SEARCH · Engineering Papers
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Algorithms for the solution of spatio-temporal recognition and classification problems are known to require intense image data computation.
As we transition towards a decarbonized economy, the integration of variable renewable energy resources and new demands (e.g., electric vehicles, heat pumps) into the electricity grid places unprecedented pressure on grid operators to effectively anticipate and manage peak load. In this context, machine learning algorithms are proving to be indispensable for accurate short-term load forecasting, a crucial task to address these challenges. This study benchmarks 6 machine learning algorithms, including three neural networks and three tree-based algorithms, across various levels of spatial aggregation and time horizons (1, 4, 8, 24, and 48 h). The central contribution of this work is the comparison and analysis of load forecasting models not only based on statistical metrics, but also based on a novel error metric, which evaluates the cost implications of forecast errors for power system stakeholders. Results show that tree-based models outperform neural networks, based on statistical metrics, and yield less skewed error distributions for most spatial scales. However, through the lens of the novel error metric, neural networks are the more competitive choice, especially for forecast horizons that exceed 8 h. The study concludes with actionable recommendations to grid operators and highlights the need for the development of error metrics that link forecasting accuracy to operational costs. To promote transparency and open science, the datasets and Python code are open-sourced via a supplementary repository.
The JPL Center for Space Microelectronics Technology (CSMT) is actively pursuing research in the neural network theory, algorithms, and electronics as well as optoelectronic neural net hardware implementations, to explore the strengths and application potential for a variety of NASA, DoD, as well as commercial application problems, where conventional computing techniques are extremely time-consuming, cumbersome, or simply non-existent. An overview of the JPL electronic neural network hardware development activities and some of the striking applications of the JPL electronic neuroprocessors are presented.
Deep reinforcement learning (DRL) has empowered a variety of artificial intelligence fields, including pattern recognition, robotics, recommendation-systems, and gaming. Similarly, graph neural networks (GNN) have also demonstrated their superior performance in supervised learning for graph-structured data. In recent times, the fusion of GNN with DRL for graph-structured environments has attracted a lot of attention. Here, this paper provides a comprehensive review of these hybrid works. These works can be classified into two categories: (1) algorithmic enhancement, where DRL and GNN complement each other for better utility; (2) application-specific enhancement, where DRL and GNN support each other. This fusion effectively addresses various complex problems in engineering and life sciences. Based on the review, we further analyze the applicability and benefits of fusing these two domains, especially in terms of increasing generalizability and reducing computational complexity. Finally, the key challenges in integrating DRL and GNN, and potential future research directions are highlighted, which will be of interest to the broader machine learning community.
We describe the PACE-MAPP algorithm that simultaneously retrieves aerosol and ocean optical parameters using multiangle and multi-channel polarimeter measurements from the SPEXone, Hyper-Angular Rainbow Polarimeter 2 (HARP2), and Ocean Color Instrument (OCI) instruments onboard the NASA Plankton, Aerosol, Cloud, ocean Ecosystem (PACE) observing system PACE-MAPP is adapted from the Research Scanning Polarimeter (RSP) Microphysical Aerosol Properties from Polarimetry (RSP-MAPP) algorithm. A key feature of the MAPP family of algorithms is the use of a coupled vector radiative transfer model such that the atmosphere and ocean are always considered together as one system. Consequently, conservation of energy ensures that negative water-leaving radiances do not occur. PACE-MAPP uses optimal estimation to simultaneously characterize the optical and microphysical properties of aerosol and ocean constituents, find the optimal solution, and reliably account for the uncertainties of each parameter. This coupled approach, together with multiangle, multi-channel polarimeter measurements, will enable retrievals of aerosol and water properties across the Earth’s oceans. The PACE-MAPP algorithm provides aerosol and ocean products for both the open ocean and coastal areas and is designed to be accurate, modular, and efficient by using fast neural networks that replace the time-consuming vector radiative transfer calculations. We provide an overview of the PACE-MAPP framework and also describe its modular components including its aerosol and hydrosol models, ocean bio-optical models, and thin cirrus model.
