Search NASA⌕ Search

SEARCH · Search NASA

Results for “operator learning”

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 505 records · Page 28
All sourcesNASA NTRSDOE OSTI
NASA NTRS2016

How Do Lessons Learned on the International Space Station (ISS) Help Plan Life Support for Mars?

How can our experience in developing and operating the International Space Station (ISS) guide the design, development, and operation of life support for the journey to Mars? The Mars deep space Environmental Control and Life Support System (ECLSS) must incorporate the knowledge and experience gained in developing ECLSS for low Earth orbit, but it must also meet the challenging new requirements of operation in deep space where there is no possibility of emergency resupply or quick crew return. The understanding gained by developing ISS flight hardware and successfully supporting a crew in orbit for many years is uniquely instructive. Different requirements for Mars life support suggest that different decisions may be made in design, testing, and operations planning, but the lessons learned developing the ECLSS for ISS provide valuable guidance.

Lessons learned↗
NASA NTRS2016

Unmanned Aircraft Systems Traffic Management (UTM) Safely Enabling UAS Operations in Low-Altitude Airspace

Unmanned Aircraft System (UAS) Traffic Management (UTM) Enabling Civilian Low-Altitude Airspace and Unmanned Aircraft System Operations What is the problem? Many beneficial civilian applications of UAS have been proposed, from goods delivery and infrastructure surveillance, to search and rescue, and agricultural monitoring. Currently, there is no established infrastructure to enable and safely manage the widespread use of low-altitude airspace and UAS operations, regardless of the type of UAS. A UAS traffic management (UTM) system for low-altitude airspace may be needed, perhaps leveraging concepts from the system of roads, lanes, stop signs, rules and lights that govern vehicles on the ground today, whether the vehicles are driven by humans or are automated. What system technologies is NASA exploring? Building on its legacy of work in air traffic management for crewed aircraft, NASA is researching prototype technologies for a UAS Traffic Management (UTM) system that could develop airspace integration requirements for enabling safe, efficient low-altitude operations. While incorporating lessons learned from the today's well-established air traffic management system, which was a response that grew out of a mid-air collision over the Grand Canyon in the early days of commercial aviation, the UTM system would enable safe and efficient low-altitude airspace operations by providing services such as airspace design, corridors, dynamic geofencing, severe weather and wind avoidance, congestion management, terrain avoidance, route planning and re-routing, separation management, sequencing and spacing, and contingency management. One of the attributes of the UTM system is that it would not require human operators to monitor every vehicle continuously. The system could provide to human managers the data to make strategic decisions related to initiation, continuation, and termination of airspace operations. This approach would ensure that only authenticated UAS could operate in the airspace. In its most mature form, the UTM system could be developed using autonomicity characteristics that include self-configuration, self-optimization and self-protection. The self-configuration aspect could determine whether the operations should continue given the current andor predicted windweather conditions. NASA envisions concepts for two types of possible UTM systems. The first type would be a Portable UTM system, which would move from between geographical areas and support operations such as precision agriculture and disaster relief. The second type of system would be a Persistent UTM system, which would support low-altitude operations and provide continuous coverage for a geographical area. Either system would require persistent communication, navigation, and surveillance (CNS) coverage to track, ensure, and monitor conformance. What is NASA doing to test the technologies? NASA's near-term goal is the development and demonstration of a possible future UTM system that could safely enable low-altitude airspace and UAS operations. Working alongside many committed government, industry and academic partners, NASA is leading the research, development and testing that is taking place in a series of activities called Technology Capability Levels (TCL), each increasing in complexity. UTM TCL1 concluded field testing in August 2015 and is undergoing additional testing at an FAA site.

