Search NASA⌕ Search

SEARCH · Search NASA

Results for “training effectiveness”

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 109 records · Page 6

Multi-fidelity learning for interatomic potentials: low-level forces and high-level energies are all you need

The promise of machine learning interatomic potentials (MLIPs) has led to an abundance of public quantum mechanical (QM) training datasets. The quality of an MLIP is directly limited by the accuracy of the energies and atomic forces in the training dataset. Unfortunately, most of these datasets are computed with relatively low-accuracy QM methods, e.g. density functional theory with a moderate basis set. Due to the increased computational cost of more accurate QM methods, e.g. coupled-cluster theory with a complete basis set (CBS) extrapolation, most high-accuracy datasets are much smaller and often do not contain atomic forces. The lack of high-accuracy atomic forces is quite troubling, as training with force data greatly improves the stability and quality of the MLIP compared to training to energy alone. Because most datasets are computed with a unique level of theory, traditional single-fidelity (SF) learning is not capable of leveraging the vast amounts of published QM data. In this study, we apply multi-fidelity learning (MFL) to train an MLIP to multiple QM datasets of different levels of accuracy, i.e. levels of fidelity. Specifically, we perform three test cases to demonstrate that MFL with both low-level forces and high-level energies yields an extremely accurate MLIP—far more accurate than a SF MLIP trained solely to high-level energies and almost as accurate as a SF MLIP trained directly to high-level energies and forces. Therefore, MFL greatly alleviates the need for generating large and expensive datasets containing high-accuracy atomic forces and allows for more effective training to existing high-accuracy energy-only datasets. Indeed, low-accuracy atomic forces and high-accuracy energies are all that are needed to achieve a high-accuracy MLIP with MFL.

36 MATERIALS SCIENCE↗

Dynamic training devices in CRM training

Pilot training effectiveness and flying safety of a seasonal tour flight company are described. The change from single pilot to two pilot operated twin otters is examined. The use of the ATC 810 training device, its possibilities and training capacity is outlined. Problem areas which may arise, emergency system and pilot/passenger interaction are analyzed.

Lawver, J.↗

Primal-Dual Differentiable Programming for Distribution System Critical Load Restoration: Preprint

Swift and reliable critical load restoration (CLR) can help make a distribution system resilient towards extreme events. To optimally achieve that, alongside practical concerns such as limiting online computational burden, some studies leverage model-free reinforcement learning (RL) to train control policies. Despite the advantages provided by RL algorithms, these approaches suffer from two issues: 1) the lack of a proper mechanism for constraint enforcement, and 2) poor sample efficiency. Therefore, in this paper, a primal-dual differentiable programming (PDDP) method is developed for guiding the training leading to a constraint-satisfying policy. Additionally, the model-based nature of the proposed method aims at improving sample efficiency. The experiment on a CLR problem demonstrates that PDDP can effectively train a control policy that both achieves desirable performance and satisfies required constraints.

differentiable programming↗

Initial Testing the Exploration Extravehicular Mobility Unit (xEMU) in Lunar Environment Simulation at the Neutral Buoyancy Lab (NBL) in 2022

Understanding how to effectively train for Extravehicular Activities (EVAs) for Artemis missions is critical. Tasked by the new Extravehicular Activity and Human Surface Mobility Program(EHP), a team of stakeholders from engineering and flight operations worked together to test the Exploration Extravehicular Mobility (xEMU) in the Neutral Buoyancy Lab(NBL) simulating the lunar environment. The NBL is a pool in which astronauts are typically trained to complete simulated International Space Station (ISS) EVA tasks in a neutral weigh-out. The overall focus for this test series was evaluating how well the NBL could be used as an effective simulated environment for Artemis lunar EVAs. New NBL support hardware was manufactured to support this test series and update the interfaces to be more representative of xEMU hardware such as a new NBL Portable Life Support System(PLSS)mock-up and umbilical, lights and cameras, and display and control unit (DCU).To simulate the 1/6th-gravity (1/6-g) environment, extra weights were added around the suit to create an accurate center of gravity representation. The Partial Gravity Weigh-out Stand (PGWS) predicted where and how much weight would be required for the 1/6th-G simulation and predicted the center of gravity placement. After the weigh-out attempt was completed, the suited participant would complete a set of directed tasks such as kneeling, object pick-up, and walking to evaluate the weigh-out. To create the lunar environment, sand, boulders, and ramps were added to the pool floor. If time allowed after weigh-out was determined acceptable, the flight operations team would direct the suited participants to complete simulated Artemis EVA tasks using flight-like tools and tool cart. The team also evaluated turning out the overhead lights in the NBL for a portion of some of the test runs and using a solar simulator to create amore realistic lunar south pole lighting environment. The test series completed one unmanned test and nine manned tests. With four of the tests, dual xEMU suited operations were completed with subjects in both the small and large xEMU sizes. Overall, the NBL was able to create an acceptable lunar landscape and ability to have two subjects work side by side in pressurized xEMUs which currently cannot be achieved in other test environments. Achieving a stable 1/6th-G lunar weigh-out and clear, consistent dual suit communications was a challenge for this series. Future work is planned to continue to improve this simulation environment.

