Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Neural Networks (Computer)”

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 811 records · Page 45

Population-based learning of load balancing policies for a distributed computer system

Effective load-balancing policies use dynamic resource information to schedule tasks in a distributed computer system. We present a novel method for automatically learning such policies. At each site in our system, we use a comparator neural network to predict the relative speedup of an incoming task using only the resource-utilization patterns obtained prior to the task's arrival. Outputs of these comparator networks are broadcast periodically over the distributed system, and the resource schedulers at each site use these values to determine the best site for executing an incoming task. The delays incurred in propagating workload information and tasks from one site to another, as well as the dynamic and unpredictable nature of workloads in multiprogrammed multiprocessors, may cause the workload pattern at the time of execution to differ from patterns prevailing at the times of load-index computation and decision making. Our load-balancing policy accommodates this uncertainty by using certain tunable parameters. We present a population-based machine-learning algorithm that adjusts these parameters in order to achieve high average speedups with respect to local execution. Our results show that our load-balancing policy, when combined with the comparator neural network for workload characterization, is effective in exploiting idle resources in a distributed computer system.

Mehra, Pankaj↗

A deep neural network regressor for phase constitution estimation in the high entropy alloy system Al-Co-Cr-Fe-Mn-Nb-Ni

High Entropy Alloys (HEAs) are composed of more than one principal element and constitute a major paradigm in metals research. The HEA space is vast and an exhaustive exploration is improbable. Therefore, a thorough estimation of the phases present in the HEA is of paramount importance for alloy design. Machine Learning presents a feasible and non-expensive method for predicting possible new HEAs on-the-fly. A deep neural network (DNN) model for the elemental system of: Mn, Ni, Fe, Al, Cr, Nb, and Co is developed using a dataset generated by high-throughput computational thermodynamic calculations using Thermo-Calc. The features list used for the neural network is developed based on literature and freely available databases. A feature significance analysis matches the reported HEAs phase constitution trends on elemental properties and further expands it by providing so far-overlooked features. The final regressor has a coefficient of determination ( r 2 ) greater than 0.96 for identifying the most recurrent phases and the functionality is tested by running optimization tasks that simulate those required in alloy design. The DNN developed constitutes an example of an emulator that can be used in fast, real-time materials discovery/design tasks.

36 MATERIALS SCIENCE↗

Uncertainties propagation in the framework of a Rod Ejection Accident modeling based on a multi-physics approach

The control of uncertainties in the field of reactor physics and their propagation in best-estimate modeling are a major issue in safety analysis. In this framework, the CEA develops a methodology to perform multi-physics simulations including uncertainties analysis. The present paper aims to present and apply this methodology for the analysis of an accidental situation such as REA (Rod Ejection Accident). This accident is characterized by a strong interaction between the different areas of the reactor physics (neutronic, fuel thermal and thermal hydraulic). The modeling is performed with CRONOS2 code. The uncertainties analysis has been conducted with the URANIE platform developed by the CEA: For each identified response from the modeling (output) and considering a set of key parameters with their uncertainties (input), a surrogate model in the form of a neural network has been produced. The set of neural networks is then used to carry out a sensitivity analysis which consists on a global variance analysis with the determination of the Sobol indices for all responses. The sensitivity indices are obtained for the input parameters by an approach based on the use of polynomial chaos. The present exercise helped to develop a methodological flow scheme, to consolidate the use of URANIE tool in the framework of parallel calculations. Finally, the use of polynomial chaos allowed computing high order sensitivity indices and thus highlighting and classifying the influence of identified uncertainties on each response of the analysis (single and interaction effects). (authors)

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Co-optimizing water-alternating-carbon dioxide injection projects using a machine learning assisted computational framework

