Results for “learning efficiency” — Page 12

NASA NTRSDate not supplied

Discovery and Analysis of Rare High-Impact Failure Modes using Adversarial RL-Informed Sampling

Adaptive learning agents have tremendous potential to handle critical tasks currently performed by humans. Unfortunately, due to their complexity, it can be difficult to verify that these learning agents do not have critical failure modes. Standard verification and validation methods often do not apply directly to learning agents and Monte Carlo methods have difficulty covering even a small fraction of the state space, especially in multiagent systems or over long time horizons. To overcome this difficulty, we demonstrate an adaptive stress-testing method based on reinforcement learning of correlations that raise the probability of failure. This approach has three key properties: (1) it is able to find rare failure modes with far greater sample efficiency than Monte Carlo methods, (2) it can estimate the true probability of a failure mode despite the inherent bias in the learning method, and (3) it is capable of learning and resampling compact representations of multimodal failure spaces. These properties are important in practice as we need to find disparate failure modes while accounting for their actual relevance. This is a significant advantage over traditional adaptive stress testing methods that give abstract likelihoods of particular failure instances, but cannot estimate the probability of a broader failure mode. We test our algorithm on a simple problem from the aviation domain where an autonomous aircraft lands in gusty wind conditions. The results suggest that we can find failure modes with far fewer samples than the Monte Carlo approach and simultaneously estimate the probability of failure.

reinforcement learning↗

NASA NTRSDate not supplied

Discovery and Analysis of Rare High-Impact Failure Modes using Adversarial RL-Informed Sampling

Adaptive learning agents have tremendous potential to handle critical tasks currently performed by humans. Unfortunately, due to their complexity, it can be difficult to verify that these learning agents do not have critical failure modes. Standard verification and validation methods often do not apply directly to learning agents and Monte Carlo methods have difficulty covering even a small fraction of the state space, especially in multiagent systems or over long time horizons. To overcome this difficulty, we demonstrate an adaptive stress-testing method based on reinforcement learning of correlations that raise the probability of failure. This approach has three key properties: (1) it is able to find rare failure modes with far greater sample efficiency than Monte Carlo methods, (2) it can estimate the true probability of a failure mode despite the inherent bias in the learning method, and (3) it is capable of learning and resampling compact representations of multimodal failure spaces. These properties are important in practice as we need to find disparate failure modes while accounting for their actual relevance. This is a significant advantage over traditional adaptive stress testing methods that give abstract likelihoods of particular failure instances, but cannot estimate the probability of a broader failure mode. We test our algorithm on a simple problem from the aviation domain where an autonomous aircraft lands in gusty wind conditions. The results suggest that we can find failure modes with far fewer samples than the Monte Carlo approach and simultaneously estimate the probability of failure.

Validation↗

NASA NTRSDate not supplied

Discovery and Analysis of Rare High-Impact Failure Modes using Adversarial RL-Informed Sampling

Adaptive learning agents have tremendous potential to handle critical tasks currently performed by humans. Unfortunately, due to their complexity, it can be difficult to verify that these learning agents do not have critical failure modes. Standard verification and validation methods often do not apply directly to learning agents and Monte Carlo methods have difficulty covering even a small fraction of the state space, especially in multiagent systems or over long time horizons. To overcome this difficulty, we demonstrate an adaptive stress-testing method based on reinforcement learning of correlations that raise the probability of failure. This approach has three key properties: (1) it is able to find rare failure modes with far greater sample efficiency than Monte Carlo methods, (2) it can estimate the true probability of a failure mode despite the inherent bias in the learning method, and (3) it is capable of learning and resampling compact representations of multimodal failure spaces. These properties are important in practice as we need to find disparate failure modes while accounting for their actual relevance. This is a significant advantage over traditional adaptive stress testing methods that give abstract likelihoods of particular failure instances, but cannot estimate the probability of a broader failure mode. We test our algorithm on a simple problem from the aviation domain where an autonomous aircraft lands in gusty wind conditions. The results suggest that we can find failure modes with far fewer samples than the Monte Carlo approach and simultaneously estimate the probability of failure.