spacesuit↗

Initial Testing of the Exploration Extravehicular Mobility Unit (xEMU) in Lunar Environment Simulation at the Neutral Buoyancy Lab (NBL) in 2022

Understanding how to effectively train for Extravehicular Activities (EVAs) for Artemis missions is critical. Tasked by the new Extravehicular Activity and Human Surface Mobility Program(EHP), a team of stakeholders from engineering and flight operations worked together to test the Exploration Extravehicular Mobility (xEMU) in the Neutral Buoyancy Lab(NBL) simulating the lunar environment. The NBL is a pool in which astronauts are typically trained to complete simulated International Space Station (ISS) EVA tasks in a neutral weigh-out. The overall focus for this test series was evaluating how well the NBL could be used as an effective simulated environment for Artemis lunar EVAs. New NBL support hardware was manufactured to support this test series and update the interfaces to be more representative of xEMU hardware such as a new NBL Portable Life Support System(PLSS)mock-up and umbilical, lights and cameras, and display and control unit (DCU).To simulate the 1/6th-gravity (1/6-g) environment, extra weights were added around the suit to create an accurate center of gravity representation. The Partial Gravity Weigh-out Stand (PGWS) predicted where and how much weight would be required for the 1/6th-G simulation and predicted the center of gravity placement. After the weigh-out attempt was completed, the suited participant would complete a set of directed tasks such as kneeling, object pick-up, and walking to evaluate the weigh-out. To create the lunar environment, sand, boulders, and ramps were added to the pool floor. If time allowed after weigh-out was determined acceptable, the flight operations team would direct the suited participants to complete simulated Artemis EVA tasks using flight-like tools and tool cart. The team also evaluated turning out the overhead lights in the NBL for a portion of some of the test runs and using a solar simulator to create amore realistic lunar south pole lighting environment. The test series completed one unmanned test and nine manned tests. With four of the tests, dual xEMU suited operations were completed with subjects in both the small and large xEMU sizes. Overall, the NBL was able to create an acceptable lunar landscape and ability to have two subjects work side by side in pressurized xEMUs which currently cannot be achieved in other test environments. Achieving a stable 1/6th-G lunar weigh-out and clear, consistent dual suit communications was a challenge for this series. Future work is planned to continue to improve this simulation environment.

spacesuit↗

Multi-task graph neural networks for simultaneous prediction of global and atomic properties in ferromagnetic systems *

Abstract We introduce a multi-tasking graph convolutional neural network, HydraGNN, to simultaneously predict both global and atomic physical properties and demonstrate with ferromagnetic materials. We train HydraGNN on an open-source ab initio density functional theory (DFT) dataset for iron-platinum with a fixed body centered tetragonal lattice structure and fixed volume to simultaneously predict the mixing enthalpy (a global feature of the system), the atomic charge transfer, and the atomic magnetic moment across configurations that span the entire compositional range. By taking advantage of underlying physical correlations between material properties, multi-task learning (MTL) with HydraGNN provides effective training even with modest amounts of data. Moreover, this is achieved with just one architecture instead of three, as required by single-task learning (STL). The first convolutional layers of the HydraGNN architecture are shared by all learning tasks and extract features common to all material properties. The following layers discriminate the features of the different properties, the results of which are fed to the separate heads of the final layer to produce predictions. Numerical results show that HydraGNN effectively captures the relation between the configurational entropy and the material properties over the entire compositional range. Overall, the accuracy of simultaneous MTL predictions is comparable to the accuracy of the STL predictions. In addition, the computational cost of training HydraGNN for MTL is much lower than the original DFT calculations and also lower than training separate STL models for each property.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Integrated Resistance and Aerobic Training Study - Sprint

Space flight causes reductions in fitness/health: (1) Cardiovascular -- reduced VO2max, cardiac output (2) Bone -- reduced bone mineral density (3) Muscle -- reduced mass, strength and endurance. Exercise is the primary countermeasure to protect against these changes and was made operational before completely mature. Research continues to identify most effective/efficient exercise programs. Crew medical tests (cardio, muscle, bone) do not yield sufficient information to fine tune the effectiveness of exercise programs, thus there is a need for more detailed testing aimed at identifying the most effective training program. The objective of this program was to obtain detailed information about crew physical fitness pre-and post-flight and evaluate new evidence based exercise prescription with higher intensity, lower duration and frequency.