In this report, a robust machine-learning-based computational framework that couples multi-layer neural network (MLNN) proxies and a multi-objective particle swarm optimizer (MOPSO) to design water-alternating-carbon dioxide injection (CO 2 -WAG) projects is presented. The proposed optimization protocol considers various objectives, including oil recovery and CO 2 storage volume. Expert MLNN systems are trained and employed as surrogate models of the high-fidelity compositional simulator in the optimization workflow. When multiple objective functions are considered, two approaches are employed to treat the objectives: the weighted sum method and the Pareto-front-based scheme. A field-scale implementation focusing on tertiary recovery in the Morrow B formation at Farnsworth Unit (FWU) is presented. The developed Pareto-optimal solutions indicate the maximal available oil production can be 1.64 × 10 7 barrels and maximal carbon storage can achieve 2.35 × 10 7 tons. Trade-offs factor is defined to divide the constructed Pareto front into 4 sections with the trade-off factors’ value ranges from 0.35 to 49.9. This work also compares the optimum solution found by the aggregative objective function and the solution repository covered by the Pareto front that considers the physical and operational constraints and reduces uncertainties involved by the multi-objective optimization process. Our comparison indicates multiple solutions exist to satisfy the objective criteria of the WAG design, and these results cannot be found using the traditional weighted sum method. The Pareto front solution can provide more options for project designers, but decisions regarding necessary trade-offs must be made using the solution repository to balance the project economics and CO 2 storage amount.

10 SYNTHETIC FUELS↗

Causally‐Informed Deep Learning to Improve Climate Models and Projections

Abstract Climate models are essential to understand and project climate change, yet long‐standing biases and uncertainties in their projections remain. This is largely associated with the representation of subgrid‐scale processes, particularly clouds and convection. Deep learning can learn these subgrid‐scale processes from computationally expensive storm‐resolving models while retaining many features at a fraction of computational cost. Yet, climate simulations with embedded neural network parameterizations are still challenging and highly depend on the deep learning solution. This is likely associated with spurious non‐physical correlations learned by the neural networks due to the complexity of the physical dynamical system. Here, we show that the combination of causality with deep learning helps removing spurious correlations and optimizing the neural network algorithm. To resolve this, we apply a causal discovery method to unveil causal drivers in the set of input predictors of atmospheric subgrid‐scale processes of a superparameterized climate model in which deep convection is explicitly resolved. The resulting causally‐informed neural networks are coupled to the climate model, hence, replacing the superparameterization and radiation scheme. We show that the climate simulations with causally‐informed neural network parameterizations retain many convection‐related properties and accurately generate the climate of the original high‐resolution climate model, while retaining similar generalization capabilities to unseen climates compared to the non‐causal approach. The combination of causal discovery and deep learning is a new and promising approach that leads to stable and more trustworthy climate simulations and paves the way toward more physically‐based causal deep learning approaches also in other scientific disciplines.

Meteorology & Atmospheric Sciences↗

Fast b -tagging at the high-level trigger of the ATLAS experiment in LHC Run 3

The ATLAS experiment relies on real-time hadronic jet reconstruction and b-tagging to record fully hadronic events containing b-jets. These algorithms require track reconstruction, which is computationally expensive and could overwhelm the high-level-trigger farm, even at the reduced event rate that passes the ATLAS first stage hardware-based trigger. In LHC Run 3, ATLAS has mitigated these computational demands by introducing a fast neural-network-based b-tagger, which acts as a low-precision filter using input from hadronic jets and tracks. It runs after a hardware trigger and before the remaining high-level-trigger reconstruction. This design relies on the negligible cost of neural-network inference as compared to track reconstruction, and the cost reduction from limiting tracking to specific regions of the detector. In the case of Standard Model HH → bb̅bb̅, a key signature relying on b-jet triggers, the filter lowers the input rate to the remaining high-level trigger by a factor of five at the small cost of reducing the overall signal efficiency by roughly 2%.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Integrated Model, Batch and Domain Parallelism in Training Neural Networks

