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.
In this presentation we investigate a novel framework for the design of autonomous fuzzy intelligent systems. The system integrates the following modules into a single autonomous entity: (1) a fuzzy expert system; (2) artificial neural network; (3) genetic algorithm; and (4) case-base reasoning. We describe the integration of these units into one intelligent structure and discuss potential applications.
The goal of optimal control is to determine a sequence of inputs for maximizing or minimizing a given performance criterion subject to the dynamics and constraints of the system under observation. This work introduces Control Physics-Informed Neural Networks (PINNs), which simultaneously learn both the system states and the optimal control signal in a single-stage framework that leverages the system’s underlying physical laws. While prior approaches often follow a two-stage process-modeling, the system first and then devising its control—the presented novel framework embeds the necessary optimality conditions directly into the network architecture and loss function. We demonstrate the effectiveness of the novel methodology by solving various open-loop optimal control problems governed by analytical, one-dimensional, and two-dimensional partial differential equations (PDEs).
Abstract Many problems in climate science require the identification of signals obscured by both the “noise” of internal climate variability and differences across models. Following previous work, we train an artificial neural network (ANN) to predict the year of a given map of annual‐mean temperature (or precipitation) from forced climate model simulations. This prediction task requires the ANN to learn forced patterns of change amidst a background of climate noise and model differences. We then apply a neural network visualization technique (layerwise relevance propagation) to visualize the spatial patterns that lead the ANN to successfully predict the year. These spatial patterns thus serve as “reliable indicators” of the forced change. The architecture of the ANN is chosen such that these indicators vary in time, thus capturing the evolving nature of regional signals of change. Results are compared to those of more standard approaches like signal‐to‐noise ratios and multilinear regression in order to gain intuition about the reliable indicators identified by the ANN. We then apply an additional visualization tool (backward optimization) to highlight where disagreements in simulated and observed patterns of change are most important for the prediction of the year. This work demonstrates that ANNs and their visualization tools make a powerful pair for extracting climate patterns of forced change.
We train neural networks to quickly generate redshift-space galaxy power spectrum covariances from a given parameter set (cosmology and galaxy bias). This covariance emulator utilizes a combination of traditional fully connected network layers and transformer architecture to accurately predict covariance matrices for the high redshift, north galactic cap sample of the BOSS DR12 galaxy catalog. We run simulated likelihood analyses with emulated and brute-force computed covariances, and we quantify the network’s performance via two different metrics: (1) difference in Χ 2 and (2) likelihood contours for simulated BOSS DR 12 analyses. We find that the emulator returns excellent results over a large parameter range. We then use our emulator to perform a reanalysis of the BOSS HighZ NGC galaxy power spectrum, and find that varying covariance with cosmology along with the model vector produces Ω m = $0.276$$^{+0.013}_{–0.015}$, H 0 = 70.2 ± 1.9 km/s/Mpc, and σ 8 = $0.674$$^{+0.058}_{–0.077}$. These constraints represent an average 0.46σ shift in best-fit values and a 5% increase in constraining power compared to fixing the covariance matrix (Ω m = 0.293 ± 0.017, H 0 = 70.3 ± 2.0 km/s/Mpc, σ 8 = $0.702$$^{+0.063}_{–0.075}$). As a result, this work demonstrates that emulators for more complex cosmological quantities than second-order statistics can be trained over a wide parameter range at sufficiently high accuracy to be implemented in realistic likelihood analyses.
Despite foreseeing tremendous speedups over conventional deep neural networks, the performance advantage of binarized neural networks (BNNs) has merely been showcased on general-purpose processors such as CPUs and GPUs. In fact, due to being unable to leverage bit-level-parallelism with a word-based architecture, GPUs have been criticized for extremely low utilization (1%) when executing BNNs. Consequently, the latest tensorcores in NVIDIA Turing GPUs start to experimentally support bit computation. In this work, we look into this brand new bit computation capability and characterize its unique features. We show that the stride of memory access can significantly affect performance delivery and a data-format co-design is highly desired to support the tensorcores for achieving superior performance than existing software solutions without tensorcores. We realize the tensorcore-accelerated BNN design, particularly the major functions for fully-connect and convolution layers — bit matrix multiplication and bit convolution. Evaluations on two NVIDIA Turing GPUs show that, with ResNet-18, our BTC-BNN design can process ImageNet at a rate of 5.6K images per second, 77% faster than state-of-the-art. Our BNN approach is released on https://github.com/pnnl/TCBNN.