Ploutz-Snyder, Lori↗

Layer-Parallel Training of Deep Residual Neural Networks

Residual neural networks (ResNets) are a promising class of deep neural networks that have shown excellent performance for a number of learning tasks, e.g., image classification and recognition. Mathematically, ResNet architectures can be interpreted as forward Euler discretizations of a nonlinear initial value problem whose time-dependent control variables represent the weights of the neural network. Hence, training a ResNet can be cast as an optimal control problem of the associated dynamical system. For similar time-dependent optimal control problems arising in engineering applications, parallel-in-time methods have shown notable improvements in scalability. This paper demonstrates the use of those techniques for efficient and effective training of ResNets. The proposed algorithms replace the classical (sequential) forward and backward propagation through the network layers with a parallel nonlinear multigrid iteration applied to the layer domain. This adds a new dimension of parallelism across layers that is attractive when training very deep networks. From this basic idea, we derive multiple layer-parallel methods. The most efficient version employs a simultaneous optimization approach where updates to the network parameters are based on inexact gradient information in order to speed up the training process. Finally, using numerical examples from supervised classification, we demonstrate that the new approach achieves a training performance similar to that of traditional methods, but enables layer-parallelism and thus provides speedup over layer-serial methods through greater concurrency.

97 MATHEMATICS AND COMPUTING↗

Analyzing and Exploring Training Recipes for Large-Scale Transformer-Based Weather Prediction

Abstract The rapid rise of deep learning (DL) in numerical weather prediction (NWP) has led to a proliferation of models which forecast atmospheric variables with comparable or superior skill than traditional physics-based NWP. However, among these leading DL models, there is a wide variance in both the training settings and architecture used. Further, the lack of thorough ablation studies makes it hard to discern which components are most critical to success. In this work, we show that it is possible to attain high forecast skill even with relatively off-the-shelf architectures, simple training procedures, and moderate compute budgets. Specifically, we train a minimally modified Swin Transformer V2 (SwinV2) on ERA5 data and find that it attains superior skill in terms of mean-square errors of deterministic forecasts when compared against the European Centre for Medium-Range Weather Forecasts’ Integrated Forecasting System (IFS). Almost all DL–NWP systems share a core set of hyperparameters and design decisions. To aid and expedite future DL–NWP research, we present an in-depth, systematic exploration of different loss functions, model sizes and depths, patch sizes, and multistep training objectives. We also examine the model performance with metrics beyond the typical accuracy (ACC) and RMSE and investigate how the performance scales with model size. Through our open-source code, scoring pipelines, and models, we share our findings on key aspects of the training pipeline. These ablations reduce the necessity for expensive hyperparameter tuning and lower the barrier to entry for future DL–NWP research. Significance Statement This study investigates the potential of using large-scale transformer-based models for weather prediction, showing that it is possible to achieve high forecast accuracy with simpler, off-the-shelf architectures. By training a minimally modified SwinV2 transformer on ERA5 data, we show that the model achieves competitive forecast skill in terms of mean-square error for key variables, outperforming the European Centre for Medium-Range Weather Forecasts’ Integrated Forecasting System (IFS) at all lead times. Our findings suggest that effective training strategies, such as multistep fine-tuning and channel-weighted losses, significantly enhance the model’s performance. However, we also highlight that these improvements come with trade-offs in other areas, such as ensemble spread and high-frequency spatial detail. This work highlights the promise of deep learning in improving weather forecasts, which could lead to better preparedness and response to weather events, ultimately benefiting society by providing more reliable weather predictions.