UTM Low-altitude↗
DOE OSTI2026

Deployment of Traditional and Hybrid Machine Learning for Critical Heat Flux Prediction in the CTF Thermal-Hydraulics Code

Critical heat flux (CHF) marks the transition from nucleate to film boiling, where heat transfer to the working fluid can rapidly deteriorate. Accurate CHF prediction is essential for efficiency, safety, and preventing equipment damage, particularly in nuclear reactors. Although widely used, empirical correlations frequently exhibit discrepancies when compared to experimental data, limiting their reliability in diverse operational conditions. Traditional machine learning (ML) approaches have demonstrated potential for CHF prediction but often suffer from limited interpretability, data scarcity, and insufficient knowledge of physical principles. Hybrid model approaches, which combine data-driven ML with base models, mitigate these concerns by incorporating prior knowledge of the domain. This study integrates an externally trained purely data-driven ML model and two hybrid models (using the Biasi and Bowring CHF correlations) within the CTF subchannel code via a custom Fortran framework. Performance was evaluated using two validation cases: a subset of the Nuclear Regulatory Commission (NRC) CHF database and the Bennett dryout experiments. In both cases, the hybrid models demonstrated significantly lower error metrics compared to conventional empirical correlations, with the best models often reducing relative error by about 5 percentage points. The pure ML model achieved comparable accuracy, outperforming the hybrid Biasi model in the NRC test case (3.3% versus 5.5% relative error) but exhibiting slightly higher error against the hybrid Bowring model in the Bennett test case (7.7% versus 6.1%). Trend analysis of error parity indicated that ML-based models reduced the tendency for CHF overprediction, improving overall accuracy. These results demonstrate that ML-based CHF models can be effectively integrated into subchannel codes and could potentially increase performance compared to conventional methods.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
NASA NTRS1994

(abstract) The IRAS All-Sky Survey: Scientific and Technical Lessons Learned

Although the Infrared Astronomy Satellite (IRAS) was launched over 10 years ago, the IRAS legacy of catalogs, images, and data archives remains a vital source of information on the infrared sky. Over 4000 papers in the refereed literature have used IRAS data, and this number continues to grow. IRAS has touched on almost all aspects of astronomy, from the solar system and the interstellar medium to the formation of stars and the formation of galaxies. In this talk I will give a brief overview of the scientific results from IRAS. I will also discuss some of the operational and technical lessons learned by IRAS that may be of relevance to the upcoming ISO mission, and to future missions that must be defined in an era of declining budgets for space science.

costs operations ISO↗
NASA NTRS2008

GSFC Technical Outreach: The Capitol College Model

In February 2005, as part of the National Aeronautic and Space Administration (NASA) Technical Outreach Program, Goddard Space Flight Center (GSFC) awarded Capitol College of Laurel, Maryland an Educational Grant to establish a Space Operation academic curriculum to meet the future needs of mission operations engineers. This was in part due to the aerospace industry and GSFC concerns that a large number of professional engineers are projected to retire in the near term with evidence showing that current enrollment in engineering schools will not produce sufficient number of space operation trained engineers that will meet industry and government demands. Capitol College, under the agreement of the Educational Grant, established the Space Operations Institute (SOI) with a new curriculum in Space Operations that was approved and certified by the State of Maryland. The SO1 programs focuses on attracting, recruiting, and training a pipeline of highly qualified engineers with experience in mission operations, system engineering and development. The selected students are integrated as members of the engineering support team in any of the missions supported by the institute. The students are mentored by professional engineers from several aerospace companies that support GSFC. Initially, the institute was involved in providing console engineers and mission planning trainees for the Upper Atmosphere Research Satellite (UARS), the Earth Radiation Budget Satellite (ERBS) and the Total Ozone Mapping Spectrometer mission (TOMS). Subsequently, the students were also involved in the technology refresh of the TOMS ground system and other mission operations development. Further mission assignment by GSFC management included participation in the Tropical Rainfall Measuring Mission (TRMM) mission operations and ground system technology refresh. The SOI program has been very successful. Since October 2005, sixty-four students have been enrolled in the SOI program and twenty-five have already graduated from the program, nineteen of whom are employed by company's supporting GSFC. Due to the success of the program, the initial grant period was extended for another period of two years. This paper presents the process that established the SOI as a viable pipeline of mission operations engineers, the lessons learned in the process of dealing with grants, and experience gained in mentoring engineering students that are responsible for particular areas of expertise and functionality. This paper can also be considered a case study and model for integrating a student team with government and industry professionals in the real world of mission operations.