Validation↗

NASA NTRS2020

Using Reinforcement Learning to Design Missed Thrust Resilient Trajectories

From ion thrusters to solar sails, spacecraft continue to adopt new and more efficient forms of propulsion. As these low-thrust propulsion meth- ods have become more prevalent, new challenges have arisen. Depending on the mission, low-thrust propulsion elements may need to thrust con- tinuously for days/months. During these thrusting periods, external fac- tors, such as a micro-meteoroid impact or a software glitch, may cause the spacecraft to prematurely cease its thrust stage. Half of all deep space missions enter a safe mode where they cannot thrust every four months. These missed thrust events can result in the complete loss of a space- craft for time-dependent trajectories like planetary rendezvous. This paper demonstrates how neural networks, trained using reinforcement learning, can autonomously correct for missed thrust events during an interplanetary trajectory.

Laipert, Frank E.↗

NASA NTRS1992

Efficient dynamic optimization of logic programs

A summary is given of the dynamic optimization approach to speed up learning for logic programs. The problem is to restructure a recursive program into an equivalent program whose expected performance is optimal for an unknown but fixed population of problem instances. We define the term 'optimal' relative to the source of input instances and sketch an algorithm that can come within a logarithmic factor of optimal with high probability. Finally, we show that finding high-utility unfolding operations (such as EBG) can be reduced to clause reordering.

Laird, Phil↗

NASA NTRS2014

Renewable Energy at NASA's Johnson Space Center

NASA's Johnson Space Center has implemented a great number of renewable energy systems. Renewable energy systems are necessary to research and implement if we humans are expected to continue to grow and thrive on this planet. These systems generate energy using renewable sources - water, wind, sun - things that we will not run out of. Johnson Space Center is helping to pave the way by installing and studying various renewable energy systems. The objective of this report will be to examine the completed renewable energy projects at NASA's Johnson Space Center for a time span of ten years, beginning in 2003 and ending in early 2014. This report will analyze the success of each project based on actual vs. projected savings and actual vs. projected efficiency. Additionally, both positive and negative experiences are documented so that lessons may be learned from past experiences. NASA is incorporating renewable energy wherever it can, including into buildings. According to the 2012 JSC Annual Sustainability Report, there are 321,660 square feet of green building space on JSC's campus. The two projects discussed here are major contributors to that statistic. These buildings were designed to meet various Leadership in Energy and Environmental Design (LEED) Certification criteria. LEED Certified buildings use 30 to 50 percent less energy and water compared to non-LEED buildings. The objectives of this project were to examine data from the renewable energy systems in two of the green buildings onsite - Building 12 and Building 20. In Building 12, data was examined from the solar photovoltaic arrays. In Building 20, data was examined from the solar water heater system. By examining the data from the two buildings, it could be determined if the renewable energy systems are operating efficiently. Objectives In Building 12, the data from the solar photovoltaic arrays shows that the system is continuously collecting energy from the sun, as shown by the graph below. Building 12 has two solar inverters, located on the second floor, that collected the data from the solar photovoltaic arrays. The data displayed here is the total energy produced by the system. These are cumulative amounts, so the last point on the graph shows all of the energy collected from the system since the start of its operation. The data shown here was manually collected from the solar inverters. However, the data is also automatically recorded through EBI. Through analysis of both sets of data it was determined that the EBI data was faulty. For example, from the manually collected data it can be determined that a total of 73 kWh of energy was collected between the dates of 1/16/2014 – 1/22/2014. The EBI data reports that approximately 17800 kWh of energy was collected during the same time frame. Not only does this exceed the time frame examined, but it also exceeds the total energy collected from the start of collection as recorded from the inverters. This leads to the belief that there is a malfunction with the automatic recording of the energy. In Building 20, data was examined from the solar water heater dating back many months and found that the pump for the solar water heater system was not operating properly, as exhibited in the graph shown below. The pump operates on a solar energy system, meaning that it collects energy throughout the day from the sun. Because of this, the system would stop operating shortly after the sun set because of a lack of sunlight. At that point, the graph should show a zero flow rate, but as exhibited in the graph below, that is not the case. It is clearly shown that the pump is continuously operating, even during the night. It was also observed that the majority of the time the pump would not turn on at all, despite good weather conditions. This led to the conclusion that the pump is malfunctioning, and needs to be examined and fixed.

