Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Trains”

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 127 records · Page 7
All sourcesNASA NTRSDOE OSTI
NASA NTRS2017

ISS Training Best Practices and Lessons Learned

Training our crew members for long duration exploration-class missions (LDEM) will have to be qualitatively and quantitatively different from current training practices. However, there is much to be learned from the extensive experience NASA has gained in training crew members for missions on board the International Space Station (ISS). Furthermore, the operational experience on board the ISS provides valuable feedback concerning training effectiveness. Keeping in mind the vast differences between current ISS crew training and training for LDEM, the needs of future crew members, and the demands of future missions, this ongoing study seeks to document current training practices and lessons learned. The goal of the study is to provide input to the design of future crew training that takes as much advantage as possible of what has already been learned and avoids as much as possible past inefficiencies. Results from this study will be presented upon its completion. By researching established training principles, examining future needs, and by using current practices in spaceflight training as test beds, this research project is mitigating program risks and generating templates and requirements to meet future training needs.

ISS↗
NASA NTRS2018

ISS Training Best Practices and Lessons Learned

Training our crew members for long-duration Deep Space Transport (DST) missions will have to be qualitatively and quantitatively different from current training practices. However, there is much to be learned from the extensive experience NASA has gained in training crew members for missions on board the International Space Station (ISS). Furthermore, the operational experience on board the ISS provides valuable feedback concerning training effectiveness. Keeping in mind the vast differences between current ISS crew training and training for DST missions, the needs of future crew members, and the demands of future missions, this ongoing study seeks to document current training practices and lessons learned. The goal of the study is to provide input to the design of future crew training that takes as much advantage as possible of what has already been learned and avoids as much as possible past inefficiencies. Results from this study will be presented upon its completion. By researching established training principles, examining future needs, and by using current practices in spaceflight training as test beds, this research project is mitigating program risks and generating templates and requirements to meet future training needs.

training↗
DOE OSTI2022

Physics-Based Method for Generating Fully Synthetic IV Curve Training Datasets for Machine Learning Classification of PV Failures

Classification machine learning models require high-quality labeled datasets for training. Among the most useful datasets for photovoltaic array fault detection and diagnosis are module or string current-voltage (IV) curves. Unfortunately, such datasets are rarely collected due to the cost of high fidelity monitoring, and the data that is available is generally not ideal, often consisting of unbalanced classes, noisy data due to environmental conditions, and few samples. In this paper, we propose an alternate approach that utilizes physics-based simulations of string-level IV curves as a fully synthetic training corpus that is independent of the test dataset. In our example, the training corpus consists of baseline (no fault), partial soiling, and cell crack system modes. The training corpus is used to train a 1D convolutional neural network (CNN) for failure classification. The approach is validated by comparing the model’s ability to classify failures detected on a real, measured IV curve testing corpus obtained from laboratory and field experiments. Results obtained using a fully synthetic training dataset achieve identical accuracy to those obtained with use of a measured training dataset. When evaluating the measured data’s test split, a 100% accuracy was found both when using simulations or measured data as the training corpus. When evaluating all of the measured data, a 96% accuracy was found when using a fully synthetic training dataset. The use of physics-based modeling results as a training corpus for failure detection and classification has many advantages for implementation as each PV system is configured differently, and it would be nearly impossible to train using labeled measured data.

Hopwood, Michael W. (ORCID:0000000161901767)↗
NASA NTRS1994

Knee-Joint Proprioception During 30-Day 6 deg Head-Down Bed Rest with Isotonic and Isokinetic Exercise Training