Willard, Jared D. [Lawrence Berkeley National Labo↗

Representing Small Commercial Building Faults in EnergyPlus, Part II: Model Validation

Automated fault detection and diagnosis (AFDD) tools based on machine-learning algorithms hold promise for lowering cost barriers for AFDD in small commercial buildings; however, access to high-quality training data for such algorithms is often difficult to obtain. To fill the gap in this research area, this study covers the development (Part I) and validation (Part II) of fault models that can be used with the building energy modeling software EnergyPlus® and OpenStudio® to generate a cost-effective training data set for developing AFDD algorithms. Part II (this paper) first presents a methodology of validating fault models with OpenStudio and then presents validation results, which are compared against measurements from a reference building. We discuss the results of our experiments with eight different faults in the reference building (a total of 39 different baseline and faulted scenarios), including our methodology for using fault models along with the reference building model to simulate the same faulted scenarios. Then, we present validation of the fault models by comparing results of simulations and experiments either quantitatively or qualitatively.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Learning to train neural networks for real-world control problems

Over the past three years, our group has concentrated on the application of neural network methods to the training of controllers for real-world systems. This presentation describes our approach, surveys what we have found to be important, mentions some contributions to the field, and shows some representative results. Topics discussed include: (1) executing model studies as rehearsal for experimental studies; (2) the importance of correct derivatives; (3) effective training with second-order (DEKF) methods; (4) the efficacy of time-lagged recurrent networks; (5) liberation from the tyranny of the control cycle using asynchronous truncated backpropagation through time; and (6) multistream training for robustness. Results from model studies of automotive idle speed control serve as examples for several of these topics.

Feldkamp, Lee A.↗

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↗

Evidence of exchange bias effect originating from the interaction between antiferromagnetic core and spin glass shell

Spin glass behavior and exchange bias effect have been observed in antiferromagnetic SrMn{sub 3}O{sub 6−x} nanoribbons synthesized via a self-sacrificing template process. The magnetic field dependence of thermoremanent magnetization and isothermal remanent magnetization shows that the sample is good correspondence to spin glass and diluted antiferromagnetic system for the applied field H < 2 T and H > 2 T, respectively. By detailed analysis of training effect using Binek's model, we argue that the observed exchange bias effect in SrMn{sub 3}O{sub 6−x} nanoribbons arises entirely from an interface exchange coupling between the antiferromagnetic core and spin glass shell. The present study is useful for understanding the nature of shell layer and the origin of exchange bias effect in other antiferromagnetic nanosystems as well.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Enhancing the Flight Safety Culture Through Training

In the 1970's, flight safety professionals became profoundly concerned about the prevalence of crew-caused accidents and incidents, and the role of human error in flight operations. As result, they initiated a change in the flight safety culture which has grown to significant proportions today. At the heart of the evolution were crew concepts such as flightdeck management, crew coordination, and cockpit resource management, concepts which seemed to target critical deficiencies. In themselves, the concepts were not new but their incorporation into training as a direct means of changing the flight safety culture was an untried, almost 'grassroots' approach. The targeted crew concepts and skills were not an integral part of the typical training program; the methods, curriculum, media, and even course content itself, would have to be developed and implemented from the bottom up. A familiar truism in the pilot culture is that you should 'Train the way you fly; Fly the way you train'. In short, training was expected to provide the pilot with practical operational skills that were consistent with the performance standards they were required to maintain and the operational demands they met on a daily basis. In short, one could not simply command crews to use good CRM; one would have to research and define these skills operationally as well as develop and implement a consistent and effective training program. Furthermore, one would need active support and collaboration among the research, industry and government communities in order to ensure acceptance and continued commitment. Additional information is contained in the original extended abstract.

Kanki, Barbara G.↗

Observation of giant exchange bias in bulk Mn{sub 50}Ni{sub 42}Sn{sub 8} Heusler alloy

We report a giant exchange bias (EB) field of 3520 Oe in bulk Mn{sub 50}Ni{sub 42}Sn{sub 8} Heusler alloy. The low temperature magnetic state of the martensite phase has been studied by DC magnetization and AC susceptibility measurements. Frequency dependence of spin freezing temperature (T{sub f}) on critical slowing down relation and observation of memory effect in zero field cooling mode confirms the super spin glass (SSG) phase at low temperatures. Large EB is attributed to the strong exchange coupling between the SSG clusters formed by small regions of ferromagnetic order embedded in an antiferromagnetic (AFM) matrix. The temperature and cooling field dependence of EB have been studied and related to the change in unidirectional anisotropy at SSG/AFM interface. The training effect also corroborates with the presence of frozen (SSG) moments at the interface and their role in EB.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Kinematic and EMG Comparison of Gait in Normal and Microgravity