McDowall, Lindsay↗

NASA NTRS1972

General and specific factors in the intersensory transfer of form.

This study assessed the relative contributions of specific and nonspecific components to intersensory transfer between vision and touch. A paired-associate paradigm was used in which visual metric figures and their tactual analogs served as stimuli, and familiar adjectives were the responses. Positive intersensory transfer, characterized by symmetry across modalities was obtained. The contribution of nonspecific learning to this effect was negligible. Intersensory transfer was found to be less efficient than the empirically determined maximum level of intrasensory transfer possible in this task.

Clark, J. L.↗

NASA NTRS1994

Architecture and evolution of Goddard Space Flight Center Distributed Active Archive Center

The Goddard Space Flight Center (GSFC) Distributed Active Archive Center (DAAC) has been developed to enhance Earth Science research by improved access to remote sensor earth science data. Building and operating an archive, even one of a moderate size (a few Terabytes), is a challenging task. One of the critical components of this system is Unitree, the Hierarchical File Storage Management System. Unitree, selected two years ago as the best available solution, requires constant system administrative support. It is not always suitable as an archive and distribution data center, and has moderate performance. The Data Archive and Distribution System (DADS) software developed to monitor, manage, and automate the ingestion, archive, and distribution functions turned out to be more challenging than anticipated. Having the software and tools is not sufficient to succeed. Human interaction within the system must be fully understood to improve efficiency to improve efficiency and ensure that the right tools are developed. One of the lessons learned is that the operability, reliability, and performance aspects should be thoroughly addressed in the initial design. However, the GSFC DAAC has demonstrated that it is capable of distributing over 40 GB per day. A backup system to archive a second copy of all data ingested is under development. This backup system will be used not only for disaster recovery but will also replace the main archive when it is unavailable during maintenance or hardware replacement. The GSFC DAAC has put a strong emphasis on quality at all level of its organization. A Quality team has also been formed to identify quality issues and to propose improvements. The DAAC has conducted numerous tests to benchmark the performance of the system. These tests proved to be extremely useful in identifying bottlenecks and deficiencies in operational procedures.

Bedet, Jean-Jacques↗

NASA NTRS2007

Integration and Testing Challenges of Small, Multiple Satellite Missions: Experiences From The Space Technology 5 Project

This brief presentation describes the mechanical and electrical integration activities and environmental testing challenges of the Space Technology 5 (ST5) Project. Lessons learned during this process are highlighted, including performing mechanical activities serially to gain efficiency through repetition and performing electrical activities based on the level of subsystem expertise available.

Sauerwein, Timothy A.↗

NASA NTRS2010

ISS Logistics Hardware Disposition and Metrics Validation

I was assigned to the Logistics Division of the International Space Station (ISS)/Spacecraft Processing Directorate. The Division consists of eight NASA engineers and specialists that oversee the logistics portion of the Checkout, Assembly, and Payload Processing Services (CAPPS) contract. Boeing, their sub-contractors and the Boeing Prime contract out of Johnson Space Center, provide the Integrated Logistics Support for the ISS activities at Kennedy Space Center. Essentially they ensure that spares are available to support flight hardware processing and the associated ground support equipment (GSE). Boeing maintains a Depot for electrical, mechanical and structural modifications and/or repair capability as required. My assigned task was to learn project management techniques utilized by NASA and its' contractors to provide an efficient and effective logistics support infrastructure to the ISS program. Within the Space Station Processing Facility (SSPF) I was exposed to Logistics support components, such as, the NASA Spacecraft Services Depot (NSSD) capabilities, Mission Processing tools, techniques and Warehouse support issues, required for integrating Space Station elements at the Kennedy Space Center. I also supported the identification of near-term ISS Hardware and Ground Support Equipment (GSE) candidates for excessing/disposition prior to October 2010; and the validation of several Logistics Metrics used by the contractor to measure logistics support effectiveness.