We propose a new integrated method of exploiting model, batch and domain parallelism for the training of deep neural networks (DNNs) on large distributed-memory computers using minibatch stochastic gradient descent (SGD). Our goal is to find an efficient parallelization strategy for a fixed batch size using $P$ processes. Our method is inspired by the communication-avoiding algorithms in numerical linear algebra. We see $P$ processes as logically divided into a $P_r \times P_c$ grid where the $P_r$ dimension is implicitly responsible for model/domain parallelism and the $P_c$ dimension is implicitly responsible for batch parallelism. In practice, the integrated matrix-based parallel algorithm encapsulates these types of parallelism automatically. We analyze the communication complexity and analytically demonstrate that the lowest communication costs are often achieved neither with pure model nor with pure data parallelism. We also show how the domain parallel approach can help in extending the theoretical scaling limit of the typical batch parallel method.

97 MATHEMATICS AND COMPUTING↗

Overview of Algorithms for Using Particle Morphology in Pre-Detonation Nuclear Forensics

A major goal in pre-detonation nuclear forensics is to infer the processing conditions and/or facility type that produced radiological material. This review paper focuses on analyses of particle size, shape, texture (“morphology”) signatures that could provide information on the provenance of interdicted materials. For example, uranium ore concentrates (UOC or yellowcake) include ammonium diuranate (ADU), ammonium uranyl carbonate (AUC), sodium diuranate (SDU), magnesium diuranate (MDU), and others, each prepared using different salts to precipitate U from solution. Once precipitated, UOCs are often dried and calcined to remove adsorbed water. The products can be allowed to react further, forming uranium oxides UO3, U3O8, or UO2 powders, whose surface morphology can be indicative of precipitation and/or calcination conditions used in their production. This review paper describes statistical issues and approaches in using quantitative analyses of measurements such as particle size and shape to infer production conditions. Statistical topics include multivariate t tests (Hotelling’s T 2 ), design of experiments, and several machine learning (ML) options including decision trees, learning vector quantization neural networks, mixture discriminant analysis, and approximate Bayesian computation (ABC). ABC is emphasized as an attractive option to include the effects of model uncertainty in the selected and fitted forward model used for inferring processing conditions.

98 NUCLEAR DISARMAMENT, SAFEGUARDS, AND PHYSICAL P↗

NeuralMie (v1.0): an aerosol optics emulator

The direct interactions of atmospheric aerosols with radiation significantly impact the Earth's climate and weather and are important to represent accurately in simulations of the atmosphere. This work introduces two contributions to enable a more accurate representation of aerosol optics in atmosphere models: (1) NeuralMie, a neural network Mie scattering emulator that can directly compute the bulk optical properties of a diverse range of aerosol populations and is appropriate for use in atmosphere simulations where aerosol optical properties are parameterized, and (2) TAMie, a fast Python-based Mie scattering code based on the Toon and Ackerman (1981) Mie scattering algorithm that can represent both homogeneous and coated particles. TAMie achieves speed and accuracy comparable to established Fortran Mie codes and is used to produce training data for NeuralMie. NeuralMie is highly flexible and can be used for a wide range of particle types, wavelengths, and mixing assumptions. It can represent core-shell scattering and, by directly estimating bulk optical properties, is more efficient than existing Mie code and Mie code emulators while incurring negligible error compared to existing aerosol optics parameterization schemes (0.08 % mean absolute percentage error).

54 ENVIRONMENTAL SCIENCES↗

A survey of visual preprocessing and shape representation techniques

Many recent theories and methods proposed for visual preprocessing and shape representation are summarized. The survey brings together research from the fields of biology, psychology, computer science, electrical engineering, and most recently, neural networks. It was motivated by the need to preprocess images for a sparse distributed memory (SDM), but the techniques presented may also prove useful for applying other associative memories to visual pattern recognition. The material of this survey is divided into three sections: an overview of biological visual processing; methods of preprocessing (extracting parts of shape, texture, motion, and depth); and shape representation and recognition (form invariance, primitives and structural descriptions, and theories of attention).

