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.
The ESnet SmartNIC is a collection of Verilog based FPGA design software, as well as drivers to interact with the FPGA. It provides an infrastructure framework for FPGA based hardware acceleration of network packet use cases. Different research and production applications can be easily written for the SmartNIC. Its advantage is that it reduces the development time for new applications by providing a pre-existing shell library for common functions.
Graph Convolutional Neural Networks (GCNs) are widely used to process large-scale graph data. Different from deep neural networks (DNNs), GCNs are sparse, irregular, and unstructured, posing unique challenges to hardware acceleration with regular processing elements (PEs). In particular, the adjacency matrix of a GCN is extremely sparse, leading to frequent but irregular memory access, low spatial/temporal data locality and poor data reuse. Furthermore, a realistic graph usually consists of unstructured data (e.g., unbalanced distributions), creating significantly different processing times and imbalanced workload for each node in GCN acceleration. To overcome these challenges, we propose an end-to-end hardware-software co-design to accelerate GCNs on resource-constrained FPGAs with the features including: (1) A custom dataflow that leverages symmetry along the diagonal of the adjacency matrix to accelerate feature aggregation for undirected graphs. We utilize either the upper or the lower triangular matrix of the adjacency matrix to perform aggregation in GCN to improve data reuse. (2) Unified compute cores for both aggregation and transform phases, with full support to the symmetry-based dataflow. These cores can be dynamically reconfigured to the systolic mode for transformation or as individual accumulators for aggregation in GCN processing. (3) Preprocessing of the graph in software to rearrange the edges and features to match the custom dataflow. This step improves the regularity in memory access and data reuse in the aggregation phase. Moreover, we quantize the GCN precision from FP32 to INT8 to reduce the memory footprint without losing the inference accuracy. We implement our accelerator design in Intel Stratix10 MX FPGA board with HBM2, and demonstrate 1.3x-110.5x improvement in end-to-end GCN latency as compared to the state-of the-art FPGA implementations, on the graph datasets of Cora, Pubmed, Citeseer and Reddit.
An autonomous system refers to the system that has the power and ability for self-governance in the performance of system functions. Autonomous systems have been actively pursued in a variety of domains such as automotive, aviation, maritime, medicine, and nuclear fields. As an unmanned concept employing the highest automation level, the autonomous system basically performs most of the work in normal operations or emergency situations. However, despite advances in technology, many researchers have noted these systems still require human actions. The nature of human actions on autonomous systems is different than the human actions that are considered in existing systems. Nevertheless, only a few studies have been conducted on 1) characterizing the different types of errors and risks associated with human actions interacting with autonomous systems and 2) how to evaluate human actions in the autonomous operations. As a starting point, this study aims to investigate differences of tasks in autonomous operation compared to those in existing nuclear power plant operation using the Event Modeling Risk Assessment Using Linked Diagram (EMRALD) software. In this paper, insights aspect of human error and time are derived out and discussed based on the output of the EMRALD models.
Machine learning technology spans many areas and today plays a significant role in addressing a wide range of problems in critical domains,i.e., healthcare, autonomous driving, finance, manufacturing, cybersecurity,etc. Metamorphic testing (MT) is considered a simple but very powerful approach in testing such computationally complex systems for which either an oracle is not available or is available but difficult to apply. Conventional metamorphic testing techniques have certain limitations in verifying deep learning-based models (i.e., convolutional neural networks (CNNs)) that have a stochastic nature (because of randomly initializing the network weights) in their training. In this article, we attempt to address this problem by using a statistical metamorphic testing (SMT) technique that does not require software testers to worry about fixing the random seeds (to get deterministic results) to verify the metamorphic relations (MRs). We propose seven MRs combined with different statistical methods to statistically verify whether the program under test adheres to the relation(s) specified in the MR(s). We further use mutation testing techniques to show the usefulness of the proposed approach in the healthcare space and test two CNN-based deep learning models (used for pneumonia detection among patients). The empirical results show that our proposed approach uncovers 85.71% of the implementation faults in the classifiers under test (CUT). Furthermore, we also propose an MRs minimization algorithm for the CUT, thus saving computational costs and organizational testing resources.
The present paper is concerned with the approaches to be employed to overcome the set of limitations in software technology which impedes currently an effective use of parallel hardware technology. The process required to solve the arising problems is found to involve four different stages. At the present time, Stage One is nearly finished, while Stage Two is under way. Tentative explorations are beginning on Stage Three, and Stage Four is more distant. In Stage One, parallelism is introduced into the hardware of a single computer, which consists of one or more processors, a main storage system, a secondary storage system, and various peripheral devices. In Stage Two, parallel execution of cooperating programs on different machines becomes explicit, while in Stage Three, new languages will make parallelism implicit. In Stage Four, there will be very high level user interfaces capable of interacting with scientists at the same level of abstraction as scientists do with each other.