Astronauts regularly perform treadmill locomotion as a part of their exercise prescription while onboard the International Space Station. Although locomotive exercise has been shown to be beneficial for bone, muscle, and cardiovascular health, astronauts return to Earth after long duration missions with net losses in all three areas [1]. These losses might be partially explained by fundamental differences in locomotive performance between normal gravity (NG) and microgravity (MG) environments. During locomotive exercise in MG, the subject must wear a waist and shoulder harness that is attached to elastomer bungees. The bungees are attached to the treadmill, and provide forces that are intended to replace gravity. However, unlike gravity, which provides a constant force upon all body parts, the bungees provide a spring force only to the harness. Therefore, subjects are subjected to two fundamental differences in MG: 1) forces returning the subject to the treadmill are not constant, and 2) forces are only applied to the axial skeleton at the waist and shoulders. The effectiveness of the exercise may also be affected by the magnitude of the gravity replacement load. Historically, astronauts have difficulty performing treadmill exercise with loads that approach body weight (BW) due to comfort and inherent stiffness in the bungee system. Although locomotion can be executed in MG, the unique requirements could result in performance differences as compared to NG. These differences may help to explain why long term training effects of treadmill exercise may differ from those found in NG. The purpose of this investigation was to compare locomotion in NG and MG to determine if kinematic or muscular activation pattern differences occur between gravitational environments.

DeWitt, John K.↗

Exchange Bias in Layered GdBaCo{sub 2}O{sub 5.5} Cobaltite

It is established that excess oxygen content δ influences the exchange bias (EB) in layered GdBa-Co{sub 2}O{sub 5+δ} cobaltite. The EB effect arises in p-type (δ > 0.5) cobaltite and disappears in n-type (δ < 0.5) cobaltite. The main parameters of EB in GdBaCo{sub 2}O{sub 5.52(2)} polycrystals are determined, including the field and temperature dependences of EB field H{sub EB}, blocking temperature T{sub B}, exchange coupling energy J{sub i} of antiferromagnet–ferromagnet (AFM–FM) interface, and dimensions of FM clusters. The training effect inherent in systems with EB has been studied. The results are explained in terms of exchange interaction between the FM and AFM phases. It is assumed that the EB originates from the coexistence of Co{sup 3+} and Co{sup 4+} ions that leads to the formation of monodomain FM clusters in the AFM matrix of cobaltite.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Intelligent Tutoring Systems for Procedural Task Training of Remote Payload Operations at NASA

Intelligent Tutoring Systems (ITSs) encode and apply the subject matter and teaching expertise of experienced instructors to provide students with individualized instruction automatically. ITSs complement training simulators by providing automated instruction when it is not economical or feasible to dedicate an instructor to each student during training simulations. Despite their proven training effectiveness and favorable operating cost, however, relatively few ITSs are in use. This is largely because it is usually costly and difficult to encode the task knowledge used by the ITS to evaluate the student's actions and assess the student's performance. Procedural tasks are tasks for which there exist procedures, guidelines, and strategies that determine the correct set of steps to be taken within each situation. To lower the cost and difficulty of creating tutoring systems for procedural task training, Stottler Henke Associates, Inc. (SHAI) worked closely with the Operations Training Group at NASA's Marshall Space Flight Center to develop the Task Tutor Toolkit (T (exp 3)), a generic tutoring system shell and scenario authoring tool. The Task Tutor Toolkit employs a case-based reasoning approach where the instructor creates a procedure template that specifies the range of student actions that are "correct" within each scenario. Because each procedure template is specific to a single scenario, the system can employ relatively simple reasoning methods to represent a correct set of actions and assess student performance. This simplicity enables a non-programmer to specify task knowledge quickly and easily by via graphical user interface, using a "demonstrate, generalize, and annotate" paradigm, that recognizes the range of possible valid actions and infers principles understood (or misunderstood) by the student when those actions are carried out. The Task Tutor Toolkit was also designed to be modular and general, so that it can be interfaced with a wide range of training simulators and support a variety of training domains. SHAI and NASA applied the Task Tutor Toolkit to create the Remote Payload Operations Tutor (RPOT). RPOT is a specific tutoring system application which lets scientists who are new to space mission operations learn to monitor and control their experiments aboard the International Space Station according to NASA payload regulations, guidelines, and procedures. The RPOT simulator lets students practice these skills by monitoring the telemetry variable values of a simple, hypothetical experiment, sending commands to the experiment, coordinating with NASA personnel via voice communication loops, and submitting and retrieving information via documents and forms. At the end of each scenario, RPOT displays the principles correctly or incorrectly demonstrated by the student, along with explanations and background information. The effectiveness of RPOT and the Task Tutor Toolkit are currently under evaluation at NASA.

Ong, James↗