ method substantially outperforms the inexact method of previous $\tt{CANA}$ versions both in speed and accuracy. We apply $\tt{CANA}$ v1.0.0 to study symmetry in 74 experimentally supported automata network models from the Cell Collective (CC) repository. The symmetry distribution is significantly different in the CC than in random automata with the same in-degree (connectivity) and bias (average output) (Kolmogorov–Smirnov test, P ≪ .001). Its spread is much wider than in a null model (IQR 0.31 versus IQR 0.20 with equal medians), demonstrating that the CC is enriched in functions with extreme symmetry or asymmetry.

Boolean networks↗

Measurement and modelling of reactive transport in geological barriers for nuclear waste containment

Compacted clays are considered as excellent candidates for barriers to radionuclide transport in future repositories for nuclear waste due to their very low hydraulic permeability. Diffusion is the dominant transport mechanism, controlled by a nano-scale pore system. Assessment of the clays' long-term containment function requires adequate modelling of such pore systems and their evolution. Existing characterisation techniques do not provide complete pore space information for effective modelling, such as pore and throat size distributions and connectivity. Special network models for reactive transport are proposed here using the complimentary character of the pore space and the solid phase. Here, this balances the insufficient characterisation information and provides the means for future mechanical–physical–chemical coupling. The anisotropy and heterogeneity of clays is represented using different length parameters and percentage of pores in different directions. Resulting networks are described as mathematical graphs with efficient discrete calculus formulation of transport. Opalinus Clay (OPA) is chosen as an example. Experimental data for the tritiated water (HTO) and U(VI) diffusion through OPA are presented. Calculated diffusion coefficients of HTO and uranium species are within the ranges of the experimentally determined data in different clay directions. This verifies the proposed pore network model and validates that uranium complexes are diffusing as neutral species in OPA. In the case of U(VI) diffusion the method is extended to account for sorption and convection. Finally, rather than changing pore radii by coarse grained mathematical formula, physical sorption is simulated in each pore, which is more accurate and realistic.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Hydropower Optimization Using Artificial Neural Network Surrogate Models of a High-Fidelity Hydrodynamics and Water Quality Model

Hydropower operations optimization subject to environmental constraints is limited by challenges associated with dimensionality and spatial and temporal resolution. The need for high-fidelity hydrodynamic and water quality models within optimization schemes is driven by improved computational capabilities, increased requirements to meet specific points of compliance with greater resolution, and the need to optimize operations of not just single reservoirs but systems of reservoirs. This study describes an important advancement for computing hourly power generation schemes for a hydropower reservoir using high-fidelity models, surrogate modeling techniques, and optimization methods. The predictive power of the high-fidelity hydrodynamic and water quality model CE-QUAL-W2 is successfully emulated by an artificial neural network, then integrated into a genetic algorithm optimization approach to maximize hydropower generation subject to constraints on dam operations and water quality. This methodology is applied to a multipurpose reservoir near Nashville, Tennessee, USA. The model successfully reproduced high-fidelity reservoir information while enabling 6.8% and 6.6% increases in hydropower production value relative to actual operations for dissolved oxygen (DO) limits of 5 and 6 mg/L, respectively, while witnessing an expected decrease in power generation at more restrictive DO constraints. Exploration of simultaneous temperature and DO constraints revealed capability to address multiple water quality constraints at specified locations. Here, the reduced computational requirements of the new modeling approach demonstrated an ability to provide decision support for reservoir operations scheduling while maintaining high-fidelity hydrodynamic and water quality information as part of the optimization decision support routines.

13 HYDRO ENERGY↗

Microstructure from joint analysis of experimental data and ab initio interactions: Hydrogenated amorphous silicon

A study of the formation of voids and molecular hydrogen in hydrogenated amorphous silicon is presented based upon a hybrid approach that involves inversion of experimental nuclear magnetic resonance data in conjunction with ab initio total-energy relaxations in an augmented solution space. The novelty of this approach is that the voids and molecular hydrogen appear naturally in the model networks unlike conventional approaches, where voids are created artificially by removing silicon atoms from the networks. Two representative models with 16 and 18 at. % of hydrogen are studied in this work. The result shows that the microstructure of the a-Si:H network consists of several microvoids and few molecular hydrogen for concentration above 15 at. % H. The microvoids are highly irregular in shape and size, and have a linear dimension of 5–7 Å. The internal surface of a microvoid is found to be decorated with 4–9 hydrogen atoms in the form of monohydride Si–H configurations as observed in nuclear magnetic resonance experiments. The microstructure consists of (0.9–1.4)% hydrogen molecules of total hydrogen in the networks. These observations are consistent with the outcome of infrared spectroscopy, nuclear magnetic resonance, and calorimetry experiments.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