The refinement of radar image analysis methods has led to a need for a systems approach to radar image processing software. Developments stimulated through satellite radar are combined with standard image processing techniques to create a user environment to manipulate and analyze airborne and satellite radar images. One aim is to create radar products for the user from the original data to enhance the ease of understanding the contents. The results are called secondary image products and derive from the original digital images. Another aim is to support interactive SAR image analysis. Software methods permit use of a digital height model to create ortho images, synthetic images, stereo-ortho images, radar maps or color combinations of different component products. Efforts are ongoing to integrate individual tools into a combined hardware/software environment for interactive radar image analysis.
A laboratory for developing robotics technology for hazardous and repetitive Shuttle and payload processing activities is discussed. An overview of the computer hardware and software responsible for integrating the laboratory systems is given. The center's anthropomorphic robot is placed on a track allowing it to be moved to different stations. Various aspects of the laboratory equipment are described, including industrial robot arm control, smart systems integration, the supervisory computer, programmable process controller, real-time tracking controller, image processing hardware, and control display graphics. Topics of research include: automated loading and unloading of hypergolics for space vehicles and payloads; the use of mobile robotics for security, fire fighting, and hazardous spill operations; nondestructive testing for SRB joint and seal verification; Shuttle Orbiter radiator damage inspection; and Orbiter contour measurements. The possibility of expanding the laboratory in the future is examined.
The present conference on computational methods for aeronautics applications discusses topics in the fields of parallel computing, multidisciplinary computational methods, grid generation, visualization methods for CFD, probabilistic modeling, numerical simulations and methodologies for different flow regimes, computational strategies and adaptive methods in CFD, and computational strategies and dynamics and control. Attention is given to the MACH system-software kernel, a multidisciplinary approach to aeroelastic analysis, interactive grid generation with control points, interactive flow visualization using stream surfaces, numerical simulations of dynamic/aerodynamic interactions, implicit mathods for the Navier-Stokes equations, and the automatic phase-space analysis of dynamical systems.
The Structures Technology Program Office (STPO) at NASA Langley Research Center has held two workshops with representatives from the commercial airframe companies to establish a plan for development of a standard cost reporting format and a cost prediction tool for conceptual and preliminary designers. This paper reviews the findings of the workshop representatives with a plan for implementation of their recommendations. The recommendations of the cost tracking and reporting committee will be implemented by reinstituting the collection of composite part fabrication data in a format similar to the DoD/NASA Structural Composites Fabrication Guide. The process of data collection will be automated by taking advantage of current technology with user friendly computer interfaces and electronic data transmission. Development of a conceptual and preliminary designers' cost prediction model will be initiated. The model will provide a technically sound method for evaluating the relative cost of different composite structural designs, fabrication processes, and assembly methods that can be compared to equivalent metallic parts or assemblies. The feasibility of developing cost prediction software in a modular form for interfacing with state of the art preliminary design tools and computer aided design (CAD) programs is assessed.
A conceptual and preliminary designers' cost prediction model has been initiated. The model will provide a technically sound method for evaluating the relative cost of different composite structural designs, fabrication processes, and assembly methods that can be compared to equivalent metallic parts or assemblies. The feasibility of developing cost prediction software in a modular form for interfacing with state of the art preliminary design tools and computer aided design programs is being evaluated. The goal of this task is to establish theoretical cost functions that relate geometric design features to summed material cost and labor content in terms of process mechanics and physics. The output of the designers' present analytical tools will be input for the designers' cost prediction model to provide the designer with a data base and deterministic cost methodology that allows one to trade and synthesize designs with both cost and weight as objective functions for optimization. The approach, goals, plans, and progress is presented for development of COSTADE (Cost Optimization Software for Transport Aircraft Design Evaluation).
The Structures Technology Program Office (STPO) at NASA LaRC has initiated development of a conceptual and preliminary designers' cost prediction model. The model will provide a technically sound method for evaluating the relative cost of different composite structural designs, fabrication processes, and assembly methods that can be compared to equivalent metallic parts or assemblies. The feasibility of developing cost prediction software in a modular form for interfacing with state-of-the-art preliminary design tools and computer aided design programs is being evaluated. The goal of this task is to establish theoretical cost functions that relate geometric design features to summed material cost and labor content in terms of process mechanics and physics. The output of the designers' present analytical tools will be input for the designers' cost prediction model to provide the designer with a database and deterministic cost methodology that allows one to trade and synthesize designs with both cost and weight as objective functions for optimization. This paper presents the team members, approach, goals, plans, and progress to date for development of COSTADE (Cost Optimization Software for Transport Aircraft Design Evaluation).
