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.
A significant hurdle confronts the software reuser attempting to select candidate components from a software repository - discriminating between those components without resorting to inspection of the implementation(s). We outline an approach to this problem based upon neural networks which avoids requiring the repository administrators to define a conceptual closeness graph for the classification vocabulary.
During periods of rapidly changing geomagnetic conditions electric fields form within the Earth’s surface and induce currents known as geomagnetically induced currents(GICs), which interact with unprotected electrical systems our society relies on. In this study, we train multi-variate Long-Short Term Memory neural networks to predict magnitude of north-south component of the geomagnetic field (|BN|) at multiple ground magnetometer stations across Alaska provided by the SuperMAG database with a future goal of predicting geomagnetic field disturbances. Each neural network is driven by solar wind and interplanetary magnetic field inputs from the NASA OMNI database spanning from 2000–2015 and is fine tuned for each station to maximize the effectiveness in predicting |BN|. The neural networks are then compared against multivariate linear regression models driven with the same inputs at each station using Heidke skill scores with thresholds at the 50, 75, 85, and 99 percentiles for |BN|. The neural network models show significant increases over the linear regression models for |BN| thresholds. We also calculate the Heidke skill scores for d|BN|/dt by deriving d|BN|/dt from |BN| predictions. However, neural network models do not show clear outperformance compared to the linear regression models. To retain the sign information and thus predict BN instead of |BN|, a secondary so-called polarity model is utilized. The polarity model is run in tandem with the neural networks predicting geomagnetic field in a coupled model approach and results in a high correlation between predicted and observed values for all stations. We find this model a promising starting point for a machine learned geomagnetic field model to be expanded upon through increased output time history and fast turnaround times.
Neural networks are being used to control a robot arm in a telerobotic operation. The concept uses neural networks for both joint and inverse kinematics in a robotic control application. An upper level neural network is trained to learn inverse kinematic mappings. The output, a trajectory, is then fed to the Decentralized Adaptive Joint Controllers. This neural network implementation has shown that the controlled arm recovers from unexpected payload changes while following the reference trajectory. The neural network-based decentralized joint controller is faster, more robust and efficient than conventional approaches. Implementations of this architecture are discussed that would relax assumptions about dynamics, obstacles, and heavy loads. This system is being developed to use with the Space Shuttle Remote Manipulator System.
This paper investigates the applicability as well as the accuracy of artificial neural networks for estimating specific parameters that describe reservoir properties based on seismic data. Our approach relies on JPL's adjoint operators general purpose neural network code to determine the best suited architecture. We believe that results presented in this work demonstrate that artificial neural networks produce surprisingly accurate estimates of the reservoir parameters.
Autonomous environment-monitoring networks (AEMNs) are artificial neural networks that are specialized for recognizing familiarity and, conversely, novelty. Like a biological neural network, an AEMN receives a constant stream of inputs. For purposes of computational implementation, the inputs are vector representations of the information of interest. As long as the most recent input vector is similar to the previous input vectors, no action is taken. Action is taken only when a novel vector is encountered. Whether a given input vector is regarded as novel depends on the previous vectors; hence, the same input vector could be regarded as familiar or novel, depending on the context of previous input vectors. AEMNs have been proposed as means to enable exploratory robots on remote planets to recognize novel features that could merit closer scientific attention. AEMNs could also be useful for processing data from medical instrumentation for automated monitoring or diagnosis. The primary substructure of an AEMN is called a spindle. In its simplest form, a spindle consists of a central vector (C), a scalar (r), and algorithms for changing C and r. The vector C is constructed from all the vectors in a given continuous stream of inputs, such that it is minimally distant from those vectors. The scalar r is the distance between C and the most remote vector in the same set. The construction of a spindle involves four vital parameters: setup size, spindle-population size, and the radii of two novelty boundaries. The setup size is the number of vectors that are taken into account before computing C. The spindle-population size is the total number of input vectors used in constructing the spindle counting both those that arrive before and those that arrive after the computation of C. The novelty-boundary radii are distances from C that partition the neighborhood around C into three concentric regions (see Figure 1). During construction of the spindle, the changing spindle radius is denoted by h. It is the final value of h, reached before beginning construction on the next spindle, that is denoted by r. During construction of a spindle, if a new vector falls between C and the inner boundary, the vector is regarded as completely familiar and no action is taken. If the new vector falls into the region between the inner and outer boundaries, it is considered unusual enough to warrant the adjustment of C and r by use of the aforementioned algorithms, but not unusual enough to be considered novel. If a vector falls outside the outer boundary, it is considered novel, in which case one of several appropriate responses could be initiation of construction of a new spindle.