The reinforcement learning techniques developed at Ames Research Center are being applied to proximity and docking operations using the Shuttle and Solar Maximum Mission (SMM) satellite simulation. In utilizing these fuzzy learning techniques, we also use the Approximate Reasoning based Intelligent Control (ARIC) architecture, and so we use two terms interchangeable to imply the same. This activity is carried out in the Software Technology Laboratory utilizing the Orbital Operations Simulator (OOS). This report is the deliverable D3 in our project activity and provides the test results of the fuzzy learning translational controller. This report is organized in six sections. Based on our experience and analysis with the attitude controller, we have modified the basic configuration of the reinforcement learning algorithm in ARIC as described in section 2. The shuttle translational controller and its implementation in fuzzy learning architecture is described in section 3. Two test cases that we have performed are described in section 4. Our results and conclusions are discussed in section 5, and section 6 provides future plans and summary for the project.
In this work, we demonstrate SONOS (silicon-oxide-nitrideoxide- silicon) analog memory arrays that are optimized for neural network inference. The devices are fabricated in a 40nm process and operated in the subthreshold regime for in-memory matrix multiplication. Subthreshold operation enables low conductances to be implemented with low error, which matches the typical weight distribution of neural networks, which is heavily skewed toward near-zero values. This leads to high accuracy in the presence of programming errors and process variations. We simulate the end-to-end neural network inference accuracy, accounting for the measured programming error, read noise, and retention loss in a fabricated SONOS array. Evaluated on the ImageNet dataset using ResNet50, the accuracy using a SONOS system is within 2.16% of floating-point accuracy without any retraining. The unique error properties and high On/Off ratio of the SONOS device allow scaling to large arrays without bit slicing, and enable an inference architecture that achieves 20 TOPS/W on ResNet50, a >10× gain in energy efficiency over state-of-the-art digital and analog inference accelerators.
Here we propose a generalized architecture for the first rapid-single-flux-quantum (RSFQ) associative memory circuit. The circuit employs hyperdimensional computing (HDC), a machine learning (ML) paradigm utilizing vectors with dimensionality in the thousands to represent information. HDC designs have small memory footprints, simple computations, and simple training algorithms compared to superconducting neural network accelerators (SNNAs), making them a better option for scalable SFQ machine learning (ML) solutions. The proposed superconducting HDC (SHDC) circuit uses entirely on-chip RSFQ memory which is tightly integrated with logic, operates at 33.3 GHz, is applicable to general ML tasks, and is manufacturable at practically useful scales given current SFQ fabrication limits. Tailored to a language recognition task, SHDC consists of ~ 2-20 M Josephson junctions (JJs) and consumes up to three times less power than an analogous CMOS HDC circuit while achieving 78-84% higher throughput. SHDC is capable of outperforming the state of the art RSFQ SNNA, SuperNPU, by 48-99% for all benchmark NN architectures tested while occupying up to 90% less area and consuming up to nine times less power. To the best of the authors' knowledge, SHDC is currently the only superconducting ML approach feasible at practically useful scales for real-world ML tasks and capable of online learning.
Here, we outline the development of a general-purpose Python-based data analysis tool for OpenFOAM. Our implementation relies on the construction of OpenFOAM applications that have bindings to data analysis libraries in Python. Double precision data in OpenFOAM is cast to a NumPy array using the NumPy C-API and Python modules may then be used for arbitrary data analysis and manipulation on flow-field information. We highlight how the proposed wrapper may be used for an in-situ online singular value decomposition (SVD) implemented in Python and accessed from the OpenFOAM solver PimpleFOAM. Here, 'in-situ' refers to a programming paradigm that allows for a concurrent computation of the data analysis on the same computational resources utilized for the partial differential equation solver. In addition, to demonstrate parallel deployments, we deploy a distributed SVD, which collects snapshot data across the ranks of a distributed simulation to compute the global left singular vectors. Crucially, both OpenFOAM and Python share the same message passing interface (MPI) communicator for this deployment which allows Python objects and functions to exchange NumPy arrays across ranks. Subsequently, we provide scaling assessments of this distributed SVD on multiple nodes of Intel Broadwell and KNL architectures for canonical test cases such as the large eddy simulations of a backward facing step and a channel flow at friction Reynolds number of 395. Finally, we demonstrate the deployment of a deep neural network for compressing the flow-field information using an autoencoder to demonstrate an ability to use state-of-the-art machine learning tools in the Python ecosystem.