Engineering Design is one of the most important areas in engineering education. Deterministic Design Methodology (DDM) is the only design method that is taught in most engineering schools. This method does not give a direct account of uncertainties in design parameters. Hence, it is impossible to quantify the uncertainties in the response and the actual safety margin remains unknown. The desire for a design methodology tha can identify the primitive (random) variables that affect the structural behavior has led to a growing interest on Probabilistic Design Methodology (PDM). This method is gaining more recognition in industries than in educational institutions. Some of the reasons for the limited use of the PDM at the moment are that many are unaware of its potentials, and most of the software developed for PDM are very recent. The central goal of the PDM project at Tennessee State University is to introduce engineering students to the method. The students participating in the project learn about PDM and the computer codes that are available to the design engineer. The software being used of this project is NESSUS (Numerical Evaluation of Stochastic Structures Under Stress) developed under NASA probabilistic structural analysis program. NESSUS has three different modules which make it a very comprehensive computer code for PDM. A research in technology transfer through course offering in PDM is in effect a Tennessee State University. The aim is to familiarize students with the problem of uncertainties in engineering design. Included in the paper are some projects on PDM carried out by some students and faculty. The areas this method is being applied at the moment include, Design of Gears (spur and worm); Design of Shafts; Design of Statistically Indeterminate Frame Structures; Design of Helical Springs; and Design of Shock Absorbers. Some of the current results of these projects are presented.
The developed methodology is aimed at improving the efficiency of executing block-structured algorithms on parallel, distributed, heterogeneous computers. The basic approach of these algorithms is to divide the flow domain into many sub- domains called blocks, and solve the governing equations over these blocks. Dynamic load balancing problem is defined as the efficient distribution of the blocks among the available processors over a period of several hours of computations. In environments with computers of different architecture, operating systems, CPU speed, memory size, load, and network speed, balancing the loads and managing the communication between processors becomes crucial. Load balancing software tools for mutually dependent parallel processes have been created to efficiently utilize an advanced computation environment and algorithms. These tools are dynamic in nature because of the chances in the computer environment during execution time. More recently, these tools were extended to a second operating system: NT. In this paper, the problems associated with this application will be discussed. Also, the developed algorithms were combined with the load sharing capability of LSF to efficiently utilize workstation clusters for parallel computing. Finally, results will be presented on running a NASA based code ADPAC to demonstrate the developed tools for dynamic load balancing.
There were three major tasks and objectives that the Tropospheric Trace Gas and Airborne Measurement Group's (TTGAMG) worked on for different aspects of this grant: 1) Migrate the data acquisition system from HP-UX to Linux, thus reducing future costs as the result of software and operating system (OS) upgrades and improving upon usability as membership in the group changes; 2) Rework the Optical Parametric Oscillator (OPO) cavities. These are the OPOs that are integral to the Georgia Institute of Technology Airborne Laser Induced Fluorescent Experiment (GITALIFE) that the TTGAMG flew on TRACE-P. The objective was to improve upon optimizing the pump laser energy and narrowing the linewidth of the UV wavelength generated by the OPOs; 3) Improve and expand the interactive website on http://tmbk2.eas.gatech.edu by adding 3-D graphing, improve the response time for Joe Surfer Dude, improve performance, usability, and expand the database. If I were to assign a letter grade to each of the above tasks, I would give the TTGAMG two Bs and an A to the tasks listed above.
In this paper we discuss a software method implmented in the Planck-LFI pseudo-correlation receivers which uses a tunable gain modulation factor, r, in the sky-load difference.