To determine if daily isotonic exercise or isokinetic exercise training coupled with daily log proprioceptive training, would influence log proprioceptive tracking responses during Bed Rest (BR), 19 men (36 +/- SD 4 years, 178 +/- 7 cm, 76.8 +/- 7.8 kg) were allocated into a NO-Exercise (NOE) training control group (n = 5), and IsoTanic Exercise (ITE, n = 7) and IsoKinetic Exercise (IKE, n = 7) training groups. Exercise training was conducted during BR for two 30-min period / d, 5 d /week. Only the IKE group performed proprioceptive training using a now isokinetic procedure with each lower extremity for 2.5 min before and after the daily exercise training sessions; proprioceptive testing occurred weekly for all groups. There were no significant differences in proprioceptive tracking scores, expressed as a percentage of the perfect score of 100, in the pro-BR ambulatory control period between the three groups. Knee extension and flexion tracking responses were unchanged with NOE during BR, but were significantly greater (*p less than 0.05) at the end of BR in both exercise groups when compared with NOE responses (extension: NOE 80.7 +/- 0.7%, ITE 82.9 +/- 0.6%, IKE 86.5* +/- 0.7%; flexion: NOE 77.6 +/- 1.50, ITE 80.0 +/- 0.8% (NS), IKE 83.6* +/- 0.8%). Although proprioceptive tracking was unchanged during BR with NOE, both lsotonic exercise training (without additional propriaceptive training) and especially isokinetic exercise training when combined with daily proprioceptive training, significantly improved knee proprioceptive tracking responses after 30 d of BR.

Bernauer, E. M.↗
NASA NTRS1997

Computer Based Training: Field Deployable Trainer and Shared Virtual Reality

Astronaut training has traditionally been conducted at specific sites with specialized facilities. Because of its size and nature the training equipment is generally not portable. Efforts are now under way to develop training tools that can be taken to remote locations, including into orbit. Two of these efforts are the Field Deployable Trainer and Shared Virtual Reality projects. Field Deployable Trainer NASA has used the recent shuttle mission by astronaut Shannon Lucid to the Russian space station, Mir, as an opportunity to develop and test a prototype of an on-orbit computer training system. A laptop computer with a customized user interface, a set of specially prepared CD's, and video tapes were taken to the Mir by Ms. Lucid. Based upon the feedback following the launch of the Lucid flight, our team prepared materials for the next Mir visitor. Astronaut John Blaha will fly on NASA/MIR Long Duration Mission 3, set to launch in mid September. He will take with him a customized hard disk drive and a package of compact disks containing training videos, references and maps. The FDT team continues to explore and develop new and innovative ways to conduct offsite astronaut training using personal computers. Shared Virtual Reality Training NASA's Space Flight Training Division has been investigating the use of virtual reality environments for astronaut training. Recent efforts have focused on activities requiring interaction by two or more people, called shared VR. Dr. Bowen Loftin, from the University of Houston, directs a virtual reality laboratory that conducts much of the NASA sponsored research. I worked on a project involving the development of a virtual environment that can be used to train astronauts and others to operate a science unit called a Biological Technology Facility (BTF). Facilities like this will be used to house and control microgravity experiments on the space station. It is hoped that astronauts and instructors will ultimately be able to share common virtual environments and, using telephone links, conduct interactive training from separate locations.

Mullen, Terence J.↗
NASA NTRS2000

Stress Counter-Response Training Via Physiological Self-Regulation During Flight Simulation

This study provided the first evaluation of a new training concept and technology aimed at training pilots to maintain physiological equilibrium during circumstances in an airplane cockpit. Thirty healthy subjects (16 males and 14 females) between the ages of 18 and 35 were randomized into two study groups, A and B. Subjects participated individually in a sequence of four study sessions. In the first visit, subjects were taught to operate a desktop fighter jet flight simulation program. In the three sessions that followed, subjects in group A were trained to minimize their autonomic deviation from baseline values while operating the desktop flight simulation. This was done by making their skin conductance and hand temperature deviations from baseline impair the functionality of the aircraft controls. Subjects also received auditory and visual cues about their autonomic deviation, and were instructed to keep these within pre-set limits to retain full control of the aircraft. Subjects in group B were subjected to periods of impaired aircraft functionality independent of their physiologic activity, and thus served as a control group. No statistically significant group differences were found in the flight performance scores from the three training sessions, and post-training flight performance scores of the two groups were not different. We conclude that this study did not provide clear support for this training methodology in optimizing pilot performance. However, a number of shortcomings in the current status of this training methodology may account for the lack of demonstrable training benefit to the experimental group. Suggested future modifications for research on this training methodology include: Limiting the amount of instrument impairment resulting from physiological deviations; conducting a greater number of physiological training sessions per subject; using pre-post training performance tests which invoke a greater amount of stress in subjects; and developing a more detailed performance scoring system.