The use of hyperspectral data to determine the abundance of constituents in a certain portion of the Earth's surface relies on the capability of imaging spectrometers to provide a large amount of information at each pixel of a certain scene. Today, hyperspectral imaging sensors are capable of generating unprecedented volumes of radiometric data. The Airborne Visible/Infrared Imaging Spectrometer (AVIRIS), for example, routinely produces image cubes with 224 spectral bands. This undoubtedly opens a wide range of new possibilities, but the analysis of such a massive amount of information is not an easy task. In fact, most of the existing algorithms devoted to analyzing multispectral images are not applicable in the hyperspectral domain, because of the size and high dimensionality of the images. The application of neural networks to perform unsupervised classification of hyperspectral data has been tested by several authors and also by us in some previous work. We have also focused on analyzing the intrinsic capability of neural networks to parallelize the whole hyperspectral unmixing process. The results shown in this work indicate that neural network models are able to find clusters of closely related hyperspectral signatures, and thus can be used as a powerful tool to achieve the desired classification. The present work discusses the possibility of using a Self Organizing neural network to perform unsupervised classification of hyperspectral images. In sections 3 and 4, the topology of the proposed neural network and the training algorithm are respectively described. Section 5 provides the results we have obtained after applying the proposed methodology to real hyperspectral data, described in section 2. Different parameters in the learning stage have been modified in order to obtain a detailed description of their influence on the final results. Finally, in section 6 we provide the conclusions at which we have arrived.
A novel analog-digital hybrid architecture based on the utilization of high density digital random access memories for the storage of the synaptic weights of a neural network, and high speed analog hardware to perform neural computation is described. An electronic neurocomputer based on such an architecture is ideally suited for investigating the dynamics, associative recall properties, and computational capabilities of neural networks and provides significant speed improvement in comparison to conventional software based neural network simulations. As a demonstration of the feasibility of the hybrid architectural concept, a prototype breadboard hybrid neurocomputer system with 32 neurons has been designed and fabricated with off-the-shelf hardware components. The performance of the breadboard system has been tested for variety of applications including associative memory and combinatorial problem solving such as Graph Coloring, and is discussed in this paper.
Winner-take-all (WTA) neural networks using pulse-firing processing elements are introduced. In the pulse-firing WTA (PWTA) networks described, input and activation signal shunting is controlled by one shared lateral inhibition signal. This organization yields an O(n) area complexity that is convenient for integrated circuit implementation. Appropriately specified network parameters allow for the accurate continuous evaluation of inputs using a signal representation compatible with established pulse-firing neural network implementations.
Channel and source coding theories are discussed. The following subject areas are covered: large constraint length convolutional codes (the Galileo code); decoder design (the big Viterbi decoder); Voyager's and Galileo's data compression scheme; current research in data compression for images; neural networks for soft decoding; neural networks for source decoding; finite-state codes; and fractals for data compression.
The fault-tolerance of analog parallel distributed implementations of a multivariable aircraft neurocontroller is analyzed by simulating weight and neuron failures in a simplified scheme of analog processing based on the functional architecture of the ETANN chip (Electrically Trainable Artificial Neural Network). The neural information processing is found to be only partially distributed throughout the set of weights of the neurocontroller synthesized with the backpropagation algorithm. Although the degree of distribution of the neural processing, and consequently the fault-tolerance of the neurocontroller, could be enhanced using Locally Distributed Weight and Neuron Approaches, a satisfactory level of fault-tolerance could only be obtained by retraining the degrated VLSI neurocontroller. The possibility of maintaining neurocontrol performance and stability in the presence of single weight of neuron failures was demonstrated through an automated retraining procedure of the neurocontroller based on a pre-programmed choice and sequence of the training parameters.