Rogers, Toneka R.↗

NASA NTRS2013

The Tailoring of Traditional Systems Engineering for the Morpheus Project

NASA's Morpheus Project has developed and tested a prototype planetary lander capable of vertical takeoff and landing that is designed to serve as a testbed for advanced spacecraft technologies. The lander vehicle, propelled by a LOX/Methane engine and sized to carry a 500kg payload to the lunar surface, provides a platform for bringing technologies from the laboratory into an integrated flight system at relatively low cost. From the beginning, one of goals for the Morpheus Project was to streamline agency processes and practices. The Morpheus project accepted a challenge to tailor the traditional NASA systems engineering approach in a way that would be appropriate for a lower cost, rapid prototype engineering effort, but retain the essence of the guiding principles. The team has produced innovative ways to create an infrastructure and approach that would challenge existing systems engineering processes while still enabling successful implementation of the current Morpheus Project. This paper describes the tailored systems engineering approach for the Morpheus project, including the processes, tools, and amount of rigor employed over the project's multiple lifecycles since the project began in FY11. Lessons learned from these trials have the potential to be scaled up and improve efficiency on a larger projects or programs.

Devolites, Jennifer L.↗

NASA NTRS2019

Health Management and Prognostics for Electric Aircraft Powertrain

W and c Any air borne vehicle needs incorporating safety as key parameter of measure, and inclusion of autonomy raises the critical need for safety under autonomous operations. Management of faults and component degradation is key as complexity in autonomous operations grow over the period of time. Therefore, in addition to basic operational requirements, an autonomous electric vehicle should be able to make accurate estimates of its current system health and take the correct decisions to complete its mission successfully. Real-time safety and state-awareness tools are therefore essential for the vehicle to be able to reach its destination in a safe and successful manner. The need for safety assurance and health management capabilities is particularly relevant for aircraft electric propulsion systems, which are relatively new and with limited historical to learn. They are critical systems requiring high power density along with reliability, resilience, efficient management of weight, and operational costs. A model- based fault diagnosis and prognostics approach of complex critical systems can successfully accomplish the safety and state awareness goal for such electric propulsion systems, enabling autonomous decision making capability for safe and efficient operation. To identify critical components in the system a Qualitative Bayesian approach using FMECA is implemented. This requires the assessment of some quantities representing the state of the electric unmanned aerial systems (e-UAS), as well as look-ahead forecasts of such states during the entire flight, presented in form of safety metrics (SM). In-service data and performance data gathered from degraded components sup- ports diagnostic and prognostic methods for these systems, but this data can be difficult to obtain as weight and packaging restrictions reduce redundancy and instrumentation on-board the vehicle. Therefore, an model-based framework should be capable or operating with limited data. In addition to data scarcity, the variability of such complex critical systems re- quires the model-based framework to reason in the presence of uncertainty, such as sensor noise, and modeling imperfections. Quantification of errors and uncertainties in the measured states and quantities is therefore a fundamental step for a precise estimation of such SMs; un-modeled uncertainty may result in erroneous state assessment and un- reliable predictions of future states of e-UAVs. Typical, centralized model-based schemes suffer from inherent disadvantages such as computational complexity, single point of failure, and scalability issues, and therefore may fail in such a complex scenario. This paper presents a methodology for developing a system level diagnostics and prognostics approach using a Qualitative Bayesian FMECA approach along with a formal uncertainty management framework for an e-UAS. In this work we demonstrate the efficacy of the framework to predict effects of sub-system level degradation on vehicle operation incorporating uncertainty management to predict future behavior under different operating conditions.