Marius, Julio L.↗
DOE OSTI2026

What do physics-informed DeepONets learn? Understanding and improving training for scientific computing applications

Physics-informed deep operator networks (DeepONets) have emerged as a promising approach toward numerically approximating the solution of partial differential equations (PDEs). In this work, we aim to develop further understanding of what is being learned by physics-informed DeepONets by assessing the universality of the extracted basis functions and demonstrating their potential toward model reduction with spectral methods. Results provide clarity about measuring the performance of a physics-informed DeepONet through the decays of singular values and expansion coefficients. In addition, we propose a transfer learning approach for improving training for physics-informed DeepONets between parameters of the same PDE as well as across different, but related, PDEs where these models struggle to train well. This approach results in significant error reduction and learned basis functions that are more effective in representing the solution of a PDE.

Deep operator networks↗
DOE OSTI2025

Efficient Training of Deep Neural Operator Networks via Randomized Sampling

Neural operators (NOs) employ deep neural networks to learn the mappings between infinitedimensional function spaces. Deep operator network (DeepONet), a popular NO architecture, has demonstrated success in the real-time prediction of complex dynamics across various scientific and engineering applications. In this work, we introduce a random sampling technique to be adopted during the training of DeepONet, aimed at improving the generalization ability of the model, while significantly reducing the computational time. The proposed approach targets the trunk network of the DeepONet model that outputs the basis functions corresponding to the spatiotemporal locations of the bounded domain on which the physical system is defined. While constructing the loss function, DeepONet training traditionally considers a uniform grid of spatiotemporal points at which all the output functions are evaluated for each iteration. This approach leads to a larger batch size, resulting in poor generalization and increased memory demands, due to the limitations of the stochastic gradient descent (SGD) optimizer. The proposed random sampling over the inputs of the trunk net mitigates these challenges, improving generalization and reducing the memory requirements during training, resulting in significant computational gains. We validate our hypothesis through three benchmark examples, demonstrating substantial reductions in training time while achieving comparable or lower overall test errors relative to the traditional training approach. Our results indicate that incorporating randomization in the trunk network inputs during training enhances the efficiency and robustness of DeepONet, offering a promising avenue for improving the framework’s performance in modeling complex physical systems.

Karumuri, Sharmila [Department of Civil & Systems ↗
NASA NTRS2022

Remote Diagnosis and Operational Response to an In-Flight Failure of the Drill Feed Mechanism Onboard the Mars Science Laboratory Rover

On November 30th, 2016 (Sol 1536 of the mission), the Mars Science Laboratory (MSL) flight operations team commanded the Curiosity rover to drill into a Martian rock to acquire the mission’s 16th rock powder sample. The drilling operation was halted when the drill feed mechanism unexpectedly stalled shortly after disengaging its magnetically-actuated friction brake at the start of a long move to extend the drill bit into contact with the target rock. The drilling campaign was put on hold and later abandoned as diagnostic experimentation revealed a tendency for the actuator to transition in and out of a high-drag state in which the mechanism could not actuate. In the subsequent months, the team conducted a variety of in-flight and ground-based investigations to determine root cause and identify mitigation strategies for continued operation. Mechanism reliability continued to degrade and by June 2017, it became clear that the mechanism could not be relied upon to support drilling operations. The team succeeded in fully extending the feed, enabling the development of new sampling techniques that would not rely on the feed mechanism. The first feed-extended drilling sample was successfully collected on Sol 2057, 521 sols after the initial feed stall. This paper discusses the anomaly investigation process, failure analysis, and feed mechanism operation mitigation strategies. It concludes with key lessons learned about mechanism design, mission architecture, and operational best practices.