Olshausen, Bruno A.↗

Space shuttle main engine fault detection using neural networks

A method for on-line Space Shuttle Main Engine (SSME) anomaly detection and fault typing using a feedback neural network is described. The method involves the computation of features representing time-variance of SSME sensor parameters, using historical test case data. The network is trained, using backpropagation, to recognize a set of fault cases. The network is then able to diagnose new fault cases correctly. An essential element of the training technique is the inclusion of randomly generated data along with the real data, in order to span the entire input space of potential non-nominal data.

Bishop, Thomas↗

Intelligent systems technology infrastructure for integrated systems

Significant advances have occurred during the last decade in intelligent systems technologies (a.k.a. knowledge-based systems, KBS) including research, feasibility demonstrations, and technology implementations in operational environments. Evaluation and simulation data obtained to date in real-time operational environments suggest that cost-effective utilization of intelligent systems technologies can be realized for Automated Rendezvous and Capture applications. The successful implementation of these technologies involve a complex system infrastructure integrating the requirements of transportation, vehicle checkout and health management, and communication systems without compromise to systems reliability and performance. The resources that must be invoked to accomplish these tasks include remote ground operations and control, built-in system fault management and control, and intelligent robotics. To ensure long-term evolution and integration of new validated technologies over the lifetime of the vehicle, system interfaces must also be addressed and integrated into the overall system interface requirements. An approach for defining and evaluating the system infrastructures including the testbed currently being used to support the on-going evaluations for the evolutionary Space Station Freedom Data Management System is presented and discussed. Intelligent system technologies discussed include artificial intelligence (real-time replanning and scheduling), high performance computational elements (parallel processors, photonic processors, and neural networks), real-time fault management and control, and system software development tools for rapid prototyping capabilities.

Lum, Henry, Jr.↗

Optoelectronic Terminal-Attractor-Based Associative Memory

Report presents theoretical and experimental study of optically and electronically addressable optical implementation of artificial neural network that performs associative recall. Shows by computer simulation that terminal-attractor-based associative memory can have perfect convergence in associative retrieval and increased storage capacity. Spurious states reduced by exploiting terminal attractors.

Liu, Hua-Kuang↗

Microwave vision for robots

Microwave Vision (MV), a concept originally developed in 1985, could play a significant role in the solution to robotic vision problems. Originally our Microwave Vision concept was based on a pattern matching approach employing computer based stored replica correlation processing. Artificial Neural Network (ANN) processor technology offers an attractive alternative to the correlation processing approach, namely the ability to learn and to adapt to changing environments. This paper describes the Microwave Vision concept, some initial ANN-MV experiments, and the design of an ANN-MV system that has led to a second patent disclosure in the robotic vision field.

Lewandowski, Leon↗

Geometry-aware training of factorized layers in tensor Tucker format

Reducing parameter redundancies in neural network architectures is crucial for achieving feasible computational and memory requirements during train and inference of large networks. Given its easy implementation and flexibility, one promising approach is layer factorization, which reshapes weight tensors into a matrix format and parameterizes it as the product of two rank-r matrices. However, this family of approaches often requires an initial full-model warm-up phase, prior knowledge of a feasible rank, and it is sensitive to parameter initialization.In this work, we introduce a novel approach to train the factors of a Tucker decomposition of the weight tensors. Our training proposal proves to be optimal in locally approximating the original unfactorized dynamics and stable for the initialization. Furthermore, the rank of each mode is dynamically updated during training.We provide a theoretical analysis of the algorithm, showing convergence, approximation and local descent guarantees. The method's performance is further illustrated through a variety of experiments, showing remarkable training compression rates and comparable or even better performance than the full baseline and alternative layer factorization strategies.