Kulkarni, Chetan↗

NASA NTRSDate not supplied

Holistic and Pragmatic Standards Processes Enable Interdisciplinary Science

Open, interdisciplinary science inevitably relies heavily on standards. Standards are those often unseen agreements that we take for granted when systems and processes are working fine. Yet standards work is perpetual, laborious, and sometimes contentious, especially for standards to work across diverse disciplines. Standards development, maintenance, and implementation is a complex, ongoing socio-technical process. NASA has developed a progressively open science policy and strategy that calls for the establishment of a data standards process reaching across the five diverse divisions of the Science Mission Directorate. This is a delicate exercise. We, therefore, seek to apply a holistic yet pragmatic approach to developing and maintaining a standards process. We adopt an ecological philosophy that focuses on the interactions within the data ecosystem and how standards facilitate those interactions. We couple high-level analysis with on the ground experimentation. We began by 1) mapping information ecosystem components (e.g. data centers, missions, services, protocols, users), 2)establishing how the components interact (e.g. sharing (meta)data, funding, personnel exchange), and 3) modelling system dynamics (e.g. creation of products from multiple data centers, redundant processes, shared services). The goal is to apply understanding of the ecosystem to real world applications (e.g. planning a new mission, implementing new policy requirements, improving process efficiency, etc.). We have also conducted studies of historical standardization efforts, documenting lessons learned and cautionary tales. We then contrast this more abstract work with real examples. We reviewed and assessed multiple existing standards development processes both within and external to NASA. We now work to implement an initial test process which can be further optimized. We seek to define a consistent approach for assigning persistent identifiers for research objects, especially for the purposes of citation. The experience from this relatively ‘simple’ test case adds a pragmatic perspective on how researchers and engineers actually work. This presentation will review the details of this methodology, our initial findings, and how they might apply to other interdisciplinary standardization efforts.

Mark A Parsons↗

NASA NTRSDate not supplied

Advancements in Blowing Dust Detection at Night via Machine Learning

This presentation introduces operational users to a machine-learning based Dust Probability product developed by the NASA SPoRT program for the application of detecting and monitoring blowing dust plumes at night. Advances in earth observing satellites has improved monitoring and detection of dust both day and night through derived imagery such as the Dust RGB. However, limitations of the RGB at night result in less contrast between dust and land surface features, as seen by the user. A Machine Learning (ML) model has been developed and applied to GOES-16 ABI to overcome this limitation and improve nighttime dust detection. The ML capability is a subset of Artificial Intelligence methods. In this case the Dust ML model was developed using a simple Random Forest (RF) model, typically used to solve classification challenges (or to provide regression type output). The goal was to leverage the strengths of the RF model to learn how to identify blowing dust, and hence, overcome the limitation of a user trying to detect blowing dust within the satellite imagery by eye alone. A brief description of the ML model development will be provided. However, the focus of the presentation will be on the initial user feedback from the assessment of this tool for the 2022 blowing dust events of March through April. During this time several users across the U.S. Southwest collaborated to apply this Dust ML product at night as a complement to the existing Dust RGB in order to determine if it provided greater operational efficiency and value.

Machine Learning↗

NASA NTRS2020

Machine Learning-Based Predictive Analytics for Aircraft Engine Conceptual Design

Big data and artificial intelligence/machine learning are transforming the global business environment. Data is now the most valuable asset for enterprises in every industry. Companies are using data-driven insights for competitive advantage. With that, the adoption of machine learning-based data analytics is rapidly taking hold across various industries, producing autonomous systems that support human decision-making. This work explored the application of machine learning to aircraft engine conceptual design. Supervised machine-learning algorithms for regression and classification were employed to study patterns in an existing, open-source database of production and research turbofan engines, and resulting in predictive analytics for use in predicting performance of new turbofan designs. Specifically, the author developed machine learning-based analytics to predict cruise thrust specific fuel consumption (TSFC) and core sizes of high-efficiency turbofan engines, using engine design parameters as the input. The predictive analytics were trained and deployed in Keras, an open-source neural networks application program interface (API) written in Python, with Google’s TensorFlow (an open source library for numerical computation) serving as the backend engine. The promising results of the predictive analytics show that machine-learning techniques merit further exploration for application in aircraft engine conceptual design.