Palsson, Olafur S.↗
NASA NTRS2000

Isometric handgrip training reduces arterial pressure at rest without changes in sympathetic nerve activity

The purpose of this study was to determine whether isometric handgrip (IHG) training reduces arterial pressure and whether reductions in muscle sympathetic nerve activity (MSNA) mediate this drop in arterial pressure. Normotensive subjects were assigned to training (n = 9), sham training (n = 7), or control (n = 8) groups. The training protocol consisted of four 3-min bouts of IHG exercise at 30% of maximal voluntary contraction (MVC) separated by 5-min rest periods. Training was performed four times per week for 5 wk. Subjects' resting arterial pressure and heart rate were measured three times on 3 consecutive days before and after training, with resting MSNA (peroneal nerve) recorded on the third day. Additionally, subjects performed IHG exercise at 30% of MVC to fatigue followed by muscle ischemia. In the trained group, resting diastolic (67 +/- 1 to 62 +/- 1 mmHg) and mean arterial pressure (86 +/- 1 to 82 +/- 1 mmHg) significantly decreased, whereas systolic arterial pressure (116 +/- 3 to 113 +/- 2 mmHg), heart rate (67 +/- 4 to 66 +/- 4 beats/min), and MSNA (14 +/- 2 to 15 +/- 2 bursts/min) did not significantly change following training. MSNA and cardiovascular responses to exercise and postexercise muscle ischemia were unchanged by training. There were no significant changes in any variables for the sham training and control groups. The results indicate that IHG training is an effective nonpharmacological intervention in lowering arterial pressure.

NASA Discipline Neuroscience↗
NASA NTRS1994

Knee-joint proprioception during 30-day 6 degrees head-down bed rest with isotonic and isokinetic exercise training

To determine if daily isotonic exercise or isokinetic exercise training coupled with daily leg proprioceptive training, would influence leg proprioceptive tracking responses during bed rest (BR), 19 men (36 +/- SD 4 years, 178 +/- 7 cm, 76.8 +/- 7.8 kg) were allocated into a no-exercise (NOE) training control group (n = 5), and isotonic exercise (ITE, n = 7) and isokinetic exercise (IKE, n = 7) training groups. Exercise training was conducted during BR for two 30-min periods.d-1, 5 d.week-1. Only the IKE group performed proprioceptive training using a new isokinetic procedure with each lower extremity for 2.5 min before and after the daily exercise training sessions; proprioceptive testing occurred weekly for all groups. There were no significant differences in proprioceptive tracking scores, expressed as a percentage of the perfect score of 100, in the pre-BR ambulatory control period between the three groups. Knee extension and flexion tracking responses were unchanged with NOE during BR, but were significantly greater (*p < 0.05) at the end of BR in both exercise groups when compared with NOE responses (extension: NOE 80.7 +/- 0.7%, ITE 82.9* +/- 0.6%, IKE 86.5* +/- 0.7%; flexion: NOE 77.6 +/- 1.5%, ITE 80.0 +/- 0.8% (NS), IKE 83.6* +/- 0.8%). Although proprioceptive tracking was unchanged during BR with NOE, both isotonic exercise training (without additional proprioceptive training) and especially isokinetic exercise training when combined with daily proprioceptive training, significantly improved knee proprioceptive tracking responses after 30 d of BR.

Controlled Clinical Trial↗
NASA NTRS2007

Development of a Human Behavior and Performance Training Curriculum for ISS Astronauts

The paper will describe the DACUM process and summarize the core competencies that were agreed upon, internationally, as important for ISS astronauts. The paper will further discuss the ongoing work being completed by the subgroup, Human Behaviour and Performance Training Working Group, including defining the competencies and behavioural markers. Finally, an overview of remaining work will be provided, including determining which competencies require formal training and which require no formal training, developing training objectives, sequencing the training, and establishing how to assess training effectiveness. DISCUSSION: Designing a common set of goals for behavioural training has been the desire of the SHBP WG since its inception in 1998. This group, along with training specialists and astronauts, are making great strides toward defining these competencies. The road ahead will be exceedingly challenging as training objectives are defined and a training flow is proposed to the MCOP; with proposed ISS crews increasing to six people in the near future, such enhanced behavioural training may be all the more essential for mission success.