Herein we present a novelty detection framework for Convolutional Neural Network (CNN) sensors that we call Sensor-Activated Feature Extraction One-Class Classification (SAFE-OCC). We show that this framework enables the safe use of computer vision sensors in process control architectures. Emergent control applications use CNN models to map visual data to a state signal that can be interpreted by the controller. Incorporating such sensors introduces a significant system operation vulnerability because CNN sensors can exhibit high prediction errors when exposed to novel (abnormal) visual data. Unfortunately, identifying such novelties in real-time is nontrivial. To address this issue, the SAFE-OCC framework leverages the convolutional blocks of the CNN to create an effective feature space to conduct novelty detection using a desired one-class classification technique. This approach engenders a feature space that directly corresponds to that used by the CNN sensor and avoids the need to derive an independent latent space. We demonstrate the effectiveness of SAFE-OCC via simulated control environments.
Big data applications are on the rise, and so is the number of data centers. The ever-increasing massive data pool needs to be periodically backed up in a secure environment. Moreover, a massive amount of securely backed-up data is required for training binary convolutional neural networks for image classification. XOR and XNOR operations are essential for large-scale data copy verification, encryption, and classification algorithms. The disproportionate speed of existing compute and memory units makes the von Neumann architecture inefficient to perform these Boolean operations. Compute-in-memory (CiM) has proved to be an optimum approach for such bulk computations. The existing CiM-based XOR/XNOR techniques either require multiple cycles for computing or add to the complexity of the fabrication process. Here, we propose a CMOS-based hardware topology for single-cycle in-memory XOR/XNOR operations. Our design provides at least 2× improvement in the latency compared with other existing CMOS-compatible solutions. We verify the proposed system through circuit/system-level simulations and evaluate its robustness using a 5000-point Monte Carlo variation analysis. This all-CMOS design paves the way for practical implementation of CiM XOR/XNOR at scaled technology nodes.
A brief summary of neural networks is presented which concentrates on the design constraints imposed. Major design issues are discussed together with analysis methods and the chosen solutions. Although the system will be capable of running on most transputer architectures, it currently is being implemented on a 40-transputer system connected to a toroidal architecture. Predictions show a performance level equivalent to that of a highly optimized simulator running on the SX-2 supercomputer.
We present an architecture of an intelligent restructurable control system to automatically detect failure of system components, assess its impact on system performance and safety, and reconfigure the controller for performance recovery. Fault detection is based on neural network associative memories and pattern classifiers, and is implemented using a multilayer feedforward network. Details of the fault detection network along with simulation results on health monitoring of a dc motor have been presented. Conceptual developments for fault assessment using an expert system and controller reconfiguration using a neural network are outlined.
Seeker is an automated extravehicular free-flying inspector CubeSat designed and built in-house at the Johnson Space Center (JSC). As a Class 1E project funded by the International Space Station (ISS) Program, Seeker had a stream-lined process to flight certification, but the vehicle had to be designed, developed, tested, and delivered within approximately one year after authority to proceed (ATP) and within a $1.8 million budget. These constraints necessitated an expedited Guidance, Navigation, and Control (GNC) development schedule. Development began with a navigation sensor trade study using Linear Covariance (LinCov) analysis and a rapid sensor down-selection process, resulting in the use of commercial off-the-shelf (COTS) sensors which could be procured quickly and subjected to in-house environmental testing to qualify them for flight. A neural network was used to enable a COTS camera to provide bearing measure-ments for visual navigation. The GNC flight software (FSW) algorithms utilized lean development practices and leveraged the Core Flight Software (CFS) architecture to rapidly develop the GNC system, tune the system parameters, and verify performance in simulation. This pace was anchored by several Hardware-Software Integration (HSI) milestones, which forced the Seeker GNC team to develop the interfaces both between hardware and software and between the GNC domains early in the project and to enable a timely delivery.