Lin, Megan Richardson↗
NASA NTRS2012

An Overview of MODIS On-orbit Operation, Calibration, and Lessons

Two nearly identical copies of the Moderate Resolution Imaging Spectroradiometer (MODIS) have successfully operated onboard the Terra and Aqua spacecraft for more than II years and 9 years since their launch in December 1999 and May 2002, respectively. MODIS is a key instrument for the NASA's Earth Observing System (EOS) missions. MODIS observations have produced an unprecedented amount and a broad range of data products and significantly benefited the science and user community. Its follow-on instrument, the Visible/Infrared Imager Radiometer Suite (VIIRS) on-board the NPOESS Preparatory Project (NPP) spacecraft, is currently scheduled for launch in late October, 2011. The NPP serves as a bridge mission between EOS and the Joint Polar Satellite System (JPSS). MODIS collects data in 36 spectral bands, covering spectral regions from visible (VIS) to long-wave infrared (L WIR), and at three different spatial resolutions. Because of its stringent design requirements, MODIS was built with a complete set of onboard calibrators, including a solar diffuser (SO), a solar diffuser stability monitor (SDSM), a blackbody (BB), a spectroradiometric calibration assembly (SRCA), and a space view (SV) port. Except for tbe SRCA, VIlRS carries the same set of onboard calibrators as MODIS. The SD/SDSM system is used together to calibrate tbe reflective solar bands (RSB). The BB is designed for the thermal emissive bands (TEB) calibration. Similar to Terra and Aqua MODIS, VIlRS will also make regular lunar observations to monitor RSB radiometric calibration stability. In this paper, we provide an overview of MODIS on-orbit operation and calibration activities and present issues identified and lessons learned from mission-long instrument operations and implementation of various calibration and characterization strategies. Examples of both Terra and Aqua MODIS instrument on-orbit performance, including their similarities and unique characteristics, are discussed in tbe context of what might be expected from and benefited to tbe NPP VIlRS operation and calibration. It is anticipated that MODIS experience and lessons will also provide valuable information for other earth observing missions/sensors.

Xiong, Jack↗
NASA NTRS1990

Lessons learned from the Hubble Space Telescope planning and scheduling system implementation and operation

The major challenge of the NASA Hubble Space Telescope (HST) Ground System has been the implementation of the planning and scheduling system. This paper describes the operational flow of HST from science proposals to the actual execution of science observations on board the HST. The paper provides a collection of lessons learned over the 10 years that have covered the definition, development, test, and initial operation of HST's planning and scheduling system.

Ruitberg, Edward O.↗
NASA NTRS1992

Spike: AI scheduling for Hubble Space Telescope after 18 months of orbital operations

This paper is a progress report on the Spike scheduling system, developed by the Space Telescope Science Institute for long-term scheduling of Hubble Space Telescope (HST) observations. Spike is an activity-based scheduler which exploits artificial intelligence (AI) techniques for constraint representation and for scheduling search. The system has been in operational use since shortly after HST launch in April 1990. Spike was adopted for several other satellite scheduling problems; of particular interest was the demonstration that the Spike framework is sufficiently flexible to handle both long-term and short-term scheduling, on timescales of years down to minutes or less. We describe the recent progress made in scheduling search techniques, the lessons learned from early HST operations, and the application of Spike to other problem domains. We also describe plans for the future evolution of the system.

Johnston, Mark D.↗
NASA NTRS2011

Habitat Demonstration Unit (HDU) Pressurized Excursion Module (PEM) Systems Integration Strategy

The Habitat Demonstration Unit (HDU) project team constructed an analog prototype lunar surface laboratory called the Pressurized Excursion Module (PEM). The prototype unit subsystems were integrated in a short amount of time, utilizing a rapid prototyping approach that brought together over 20 habitation-related technologies from a variety of NASA centers. This paper describes the system integration strategies and lessons learned, that allowed the PEM to be brought from paper design to working field prototype using a multi-center team. The system integration process was based on a rapid prototyping approach. Tailored design review and test and integration processes facilitated that approach. The use of collaboration tools including electronic tools as well as documentation enabled a geographically distributed team take a paper concept to an operational prototype in approximately one year. One of the major tools used in the integration strategy was a coordinated effort to accurately model all the subsystems using computer aided design (CAD), so conflicts were identified before physical components came together. A deliberate effort was made following the deployment of the HDU PEM for field operations to collect lessons learned to facilitate process improvement and inform the design of future flight or analog versions of habitat systems. Significant items within those lessons learned were limitations with the CAD integration approach and the impact of shell design on flexibility of placing systems within the HDU shell.