VanderArk, Steve↗
NASA NTRS2008

Human Factors in Training

Future space missions will be significantly longer than current Shuttle missions and new systems will be more complex than current systems. Increasing communication delays between crews and Earth-based support means that astronauts need to be prepared to handle the unexpected on their own. As crews become more autonomous, their potential span of control and required expertise must grow to match their autonomy. It is not possible to train for every eventuality ahead of time on the ground, or to maintain trained skills across long intervals of disuse. To adequately prepare NASA personnel for these challenges, new training approaches, methodologies, and tools are required. This research project aims at developing these training capabilities. Training efforts in FY07 strongly focused on crew medical training, but also began exploring how Space Flight Resource Management training for Mission Operations Directorate (MOD) Flight Controllers could be integrated with systems training for optimal Mission Control Center operations. Beginning in January 2008, the training research effort will include team training prototypes and tools. The Training Task addresses Program risks that lie at the intersection of the following three risks identified by the Project: 1) Risk associated with poor task design; 2) Risk of error due to inadequate information; 3) Risk associated with reduced safety and efficiency due to poor human factors design.

Barshi, Immanuel↗
NASA NTRS2010

The Right Stuff: A Look Back at Three Decades of Flight Controller Training for Space Shuttle Mission Operations

This paper will summarize the thirty-year history of Space Shuttle operations from the perspective of training in NASA Johnson Space Center's Mission Control Center. It will focus on training and development of flight controllers and instructors, and how training practices have evolved over the years as flight experience was gained, new technologies developed, and programmatic needs changed. Operations of human spaceflight systems is extremely complex, therefore the training and certification of operations personnel is a critical piece of ensuring mission success. Mission Control Center (MCC-H), at the Lyndon B. Johnson Space Center, in Houston, Texas manages mission operations for the Space Shuttle Program, including the training and certification of the astronauts and flight control teams. This paper will give an overview of a flight control team s makeup and responsibilities during a flight, and details on how those teams are trained and certified. The training methodology for developing flight controllers has evolved significantly over the last thirty years, while the core goals and competencies have remained the same. In addition, the facilities and tools used in the control center have evolved. These changes have been driven by many factors including lessons learned, technology, shuttle accidents, shifts in risk posture, and generational differences. Flight controllers will share their experiences in training and operating the Space Shuttle throughout the Program s history. A primary method used for training Space Shuttle flight control teams is by running mission simulations of the orbit, ascent, and entry phases, to truly "train like you fly." The audience will learn what it is like to perform a simulation as a shuttle flight controller. Finally, we will reflect on the lessons learned in training for the shuttle program, and how those could be applied to future human spaceflight endeavors.

Dittemore, Gary D.↗
NASA NTRS2015

The Evolution of On-Board Emergency Training for the International Space Station Crew

The crew of the International Space Station (ISS) receives extensive ground‐training in order to safely and effectively respond to any potential emergency event while on‐orbit, but few people realize that their training is not concluded when they launch into space. The evolution of the emergency On‐ Board Training events (OBTs) has recently moved from paper "scripts" to an intranet‐based software simulation that allows for the crew, as well as the flight control teams in Mission Control Centers across the world, to share in an improved and more realistic training event. This emergency OBT simulator ensures that the participants experience the training event as it unfolds, completely unaware of the type, location, or severity of the simulated emergency until the scenario begins. The crew interfaces with the simulation software via iPads that they keep with them as they translate through the ISS modules, receiving prompts and information as they proceed through the response. Personnel in the control centers bring up the simulation via an intranet browser at their console workstations, and can view additional telemetry signatures in simulated ground displays in order to assist the crew and communicate vital information to them as applicable. The Chief Training Officers and emergency instructors set the simulation in motion, choosing the type of emergency (rapid depressurization, fire, or toxic atmosphere) and specific initial conditions to emphasize the desired training objectives. Project development, testing, and implementation was a collaborative effort between ISS emergency instructors, Chief Training Officers, Flight Directors, and the Crew Office using commercial off the shelf (COTS) hardware along with simulation software created in‐house. Due to the success of the Emergency OBT simulator, the already‐developed software has been leveraged and repurposed to develop a new emulator used during fire response ground‐training to deliver data that the crew receives from the handheld Compound Specific Analyzer for Combustion Products (CSA‐CP). This CSA‐CP emulator makes use of a portion of codebase from the Emergency OBT simulator dealing with atmospheric contamination during fire scenarios, and feeds various data signatures to crew via an iPod Touch with a flight‐like CSA‐CP display. These innovative simulations, which make use of COTS hardware with custom in‐house software, have yielded drastic improvements to emergency training effectiveness and risk reduction for ISS crew and flight control teams during on‐orbit and ground training events.