The TSPT (Time Series Product Tool) software was custom-designed for NASA to rapidly create and display single-band and band-combination time series, such as NDVI (Normalized Difference Vegetation Index) images, for wide-area crop surveillance and for other time-critical applications. The TSPT, developed in MATLAB, allows users to create and display various MODIS (Moderate Resolution Imaging Spectroradiometer) or simulated VIIRS (Visible/Infrared Imager Radiometer Suite) products as single images, as time series plots at a selected location, or as temporally processed image videos. Manually creating these types of products is extremely labor intensive; however, the TSPT development tool makes the process simplified and efficient. MODIS is ideal for monitoring large crop areas because of its wide swath (2330 km), its relatively small ground sample distance (250 m), and its high temporal revisit time (twice daily). Furthermore, because MODIS imagery is acquired daily, rapid changes in vegetative health can potentially be detected. The new TSPT technology provides users with the ability to temporally process high-revisit-rate satellite imagery, such as that acquired from MODIS and from its successor, the VIIRS. The TSPT features the important capability of fusing data from both MODIS instruments onboard the Terra and Aqua satellites, which drastically improves cloud statistics. With the TSPT, MODIS metadata is used to find and optionally remove bad and suspect data. Noise removal and temporal processing techniques allow users to create low-noise time series plots and image videos and to select settings and thresholds that tailor particular output products. The TSPT GUI (graphical user interface) provides an interactive environment for crafting what-if scenarios by enabling a user to repeat product generation using different settings and thresholds. The TSPT Application Programming Interface provides more fine-tuned control of product generation, allowing experienced programmers to bypass the GUI and to create more user-specific output products, such as comparison time plots or images. This type of time series analysis tool for remotely sensed imagery could be the basis of a large-area vegetation surveillance system. The TSPT has been used to generate NDVI time series over growing seasons in California and Argentina and for hurricane events, such as Hurricane Katrina.
A structural analysis, sizing optimization, and weight prediction study was performed by Collier Research Corporation and NASA Glenn on a spherical cryogenic hydrogen tank. The tank consisted of an inner and outer wall separated by a vacuum for thermal insulation purposes. HyperSizer (Collier Research and Development Corporation), a commercial automated structural analysis and sizing software package was used to design the lightest feasible tank for a given overall size and thermomechanical loading environment. Weight trade studies were completed for different panel concepts and metallic and composite material systems. Extensive failure analyses were performed for each combination of dimensional variables, materials, and layups to establish the structural integrity of tank designs. Detailed stress and strain fields were computed from operational temperature changes and pressure loads. The inner tank wall is sized by the resulting biaxial tensile stresses which cause it to be strength driven, and leads to an optimum panel concept that need not be stiffened. Conversely, the outer tank wall is sized by a biaxial compressive stress field, induced by the pressure differential between atmospheric pressure and the vacuum between the tanks, thereby causing the design to be stability driven and thus stiffened to prevent buckling. Induced thermal stresses become a major sizing driver when a composite or hybrid composite/metallic material systems are used for the inner tank wall for purposes such as liners to contain the fuel and reduce hydrogen permeation.
The Orion spacecraft will replace the space shuttle and will be the first human spacecraft since the Apollo program to leave low earth orbit. This vehicle will serve as the cornerstone of a complete space transportation system with a myriad of mission requirements necessitating rendezvous to multiple vehicles in earth orbit, around the moon and eventually beyond . These goals will require a complex and robust vehicle that is, significantly different from both the space shuttle and the command module of the Apollo program. Historically, orbit operations have been accomplished with heavy reliance on ground support and manual crew reconfiguration and monitoring. One major difference with Orion is that automation will be incorporated as a key element of the man-vehicle system. The automated system will consist of software devoted to transitioning between events based on a master timeline. This effectively adds a layer of high level sequencing that moves control of the vehicle from one phase to the next. This type of automated control is not entirely new to spacecraft since the shuttle uses a version of this during ascent and entry operations. During shuttle orbit operations however many of the software modes and hardware switches must be manually configured through the use of printed procedures and instructions voiced from the ground. The goal of the automation scheme on Orion is to extend high level automation to all flight phases. The move towards automation represents a large shift from current space shuttle operations, and so these new systems will be adopted gradually via various safeguards. These include features such as authority-to-proceed, manual down modes, and functional inhibits. This paper describes the contrast between the manual and ground approach of the space shuttle and the proposed automation of the Orion vehicle. I will introduce typical orbit operations that are common to all rendezvous missions and go on to describe the current Orion automation architecture and contrast it with shuttle rendezvous techniques and circumstances. The shuttle rendezvous profile is timed to take approximately 3 days from orbit insertion to docking at the International Space Station (ISS). This process can be divided into 3 phases: far-field, mid-field and proximity operations. The far-field stage is characterized as the most quiescent phase. The spacecraft is usually too far to navigate using relative sensors and uses the Inertial Measurement Units (IMU s) to numerically solve for its position. The maneuvers are infrequent, roughly twice per day, and are larger than other burns in the profile. The shuttle uses this opportunity to take extensive ground based radar updates and keep high fidelity orbit states on the ground. This state is then periodically uplinked to the shuttle computers. The targeting solutions for burn maneuvers are also computed on the ground and uplinked. During the burn the crew is responsible for setting the shuttle attitude and configuring the propulsion system for ignition. Again this entire process is manually driven by both crew and ground activity. The only automatic processes that occur are associated with the real-time execution of the burn. The Orion automated functionality will seek to relieve the workload of both the crew and ground during this phase