A key question is how to utilize civilian government agencies along with an industrial consortium to successfully complement the so far primarily defense-oriented neural network research. Civilian artificial neural system projects, such as artificial cerebellar neurocontrollers aimed at duplicating nature's existing neural network solutions for adaptive sensorimotor coordination, are proposed by such a synthesis. The cerebellum provides an intelligent interface between higher possibly symbolic levels of human intelligence and repetitious demands of real world conventional controllers. The generation of such intelligent interfaces could be crucial to the economic feasibility of the human settlement of space and an improvement in telerobotics techniques to permit the cost-effective exploitation of nonterrestrial materials and planetary exploration and monitoring. The authors propose a scientific framework within which such interagency activities could effectively cooperate.
Viewgraphs on a neurocomputing approach to model determination of large space structures are presented. Topics covered include: background on neural networks; math modeling with neural networks; computer architectures for real-time modeling; and hardware and software implementation considerations, OPTIMA/1 data acquisition, modeling, and control workstation.
This paper addresses the need to predict the flight energy consumption of aerial vehicles in the presence of wind using machine learning techniques. The presented work is critical to achieving sustainable and efficient operations for Advanced Air Mobility (AAM) and to evaluating the readiness of the ground-supporting energy infrastructure, e.g., electric grid and AAM portals. The flight energy consumption is described using the "energy per meter" (EPM) metric. We present a comparison study of influential machine learning techniques in predicting EPM using real-world flight test data. We presented new results of using the Decision Tree, Random Forest, and linear regression techniques, along with our previous results using the Recurrent Neural Network and Feed Forward Neural Network techniques. The comparison results show that the Linear Regression method outperforms other methods on the basis of the Mean Squared Error and error variance.
We present a neural network based multi-‐spectral image segmentation method. A neural network is trained on the selected features of both the objects and background in the longwave (LW) Infrared (IR) images. Multiple iterations of training are performed until the accuracy of the segmentation reaches satisfactory level. The segmentation boundary of the LW image is used to segment the midwave (MW) and shortwave (SW) IR images. A second neural network detects the local discontinuities and refines the accuracy of the local boundaries. The neural net based segmentation method is compared with Wavelet-‐threshold and Grab-‐Cut methods. Test results have shown increased accuracy and robustness of this segmentation scheme for multi-‐spectral IR images.
Measurement of three-dimensional (3-D) three-component velocity fields is of great importance in both ground and space experiments for understanding materials processing and fluid physics. The experiments in these fields most likely inhibit the application of conventional planar probes for observing 3-D phenomena. Here, we present the investigation results of stereoscopic tracking velocimetry (STV) for measuring 3-D velocity fields, which include diagnostic technology development, experimental velocity measurement, and comparison with analytical and numerical computation. STV is advantageous in system simplicity for building compact hardware and in software efficiency for continual near-real-time monitoring. It has great freedom in illuminating and observing volumetric fields from arbitrary directions. STV is based on stereoscopic observation of particles-Seeded in a flow by CCD sensors. In the approach, part of the individual particle images that provide data points is likely to be lost or cause errors when their images overlap and crisscross each other especially under a high particle density. In order to maximize the valid recovery of data points, neural networks are implemented for these two important processes. For the step of particle overlap decomposition, the back propagation neural network is utilized because of its ability in pattern recognition with pertinent particle image feature parameters. For the step of particle tracking, the Hopfield neural network is employed to find appropriate particle tracks based on global optimization. Our investigation indicates that the neural networks are very efficient and useful for stereoscopically tracking particles. As an initial assessment of the diagnostic technology performance, laminar water jets with and without pulsation are measured. The jet tip velocity profiles are in good agreement with analytical predictions. Finally, for testing in material processing applications, a simple directional solidification apparatus is built for experimenting with a metal analog of succinonitrile. Its 3-D velocity field at the liquid phase is then measured to be compared with those from numerical computation. Our theoretical, numerical, and experimental investigations have proven STV to be a viable candidate for reliably measuring 3-D flow velocities. With current activities are focused on further improving the processing efficiency, overall accuracy, and automation, the eventual efforts of broad experimental applications and concurrent numerical modeling validation will be vital to many areas in fluid flow and materials processing.