Trigger (interesting events) detection is crucial to high-energy and nuclear physics experiments because it improves data acquisition efficiency. It also plays a vital role in facilitating the downstream offline data analysis process. The sPHENIX detector, located at the Relativistic Heavy Ion Collider in Brookhaven National Laboratory, is one of the largest nuclear physics experiments on a world scale and is optimized to detect physics processes involving charm and beauty quarks. Furthermore, these particles are produced in collisions involving two proton beams, two gold nuclei beams, or a combination of the two and give critical insights into the formation of the early universe. This paper presents a model architecture for trigger detection with geometric information from two fast silicon detectors. Transverse momentum is introduced as an intermediate feature from physics heuristics. We also prove its importance through our training experiments. Each event consists of tracks and can be viewed as a graph. A bipartite graph neural network is integrated with the attention mechanism to design a binary classification model. Compared with the state-of-the-art algorithm for trigger detection, our model is parsimonious and increases the accuracy and the AUC score by more than 15%.
The traditional AI answer to the decision making problem for a robot is planning. However, planning is usually CPU-time consuming, depending on the availability and accuracy of a world model. The Dyna system generally described in earlier work, uses trial and error to learn a world model which is simultaneously used to plan reactions resulting in optimal action sequences. It is an attempt to integrate planning, reactive, and learning systems. The architecture of Dyna is presented. The different blocks are described. There are three main components of the system. The first is the world model used by the robot for internal world representation. The input of the world model is the current state and the action taken in the current state. The output is the corresponding reward and resulting state. The second module in the system is the policy. The policy observes the current state and outputs the action to be executed by the robot. At the beginning of program execution, the policy is stochastic and through learning progressively becomes deterministic. The policy decides upon an action according to the output of an evaluation function, which is the third module of the system. The evaluation function takes the following as input: the current state of the system, the action taken in that state, the resulting state, and a reward generated by the world which is proportional to the current distance from the goal state. Originally, the work proposed was as follows: (1) to implement a simple 2-D world where a 'robot' is navigating around obstacles, to learn the path to a goal, by using lookup tables; (2) to substitute the world model and Q estimate function Q by neural networks; and (3) to apply the algorithm to a more complex world where the use of a neural network would be fully justified. In this paper, the system design and achieved results will be described. First we implement the world model with a neural network and leave Q implemented as a look up table. Next, we use a lookup table for the world model and implement the Q function with a neural net. Time limitations prevented the combination of these two approaches. The final section discusses the results and gives clues for future work.
The present invention enhances the bit resolution of a CCD/CID MVM processor by storing each bit of each matrix element as a separate CCD charge packet. The bits of each input vector are separately multiplied by each bit of each matrix element in massive parallelism and the resulting products are combined appropriately to synthesize the correct product. In another aspect of the invention, such arrays are employed in a pseudo-spectral method of the invention, in which partial differential equations are solved by expressing each derivative analytically as matrices, and the state function is updated at each computation cycle by multiplying it by the matrices. The matrices are treated as synaptic arrays of a neural network and the state function vector elements are treated as neurons. In a further aspect of the invention, moving target detection is performed by driving the soliton equation with a vector of detector outputs. The neural architecture consists of two synaptic arrays corresponding to the two differential terms of the soliton-equation and an adder connected to the output thereof and to the output of the detector array to drive the soliton equation.
The internet is turning out to be an integral part of every-one's lives as more and more devices are being connected to serve societal needs. Our work is motivated by two ma-jor observations. Firstly, one drawback of connecting to the network is the threat of network attacks that can compromise users' private information, leading to data loss and adversely affecting productivity. There are several traditional security mechanisms to defend against these attacks, such as firewalls, virtual private networks (VPNs), demilitarized zones (DMZs), and vulnerability scanners. One way to prevent these attacks is early detection and prevention. However, these kinds of architecture do not scale very well because of their centralized nature. Secondly, we observe from heuristics and data set distributions that the majority of the requests made to a server are innocuous. Therefore, almost all server request data sets are highly imbalanced, weighted highly towards the harmless requests.