Gill, Tracy↗
DOE OSTI2025

Deep Learning–Assisted Multiobjective Optimization of Geological CO 2 Storage Performance under Geomechanical Risks

In geological CO 2 storage, designing the optimal well control strategy for CO 2 injection to maximize CO 2 storage while minimizing the associated geomechanical risks is not trivial. This challenge arises due to pressure buildup, CO 2 plume migration, the highly nonlinear nature of geomechanical responses to rock-fluid interaction, and the high computational cost associated with coupled flow and geomechanics simulations. In this paper, we introduce a novel optimization framework to address these challenges. The optimization problem is formulated as follows: maximize total CO 2 storage while minimizing geomechanical risks by adjusting the injection schedules within bounded constraints. The geomechanical risks are primarily driven by injection-induced pressure build-up, which is characterized by ground displacement and the induced microseismicity. We used the Fourier neural operator (FNO)-based deep learning model to construct surrogate models, replacing the time-consuming coupled flow and geomechanics simulations for evaluating the aforementioned objective functions. The developed surrogate models have been incorporated into a multiobjective optimization framework through a genetic algorithm to reduce the computational burden. The proposed optimization framework reduces the computational cost from approximately 2,400 hours, when using objective function evaluations based on physics-based simulations, to around 20 minutes. A set of Pareto-optimal solutions of the proposed workflow yields nontrivial optimal decisions, reducing the microseismicity potential and the vertical displacement. This Pareto front highlights the optimal trade-offs between CO 2 storage amount, safety, and ground displacement, emphasizing the need for careful optimization and management of injection strategies to achieve a balanced outcome. The novelty of this work is twofold. First, we demonstrate the importance of incorporating the minimization of the geomechanical risks as objective functions into the CO 2 storage optimization workflow to mitigate the potential risk of induced microseismicity and ground displacement. Second, we leverage the FNO-based surrogate models to optimize a real-field CO 2 storage operation.

42 ENGINEERING↗
DOE OSTI2024

Predicting Critical Transitions in Multiscale Data

Predicting the dynamics of complex nonlinear systems remains a challenging problem both in dynamical systems theory as well as real world science and engineering applications. Data-driven methods utilizing the latest advances in machine learning (ML) provide a promising new paradigm for this task. Our work centered on Reservoir Computing (RC), which has shown itself to be capable of skillfully predicting chaotic dynamics in multiscale systems. In the first part of the work, the focus is on how to improve predictions of critical transitions in a class of slow-fast metastable systems in which the equations are known. An additional goal was to determine whether a relationship exists between RC and Koopman operator theory, to improve the efficiency and broaden the applicability of the approach. In the second part of this work, a variation on the RC model known as Reconstructive Reservoir Computing (RRC) is applied to real-world data to identify anomalies.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS2018

SORCE Daylight-Only Operations

The recent experience of the SORCE flight operations team offers an excellent example of innovative engineering using limited resources. The goal of this paper is to extend to the space operations community the lessons learned during this critical redesign in order to aid other missions facing equally daunting challenges. The end result is a mission extended well beyond its designed life continuing to return important data to the science community to extend the climate record.

Spacecraft Anomaly Recovery↗
NASA NTRS1995

Lessons learned from the introduction of autonomous monitoring to the EUVE science operations center