LaBuff, Skyler↗
NASA NTRSDate not supplied

Shoulder Postures in EVA Training in Reduced Gravity Analogues

Shoulder Postures in EVA Training in Reduced Gravity Analogues K. Guhl1, L. Vu2, H. Kim3, S. Rajulu4 1KBR Inc., Houston, TX, 2Aegis Aerospace Inc., Houston, TX, 3Leidos Innovations, Houston, TX, 4NASA Johnson Space Center, Houston, TX. During extravehicular activities (EVAs) and EVA training in both the Neutral Buoyancy Laboratory (NBL) and at the Active Response Gravity Offload System (ARGOS), crewmembers perform a variety of hand-intensive tasks with frequent arm/shoulder repositioning while wearing a pressurized spacesuit. As a result, crewmembers may experience ergonomic stressors such as awkward shoulder postures. The ergonomic shoulder risk is also compounded by limited or restricted shoulder mobility of the spacesuit, extreme work positions such as overhead tasks, and tasks with heavy tools and repetitive motions. Prolonged or frequent shoulder elevation and overhead work, in particular, can lead to excessive stresses and musculoskeletal injuries of the shoulder joints. Future EVA missions, specifically lunar surface EVAs, will also be longer in duration and thus increase the exposure to awkward shoulder postures. In this study, we aimed to assess the ergonomic risk of awkward shoulder postures in simulated lunar surface EVAs by quantifying when and how long the arms are raised above the chest level. We assessed video recordings of pilot lunar EVA simulations (3 lunar trials each in the NBL and at ARGOS and 2 microgravity EVA trials in the NBL). These runs consisted of both training and engineering test objectives. Actual demands for shoulder use varied for different EVA types and analogues, but many EVA runs have common tasks and require similar motion components. A video observation and event logging software was used to document the duration and occurrence of the subject’s arms being raised throughout the video recordings of each EVA run. An arm raised instance was classified as the arm being at 90 degrees or above with relation to gravity for lunar EVA training events and with relation to the body for microgravity EVA training events. Such events included EVA hardware maintenance or heavy geology sampling tool operations. Events where the arm load was partially supported by external objects, like climbing a ladder or leaning against a surface were separately identified and excluded. Statistical analysis was performed to summarize the timing, frequency, and durations of the arm raise events and compared across the different EVA tasks and analogue types. Preliminary observations indicated that the total duration and number of arm raised instances were surprisingly smaller for lunar surface EVA training as compared to microgravity EVA training. The observed difference may be attributed to differing task demands and unique environmental characteristics found in lunar EVAs in comparison to microgravity EVAs. A detailed statistical analysis will be performed between lunar surface EVA training events and microgravity EVA training events in the final submission. Overall, this analysis is expected to provide insight into how ergonomic recommendations can be refined for lunar EVA training with pressurized suits. It may also inform task design and influence suit padding design to better protect crewmembers during future lunar training.