Response functions are a key quantity to describe the near-equilibrium dynamics of strongly interacting many-body systems. Recent techniques that attempt to overcome the challenges of calculating these ab initio have employed expansions in terms of orthogonal polynomials. We employ a neural network prediction algorithm to reconstruct a response function 𝑆(𝜔) defined over a range in frequencies 𝜔. Here, we represent the calculated response function as a truncated Chebyshev series whose coefficients can be optimized to reduce the representation error. We compare the quality of response functions obtained using coefficients calculated using a neural network (NN) algorithm with those computed using the Gaussian integral transform (GIT) method. In the regime where only a small number of terms in the Chebyshev series are retained, we find that the NN scheme outperforms the GIT method.
We propose AGS-GNN, a novel attribute-guided sampling algorithm for Graph Neural Networks (GNNs) that exploits node features and connectivity structure of a graph while simultaneously adapting for both homophily and heterophily in graphs. (In homophilic graphs vertices of the same class are more likely to be connected, and vertices of different classes tend to be linked in heterophilic graphs.) While GNNs have been successfully applied to homophilic graphs, their application to heterophilic graphs remains challenging. The best-performing GNNs for heterophilic graphs do not fit the sampling paradigm, suffer high computational costs, and are not inductive. We employ samplers based on feature-similarity and feature-diversity to select subsets of neighbors for a node, and adaptively capture information from homophilic and heterophilic neighborhoods using dual channels. Currently, AGS-GNN is the only algorithm that we know of that explicitly controls homophily in the sampled subgraph through similar and diverse neighborhood samples. For diverse neighborhood sampling, we employ submodularity, which was not used in this context prior to our work. The sampling distribution is pre-computed and highly parallel, achieving the desired scalability. Using an extensive dataset consisting of 35 small (<=100K nodes) and large (>100K nodes) homophilic and heterophilic graphs, we demonstrate the superiority of AGS-GNN compare to the current approaches in the literature. AGS-GNN achieves comparable test accuracy to the best-performing heterophilic GNNs, even outperforming methods using the entire graph for node classification. AGS-GNN also converges faster compared to methods that sample neighborhoods randomly, and can be incorporated into existing GNN models that employ node or graph sampling.
The research is concerned with developing a new approach to enhancing fault tolerance of flight control systems. The original motivation for fault-tolerant control comes from the need for safe operation of control elements (e.g. actuators) in the event of hardware failures in high reliability systems. One such example is modem space vehicle subjected to actuator/sensor impairments. A major task in flight control is to revise the control policy to balance impairment detectability and to achieve sufficient robustness. This involves careful selection of types and parameters of the controllers and the impairment detecting filters used. It also involves a decision, upon the identification of some failures, on whether and how a control reconfiguration should take place in order to maintain a certain system performance level. In this project new flight dynamic model under uncertain flight conditions is considered, in which the effects of both ramp and jump faults are reflected. Stabilization algorithms based on neural network and adaptive method are derived. The control algorithms are shown to be effective in dealing with uncertain dynamics due to external disturbances and unpredictable faults. The overall strategy is easy to set up and the computation involved is much less as compared with other strategies. Computer simulation software is developed. A serious of simulation studies have been conducted with varying flight conditions.
Data mining is widely used to mine business, engineering, and scientific data. Data mining uses pattern based queries, searches, or other analyses of one or more electronic databases/datasets in order to discover or locate a predictive pattern or anomaly indicative of system failure, criminal or terrorist activity, etc. There are various algorithms, techniques and methods used to mine data; including neural networks, genetic algorithms, decision trees, nearest neighbor method, rule induction association analysis, slice and dice, segmentation, and clustering. These algorithms, techniques and methods used to detect patterns in a dataset, have been used in the development of numerous open source and commercially available products and technology for data mining. Data mining is best realized when latent information in a large quantity of data stored is discovered. No one technique solves all data mining problems; challenges are to select algorithms or methods appropriate to strengthen data/text mining and trending within given datasets. In recent years, throughout industry, academia and government agencies, thousands of data systems have been designed and tailored to serve specific engineering and business needs. Many of these systems use databases with relational algebra and structured query language to categorize and retrieve data. In these systems, data analyses are limited and require prior explicit knowledge of metadata and database relations; lacking exploratory data mining and discoveries of latent information. This presentation introduces MatLab(R) (MATrix LABoratory), an engineering and scientific data analyses tool to perform data mining. MatLab was originally intended to perform purely numerical calculations (a glorified calculator). Now, in addition to having hundreds of mathematical functions, it is a programming language with hundreds built in standard functions and numerous available toolboxes. MatLab's ease of data processing, visualization and its enormous availability of built in functionalities and toolboxes make it suitable to perform numerical computations and simulations as well as a data mining tool. Engineers and scientists can take advantage of the readily available functions/toolboxes to gain wider insight in their perspective data mining experiments.