The University of California at Berkeley's (UCB) Center for Extreme Ultraviolet Astrophysics (CEA), in conjunction with NASA's Ames Research Center (ARC), has implemented an autonomous monitoring system in the Extreme Ultraviolet Explorer (EUVE) science operations center (ESOC). The implementation was driven by a need to reduce operations costs and has allowed the ESOC to move from continuous, three-shift, human-tended monitoring of the science payload to a one-shift operation in which the off shifts are monitored by an autonomous anomaly detection system. This system includes Eworks, an artificial intelligence (AI) payload telemetry monitoring package based on RTworks, and Epage, an automatic paging system to notify ESOC personnel of detected anomalies. In this age of shrinking NASA budgets, the lessons learned on the EUVE project are useful to other NASA missions looking for ways to reduce their operations budgets. The process of knowledge capture, from the payload controllers for implementation in an expert system, is directly applicable to any mission considering a transition to autonomous monitoring in their control center. The collaboration with ARC demonstrates how a project with limited programming resources can expand the breadth of its goals without incurring the high cost of hiring additional, dedicated programmers. This dispersal of expertise across NASA centers allows future missions to easily access experts for collaborative efforts of their own. Even the criterion used to choose an expert system has widespread impacts on the implementation, including the completion time and the final cost. In this paper we discuss, from inception to completion, the areas where our experiences in moving from three shifts to one shift may offer insights for other NASA missions.

Lewis, M.↗
NASA NTRS2009

Active Learning with Irrelevant Examples

An improved active learning method has been devised for training data classifiers. One example of a data classifier is the algorithm used by the United States Postal Service since the 1960s to recognize scans of handwritten digits for processing zip codes. Active learning algorithms enable rapid training with minimal investment of time on the part of human experts to provide training examples consisting of correctly classified (labeled) input data. They function by identifying which examples would be most profitable for a human expert to label. The goal is to maximize classifier accuracy while minimizing the number of examples the expert must label. Although there are several well-established methods for active learning, they may not operate well when irrelevant examples are present in the data set. That is, they may select an item for labeling that the expert simply cannot assign to any of the valid classes. In the context of classifying handwritten digits, the irrelevant items may include stray marks, smudges, and mis-scans. Querying the expert about these items results in wasted time or erroneous labels, if the expert is forced to assign the item to one of the valid classes. In contrast, the new algorithm provides a specific mechanism for avoiding querying the irrelevant items. This algorithm has two components: an active learner (which could be a conventional active learning algorithm) and a relevance classifier. The combination of these components yields a method, denoted Relevance Bias, that enables the active learner to avoid querying irrelevant data so as to increase its learning rate and efficiency when irrelevant items are present. The algorithm collects irrelevant data in a set of rejected examples, then trains the relevance classifier to distinguish between labeled (relevant) training examples and the rejected ones. The active learner combines its ranking of the items with the probability that they are relevant to yield a final decision about which item to present to the expert for labeling. Experiments on several data sets have demonstrated that the Relevance Bias approach significantly decreases the number of irrelevant items queried and also accelerates learning speed.

Wagstaff, Kiri↗
DOE OSTI2025

Synergistic learning with multi-task DeepONet for efficient PDE problem solving

Multi-task learning (MTL) is an inductive transfer mechanism designed to leverage useful information from multiple tasks to improve generalization performance compared to single-task learning. It has been extensively explored in traditional machine learning to address issues such as data sparsity and overfitting in neural networks. In this work, we apply MTL to problems in science and engineering governed by partial differential equations (PDEs). However, implementing MTL in this context is complex, as it requires task-specific modifications to accommodate various scenarios representing different physical processes. To this end, we present a multi-task deep operator network (MT-DeepONet) to learn solutions across various functional forms of source terms in a PDE and multiple geometries in a single concurrent training session. We introduce modifications in the branch network of the vanilla DeepONet to account for various functional forms of a parameterized coefficient in a PDE. Additionally, we handle parameterized geometries by introducing a binary mask in the branch network and incorporating it into the loss term to improve convergence and generalization to new geometry tasks. Our approach is demonstrated on three benchmark problems: (1) learning different functional forms of the source term in the Fisher equation; (2) learning multiple geometries in a 2D Darcy Flow problem and showcasing better transfer learning capabilities to new geometries; and (3) learning 3D parameterized geometries for a heat transfer problem and demonstrate the ability to predict on new but similar geometries. Finally, our MT-DeepONet framework offers a novel approach to solving PDE problems in engineering and science under a unified umbrella based on synergistic learning that reduces the overall training cost for neural operators.

42 ENGINEERING↗