Kaitlyn Lea Guhl↗
NASA NTRSDate not supplied

GL4U: Using Space Biology Omics Data to Provide Bioinformatics Training for Students and Educators

NASA’s GeneLab project provides researchers open access to space-relevant multi-omics data via the Open Science Data Repository (OSDR) that can be mined to understand the effects of spaceflight on biological systems. To maximize the number of scientists who understand and utilize GeneLab data and data processing pipelines, GeneLab created GeneLab for Colleges and Universities (GL4U). GL4U provides space biology-relevant training in bioinformatics to the next generation of scientists through direct (training students) and indirect (training educators) approaches. The GL4U pilot programs were conducted in June 2021 (direct training) and 2022 (indirect training). During the pilots, students and educators at Historically Black Colleges and Universities (HBCUs) and Minority Serving Institutions (MSIs) participated in a week-long (direct training) or two-week-long (indirect training) bootcamp consisting of space biology-specific lectures and hands-on instruction using Jupyter Notebooks to analyze space biology RNA sequencing data from OSDR. During the educator pilot, participants received materials, training, and the necessary compute resources to enable them to run the bootcamp at their home institutions, thereby extending the reach of this initiative. In July 2023, GeneLab is partnering with JPL to expand GL4U to include amplicon sequencing (Amp-Seq) analysis training. During the GL4U Amp-Seq bootcamp, student and educator participants will receive training on how to analyze and interpret Amp-Seq data using the NASA GeneLab data processing pipeline. All bootcamp material, including instructions for requesting compute resources, will be made publicly available on GitHub for educators to teach the GL4U content in subsequent semesters. GL4U provides undergraduate students from underrepresented groups the opportunity to learn about NASA and Space Biology, and to enhance their career prospects by gaining hands-on experience analyzing omics data, a skillset that is highly applicable and marketable in the life sciences. We present results from pre- and post-training surveys completed by all participants of the Amp-Seq bootcamp.

Amanda M Saravia-Butler↗
NASA NTRS1992

Human torque velocity adaptations to sprint, endurance, or combined modes of training

We had groups of athletes perform sprint and endurance run training independently or concurrently for 8 weeks to examine the voluntary in vivo mechanical responses to each type of training. Pre- and posttraining angle-specific peak torque during knee extension and flexion were determined at 0, 0.84, 1.65, 2.51, 3.35, 4.19, and 5.03 radian.sec-1 and normalized for lean body mass. Knee extension torque in the sprint-trained group increased across all test velocities, the endurance-trained group increased at 2.51, 3.34, 4.19, and 5.03 radian.sec-1, and the group performing the combined training showed no change at any velocity. Knee flexion torque of the sprint and combined groups decreased at 0.84, 1.65, and 2.51 radian.sec-1. Knee flexion torque in the sprint-trained group also decreased at 0 radian.sec-1 and in the combined group at 3.34 radian.sec-1. Knee flexion torque in the endurance-trained group showed no change at any velocity of contraction. Mean knee flexion:extension ratios across the test velocities significantly decreased in the sprint-trained group. Knee extension endurance during 30 seconds of maximal contractions significantly increased in all groups. Only the sprint-trained group showed a significant increase in endurance of the knee flexors. These data suggest that changes in the voluntary in vivo mechanical characteristics of knee extensor and flexor skeletal muscles are specific to the type of run training performed.

NASA Discipline Musculoskeletal↗
DOE OSTI2020

Active Learning A Neural Network Model For Gold Clusters & Bulk From Sparse First Principles Training Data