deep-learning↗

NASA NTRS2022

Use of Machine Learning to Reduce Uncertainties in Particle Number Concentration and Aerosol Indirect Radiative Forcing Predicted by Climate Models

The radiative forcing of anthropogenic aerosols associated with aerosol–cloud interactions (RF(sub aci)) remains the largest source of uncertainty in climate prediction. The calculation of particle number concentration (PNC), one of the critical parameters affecting RF(sub aci), is generally simplified in climate models. Here we employ outputs from long-term (30-years) simulations of a global size-resolved (sectional) aerosol microphysics model and a machine-learning tool to develop a Random Forest Regression Model (RFRM) for PNC. We have implemented the PNC RFRM in GISS-ModelE2.1 with a mass-based One-Moment Aerosol module, which is one of CMIP6 models. Compared to the default setting, the GISS-ModelE2.1 simulation based on RFRM reduces the changes of cloud droplet number concentration associated with anthropogenic emissions, and decreases the RF(sub aci) from −1.46 W⋅m(exp −2) to −1.11 W⋅m(exp −2). This work highlights a promising approach based on machine learning to reduce uncertainties of climate models in predicting PNC and RF(sub aci) without compromising their computing efficiency.

Radiative forcing↗

NASA NTRS2024

Automatic Detection and Classification of Aurora in THEMIS All‐Sky Images

We report a novel machine-learning algorithm for automatically detecting and classifying aurora in all–sky images (ASI) that is largely trained without requiring ground–truth labels. By including a small number of labeled images, we are able to automatically label all of the approximately 700 million images in the Time History of Events and Macroscale Interactions during Substorms (THEMIS) ASI data set from 2008 to 2022. We use a two–stage approach. In the first stage, we adapt the Simple framework for Contrastive Learning of Representations (SimCLR) algorithm to learn latent representations of THEMIS all–sky images. We then finetune a classifier network on the latent representations our model learns of the manually labeled Oslo aurora THEMIS (OATH) data set. We demonstrate that this two–stage approach achieves excellent classification results on data for which there is no current ML classification benchmark. The outcome of this work will facilitate efficient information retrieval for researchers interested in specific categories of aurora and will enable large scale statistical studies and machine learning analyses of THEMIS all–sky images that have not previously been possible. To demonstrate possible ways to utilize this database, we performed a statistical analysis of the occurrence rates of auroral labels with respect to solar wind parameters, interplanetary magnetic field vector, and geomagnetic indices. We further investigate the occurrence rates of auroral phenomena in the annotated data set and their geoeffectiveness by utilizing the co–located THEMIS ground magnetometer data set.

Jeremiah W Johnson↗

NASA NTRSDate not supplied

A Machine Learning-Derived Atomistic Potential for Y2Si2O7

Incorporation of SiC/SiC ceramic matrix composite (CMC) hot section components into aircraft engines promises to increase efficiency and safety. However, SiC/SiC CMCs are subject to water vapor-induced oxidation and recession at the high temperatures of engine operation, and thus environmental barrier coatings (EBCs) are required to reduce this degradation and enable their widespread adoption. An understanding of EBCs failure mechanisms, including thermochemical and thermomechanical mechanisms, is essential as coating degradation leads to reduced CMC component service life. Computational modeling approaches can provide insight into EBC material properties important for coating design. However, density functional theory (DFT) is computationally expensive and atomistic potentials are lacking for materials of interest. In this work, we utilize a machine learning approach and DFT training data to parameterize atomistic potentials for two candidate EBC materials, Y2Si2O7 and Yb2Si2O7. These potentials enable near DFT-accurate calculations of thermodynamic and thermomechanical properties essential to EBC design.

Cameron J Bodenschatz↗