Understanding orbital mechanics is essential in space flight and navigation applications, and leveraging modern force models for flight path projection remains an important aspect in space mission design and operation. However, force models do not capture all the dynamics or perturbations in the space environment and thus are subject to errors in predicting the state vectors. The further out the predicted miss distance between spacecraft is from the time of closest approach (TCA), the larger the propagated errors in the predicted miss distance at TCA is. The dependency on these force models for spacecraft flight state prediction calls for a more reliable method that can quantify, or even reduce, these propagated errors. With recent advances in the field artificial intelligence, specifically in machine and deep learning algorithms, a model that implements these approaches can improve on the modern force model approach. The goal for this work is to provide an early-information decision-making threshold, in order to prioritize risk assessment implementation, given the ongoing increase of space objects. In analyzing the relationship of several parameters from conjunction data messages(CDMs) and solar information, early information becomes viable in miss distance prediction with unsupervised learning techniques, which learn the parameters that are linked together with miss distance and probability of collision (Pc) variables. Another approach implemented for identifying relationships within CDMs is supervised learning, in which a shallow neural network binary classifier learns to distinguish events with Pc values¡108. These parameters detected in the unsupervised process are then applied to a regression neural network, which predicts the miss distance at TCA for a specific event within a given uncertainty bound. For the regression neural network, a Long Short Term Memory (LSTM) neural network is implemented, which yields memory about each time step in an event. Using an LSTM network, the model learns to predict miss distance within 0.2km of the value measured at TCA. Although there is a limited amount of "close miss" data to train a network, the network learns to associate parameters, like large energy dissipation rates with the secondary object, with an elevated Pc
The ultimate goal of this report was to integrate the powerful tools of artificial intelligence into the traditional process of software development. To maintain the US aerospace competitive advantage, traditional aerospace and software engineers need to more easily incorporate the technology of artificial intelligence into the advanced aerospace systems being designed today. The future goal was to transition artificial intelligence from an emerging technology to a standard technology that is considered early in the life cycle process to develop state-of-the-art aircraft automation systems. This report addressed the future goal in two ways. First, it provided a matrix that identified typical aircraft automation applications conducive to various artificial intelligence methods. The purpose of this matrix was to provide top-level guidance to managers contemplating the possible use of artificial intelligence in the development of aircraft automation. Second, the report provided a methodology to formally evaluate neural networks as part of the traditional process of software development. The matrix was developed by organizing the discipline of artificial intelligence into the following six methods: logical, object representation-based, distributed, uncertainty management, temporal and neurocomputing. Next, a study of existing aircraft automation applications that have been conducive to artificial intelligence implementation resulted in the following five categories: pilot-vehicle interface, system status and diagnosis, situation assessment, automatic flight planning, and aircraft flight control. The resulting matrix provided management guidance to understand artificial intelligence as it applied to aircraft automation. The approach taken to develop a methodology to formally evaluate neural networks as part of the software engineering life cycle was to start with the existing software quality assurance standards and to change these standards to include neural network development. The changes were to include evaluation tools that can be applied to neural networks at each phase of the software engineering life cycle. The result was a formal evaluation approach to increase the product quality of systems that use neural networks for their implementation.
An area of application of neural networks is considered. A neuron is modeled as a linear threshold gate, and the network architecture considered is the layered feedforward network. It is shown how common arithmetic functions such as multiplication and sorting can be efficiently computed in a shallow neural network. Some known results are improved by showing that the product of two n-bit numbers and sorting of n n-bit numbers can be computed by a polynomial-size neural network using only four and five unit delays, respectively. Moreover, the weights of each threshold element in the neural networks require O(log n)-bit (instead of n-bit) accuracy. These results can be extended to more complicated functions such as multiple products, division, rational functions, and approximation of analytic functions.