Small metal clusters are of fundamental scientific interest and of tremendous significance in catalysis. These nanoscale clusters display diverse geometries and structural motifs depending on the cluster size; a knowledge of this size-dependent structural motifs and their dynamical evolution has been of longstanding interest. Given the high computational cost of first-principles calculations, molecular modeling and atomistic simulations such as molecular dynamics (MD) has proven to be an important complementary tool to aid this understanding. Classical MD typically employ predefined functional forms which limits their ability to capture such complex size-dependent structural and dynamical transformation. Neural Network (NN) based potentials represent flexible alternatives and in principle, well-trained NN potentials can provide high level of flexibility, transferability and accuracy on-par with the reference model used for training. A major challenge, however, is that NN models are interpolative and requires large quantities (similar to 10 4 or greater) of training data to ensure that the model adequately samples the energy landscape both near and far-from-equilibrium. A highly desirable goal is minimize the number of training data, especially if the underlying reference model is first-principles based and hence expensive. In this work, we introduce an active learning (AL) scheme that trains a NN model on-the-fly with minimal amount of first-principles based training data. Our AL workflow is initiated with a sparse training dataset (similar to 1 to 5 data points) and is updated on-the-fly via a Nested Ensemble Monte Carlo scheme that iteratively queries the energy landscape in regions of failure and updates the training pool to improve the network performance. Using a representative system of gold clusters, we demonstrate that our AL workflow can train a NN with similar to 500 total reference calculations. Using an extensive DFT test set of similar to 1100 configurations, we show that our AL-NN is able to accurately predict both the DFT energies and the forces for clusters of a myriad of different sizes. Our NN predictions are within 30 meV/atom and 40 meV/angstrom of the reference DFT calculations. Moreover, our AL-NN model also adequately captures the various size-dependent structural and dynamical properties of gold clusters in excellent agreement with DFT calculations and available experiments. We finally show that our AL-NN model also captures bulk properties reasonably well, even though they were not included in the training data.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

The impacts of training on change deafness and build-up in a flicker task

Performance on auditory change detection tasks can be improved by training. We examined the stimulus specificity of these training effects in behavior and ERPs. A flicker change detection task was employed in which spatialized auditory scenes were alternated until a "change" or "same" response was made. For half of the trials, scenes were identical. The other half contained changes in the spatial locations of objects from scene to scene. On Day 1, participants were either trained on this auditory change detection task (trained group), or trained on a non-auditory change detection task (control group). On Day 2, all participants were tested on the flicker task while EEG was recorded. The trained group showed greater change detection accuracy than the control group. They were less biased to respond "same" and showed full generalization of learning from trained to novel auditory objects. ERPs for "change" compared to "same" trials showed more negative going P1, N1, and P2 amplitudes, as well as a larger P3b amplitude. The P3b amplitude also differed between the trained and control group, with larger amplitudes for the trained group. Analysis of ERPs to scenes viewed prior to a decision revealed build-up of a difference between "change" and "same" trials in N1 and P2. Results demonstrate that training has an impact early in the "same" versus "change" decision-making process, and that the flicker paradigm combined with the ERP method can be used to study the build-up of change detection in auditory scenes.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2020

How Low Can You Go? Using Synthetic 3D Imagery to Drastically Reduce Real-World Training Data for Object Detection

Deep convolutional neural networks (DCNNs) currently provide state-of-the-art performance on image classification and object detection tasks, and there are many global security mission areas where such models could be extremely useful. Crucially, the success of these models is driven in large part by the widespread availability of high-quality open source data sets such as Image Net, Common Objects in Context (COCO), and KITTI, which contain millions of images with thousands of unique labels. However, global security relevant objects-of-interest can be difficult to obtain: relevant events are low frequency and high consequence; the content of relevant images is sensitive; and adversaries and proliferators seek to obscure their activities. For these cases where exemplar data is hard to come-by, even fine-tuning an existing model with available data can be effectively impossible. Recent work demonstrated that models can be trained using a combination of real-world and synthetic images generated from 3D representations; that such models can exceed the performance of models trained using real-world data alone; and that the generated images need not be perfectly realistic (Tremblay, et al., 2018). However, this approach still required hundreds to thousands of real-world images for training and fine tuning, which for sparse, global security-relevant datasets can be an unrealistic hurdle. In this research, we validate the performance and behavior of DCNN models as we drive the number of real-world images used for training object detection tasks down to a minimal set. We perform multiple experiments to identify the best approach to train DCNNs from an extremely small set of real-world images. In doing so, we: Develop state-of-the-art, parameterized 3D models based on real-world images and sample from their parameters to increase the variance in synthetic image training data; Use machine learning explainability techniques to highlight and correct through targeted training the biases that result from training using completely synthetic images; and Validate our results by comparing the performance of the models trained on synthetic data to one another, and to a control model created by fine-tuning an existing ImageNet-trained model with a limited number (hundreds) of real-world images.

97 MATHEMATICS AND COMPUTING↗