Zangrando, Emanuele [Gran Sasso Science Institute ↗

End-to-End Optimization for Battery Materials and Molecules by Combining Graph Neural Networks and Reinforcement Learning

The National Renewable Energy Laboratory (NREL), together with the Colorado School of Mines (CSM) and Colorado State University (CSU), has developed a machine learning-enhanced approach to the design of new battery materials. Currently, such materials are designed in part via numerous expensive high-fidelity computational simulations that predict the performance of a given composition. Even with computational screening tools, the vast landscape of possible molecular or crystal structures exceeds current and future computational capacity. Improving the efficiency by which new materials can be optimized will therefore disrupt the cost, risk, and time required to bring new energy solutions to the marketplace. Predicting the properties of an organic molecule or periodic crystalline material given its structure has grown increasingly common. These approaches leverage large-scale computational and experimental databases and ML approaches such as graph neural networks. The inverse design problem of finding a material that possesses desired properties is substantially more challenging, since enumerating all valid material structures is not feasible. In this project, we leveraged recent success in reinforcement learning to efficiently navigate this high-dimensional search space. Just as algorithms can find the optimal chess moves from nearly limitless options, we train an approach to evolve a simple starting structure into a complex structure that possess the desired properties. Our solution has been demonstrated by applying it to two related design application tasks for short- and long-term energy storage, respectively: (1) the design of solid-state ion conductors and (2) the design of organic redox-active materials. The project has resulted an open-source software library for material design, documented examples of applying the library to both organic and inorganic material optimization, and peer-reviewed publications detailing the data, computational models, and resulting candidate materials.

25 ENERGY STORAGE↗

Differentiable hybrid neural network approach for enhancing reactor dynamics simulations

Reactor dynamics simulations provide essential insights into the time-dependent behavior of nuclear reactors under various operating conditions. However, high-fidelity simulations can be computationally intensive, requiring significant computational resources. Here, to address this challenge, this study employs a differentiable hybrid model that utilizes neural networks as a corrector to enhance the performance of a low-fidelity simulation, aligning its predictions with those of a high-fidelity simulation. Low-fidelity and high-fidelity simulations were obtained by adjusting the mesh size in the System Dynamics Analysis Tool. The differentiable hybrid model was trained in two approaches: time-step-wise and sequence-wise. It was then applied to simulate various transients in a molten salt reactor. Its performance was evaluated by comparing its responses to transients against those of the high-fidelity simulation. An additional approach was performed using a data-driven model to correct the low-fidelity simulation. In comparison, the differentiable hybrid model showed significant improvements in transient prediction, effectively addressing the limitations of the low-fidelity simulations. The results highlighted the robustness of the differentiable hybrid model in both training approaches. It delivered simulations that were at least 3.8 times faster than high-fidelity models. In the time-step-wise approach, it achieved at least a 39% improvement in accuracy. In the sequence-wise approach, it showed at least an 81% accuracy improvement over the full transient. This approach offers a promising path for improving computational efficiency without compromising accuracy in nuclear reactor simulations, making it suitable for real-time digital twin applications.

42 - ENGINEERING↗

Reducing Communication in Graph Neural Network Training

Graph Neural Networks (GNNs) are powerful and flexible neural networks that use the naturally sparse connectivity information of the data. GNNs represent this connectivity as sparse matrices, which have lower arithmetic intensity and thus higher communication costs compared to dense matrices, making GNNs harder to scale to high concurrencies than convolutional or fully-connected neural networks. Here, we introduce a family of parallel algorithms for training GNNs and show that they can asymptotically reduce communication compared to previous parallel GNN training methods. We implement these algorithms, which are based on 1D, 1. 5D, 2D, and 3D sparse-dense matrix multiplication, using torch.distributed on GPU-equipped clusters. Our algorithms optimize communication across the full GNN training pipeline. We train GNNs on over a hundred GPUs on multiple datasets, including a protein network with over a billion edges.

97 MATHEMATICS AND COMPUTING↗