Data mining is widely used to mine business, engineering, and scientific data. Data mining uses pattern based queries, searches, or other analyses of one or more electronic databases/datasets in order to discover or locate a predictive pattern or anomaly indicative of system failure, criminal or terrorist activity, etc. There are various algorithms, techniques and methods used to mine data; including neural networks, genetic algorithms, decision trees, nearest neighbor method, rule induction association analysis, slice and dice, segmentation, and clustering. These algorithms, techniques and methods used to detect patterns in a dataset, have been used in the development of numerous open source and commercially available products and technology for data mining. Data mining is best realized when latent information in a large quantity of data stored is discovered. No one technique solves all data mining problems; challenges are to select algorithms or methods appropriate to strengthen data/text mining and trending within given datasets. In recent years, throughout industry, academia and government agencies, thousands of data systems have been designed and tailored to serve specific engineering and business needs. Many of these systems use databases with relational algebra and structured query language to categorize and retrieve data. In these systems, data analyses are limited and require prior explicit knowledge of metadata and database relations; lacking exploratory data mining and discoveries of latent information. This presentation introduces MatLab(TradeMark)(MATrix LABoratory), an engineering and scientific data analyses tool to perform data mining. MatLab was originally intended to perform purely numerical calculations (a glorified calculator). Now, in addition to having hundreds of mathematical functions, it is a programming language with hundreds built in standard functions and numerous available toolboxes. MatLab's ease of data processing, visualization and its enormous availability of built in functionalities and toolboxes make it suitable to perform numerical computations and simulations as well as a data mining tool. Engineers and scientists can take advantage of the readily available functions/toolboxes to gain wider insight in their perspective data mining experiments.
Earthquake location algorithms typically require travel time calculation. Doing this calculation in 3D, despite advances in algorithm efficiency and computational power, can still be prohibitively expensive in terms of resources and storage. Implementation of high-resolution 3D models in routine earthquake location would be a significant step forward in most of the world. Machine learning algorithms have potential to act as substitutes for travel time calculation algorithms or stored travel time tables. We investigate EikoNet - a physics informed neural network machine learning model that estimates travel times very quickly and comes with negligible memory-overhead. Specifically, we apply EikoNet to the Wasatch Fault Community Velocity Model (WFCVM), a highly detailed and complex 3D velocity model of the Salt Lake City, UT region. While routine locations in the area and studies of the 2020 Magna, UT earthquake sequence used a 1D velocity model, a 3D model may help better our understanding the structure of the major fault in the region. Our primary goal was to test the speed, memory requirements, and accuracy of EikoNet compared to a reference eikonal solver. We find that while the EikoNet is exceedingly fast and requires little memory overhead, achieving acceptable accuracy in estimated travel times is difficult and requires extensive computational resources.
This repository implements an iterative magnitude pruning algorithm for pruning neural networks in PyTorch. The pruning method involves gradually removing less significant weights from the model to achieve a specified sparsity, followed by fine-tuning the pruned model to recover performance.
A new algorithm is presented to discriminate reconstructed hadronic decays of tau leptons ($\tau_\mathrm{h}$) that originate from genuine tau leptons in the CMS detector against $\tau_\mathrm{h}$ candidates that originate from quark or gluon jets, electrons, or muons. The algorithm inputs information from all reconstructed particles in the vicinity of a $\tau_\mathrm{h}$ candidate and employs a deep neural network with convolutional layers to efficiently process the inputs. This algorithm leads to a significantly improved performance compared with the previously used one. For example, the efficiency for a genuine $\tau_\mathrm{h}$ to pass the discriminator against jets increases by 10-30% for a given efficiency for quark and gluon jets. Furthermore, a more efficient $\tau_\mathrm{h}$ reconstruction is introduced that incorporates additional hadronic decay modes. The superior performance of the new algorithm to discriminate against jets, electrons, and muons and the improved $\tau_\mathrm{h}$ reconstruction method are validated with LHC proton-proton collision data at $\sqrt{s} =$ 13 TeV.
Convolutional neural networks (CNNs) are becoming more and more important for solving challenging and critical problems in many fields. CNN inference applications have been deployed in safety-critical systems, which may suffer from soft errors caused by high-energy particles, high temperature, or abnormal voltage. Of critical importance is ensuring the stability of the CNN inference process against soft errors. Traditional fault tolerance methods are not suitable for CNN inference because error-correcting code is unable to protect computational components, instruction duplication techniques incur high overhead, and existing algorithm-based fault tolerance (ABFT) techniques cannot protect all convolution implementations. In this paper, we focus on how to protect the CNN inference process against soft errors as efficiently as possible, with the following three contributions. (1) We propose several systematic ABFT schemes based on checksum techniques and analyze their fault protection ability and runtime thoroughly. Unlike traditional ABFT based on matrix-matrix multiplication, our schemes support any convolution implementations. (2) We design a novel workflow integrating all the proposed schemes to obtain a high detection/correction ability with limited total runtime overhead. (3) We perform our evaluation using ImageNet with well-known CNN models including AlexNet, VGG-19, ResNet-18, and YOLOv2. Here, experimental results demonstrate that our implementation can handle soft errors with very limited runtime overhead (4%~8% in both error-free and error-injected situations).
The motivation of this research came about when a neural network direct adaptive control scheme was applied to control the tip position of a flexible robotic arm. Satisfactory control performance was not attainable due to the inherent non-minimum phase characteristics of the flexible robotic arm tip. Most of the existing neural network control algorithms are based on the direct method and exhibit very high sensitivity, if not unstable, closed-loop behavior. Therefore, a neural self-tuning control (NSTC) algorithm is developed and applied to this problem and showed promising results. Simulation results of the NSTC scheme and the conventional self-tuning (STR) control scheme are used to examine performance factors such as control tracking mean square error, estimation mean square error, transient response, and steady state response.
The motivation of this research came about when a neural network direct adaptive control scheme was applied to control the tip position of a flexible robotic arm. Satisfactory control performance was not attainable due to the inherent non-minimum phase characteristics of the flexible robotic arm tip. Most of the existing neural network control algorithms are based on the direct method and exhibit very high sensitivity if not unstable closed-loop behavior. Therefore a neural self-tuning control (NSTC) algorithm is developed and applied to this problem and showed promising results. Simulation results of the NSTC scheme and the conventional self-tuning (STR) control scheme are used to examine performance factors such as control tracking mean square error, estimation mean square error, transient response, and steady state response.
Abstract A real-time capable core Ion Cyclotron Range of Frequencies (ICRF) heating model on NSTX and WEST is developed. The model is based on two nonlinear regression algorithms, the random forest ensemble of decision trees and the multilayer perceptron neural network. The algorithms are trained on TORIC ICRF spectrum solver simulations of the expected flat-top operation scenarios in NSTX and WEST assuming Maxwellian plasmas. The surrogate models are shown to successfully capture the multi-species core ICRF power absorption predicted by the original model for the high harmonic fast wave and the ion cyclotron minority heating schemes while reducing the computational time by six orders of magnitude. Although these models can be expanded, the achieved regression scoring, computational efficiency and increased model robustness suggest these strategies can be implemented into integrated modeling frameworks for real-time control applications.
Abstract The advancement of technology has resulted in a rapid increase in supernova (SN) discoveries. The Subaru/Hyper Suprime-Cam (HSC) transient survey, conducted from fall 2016 through spring 2017, yielded 1824 SN candidates. This gave rise to the need for fast type classification for spectroscopic follow-up and prompted us to develop a machine learning algorithm using a deep neural network with highway layers. This algorithm is trained by actual observed cadence and filter combinations such that we can directly input the observed data array without any interpretation. We tested our model with a dataset from the LSST classification challenge (Deep Drilling Field). Our classifier scores an area under the curve (AUC) of 0.996 for binary classification (SN Ia or non-SN Ia) and 95.3% accuracy for three-class classification (SN Ia, SN Ibc, or SN II). Application of our binary classification to HSC transient data yields an AUC score of 0.925. With two weeks of HSC data since the first detection, this classifier achieves 78.1% accuracy for binary classification, and the accuracy increases to 84.2% with the full dataset. This paper discusses the potential use of machine learning for SN type classification purposes.