Search NASA⌕ Search

SEARCH · Search NASA

Results for “Random effect model”

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 739 records · Page 41

Fine particulate concentrations over East Asia derived from aerosols measured by the Advanced Himawari Imager using machine learning

Fine particulate matter with a diameter below 2.5 μm (PM 2.5 ) is deleterious to the cardiovascular and respiratory systems. It is often difficult to assess the effects of PM 2.5 on human health over regions with limited ground monitoring sites, especially in East Asia. As an alternative, we estimated near-surface PM 2.5 concentrations by analyzing Advanced Himawari Imager (AHI) Yonsei Aerosol Retrieval (YAER) products. This study incorporates daytime data for East Asia covering the Korean Peninsula, China, Japan, Southeast Asia, and southern Mongolia. We collocated AHI YAER product pixels with meteorological, land-cover, and other ancillary data for the period from March 2018 to February 2019. To estimate PM 2.5 concentrations over wide areas spanning many countries displaying various relationships between aerosol optical depth and PM 2.5 , monthly models were developed by considering both the spatial and temporal characteristics of ground-based PM 2.5 measurements. Random forest machine learning model estimated ground-level mass concentrations of PM 2.5 ; subsequent 10-fold cross validation (CV) yielded a CV R 2 value of 0.81 and a CV root mean squared error (RMSE) of 12.3 μg m -3 . We investigated the spatial pattern of PM 2.5 concentrations over multiple countries and seasonal variation in PM 2.5 concentrations. Diurnal variation of a severe PM 2.5 event in the Korean Peninsula was investigated as a case study. The model captured the extremely heterogeneous spatial distribution of PM 2.5 concentrations peaked around local noon. To measure the capability of the developed model to estimate PM 2.5 concentrations in areas with few in-situ data, its predictive performance was evaluated using a dataset independent of the training process with an R 2 of 0.60 and RMSE of 8.18 μg m −3 . This study demonstrates the potential for satellite-based PM 2.5 estimation for areas with insufficient measuring stations.

Pm2.5↗

Partition-based Feasible Integer Solution Pre-computation for Hybrid Model Predictive Control

For multiparametric mixed-integer convex programming problems such as those encountered in hybrid model predictive control, we propose an algorithm for generating a feasible partition of a subset of the parameter space. The result is a static map from the current parameter to a suboptimal integer solution such that the remaining convex program is feasible. Convergence is proved with a new insight that the overlap among the feasible parameter sets of each integer solution governs the partition complexity. The partition is stored as a tree which makes querying the feasible solution efficient. The algorithm can be used to warm start a mixed integer solver with a real-time guarantee or to provide a reference integer solution in several suboptimal MPC schemes. The algorithm is tested on randomly generated systems with up to six states, demonstrating the effectiveness of the approach.

Bayard, David S.↗

Sputtering Erosion in the Ion Thruster

During the first phase of this research, the sputtering yields of molybdenum by low energy (100 eV and higher) xenon ions were measured by using the methods of secondary neutral mass spectrometry (SNMS) and Rutherford backscattering spectrometry (RBS). However, the measured sputtering yields were found to be far too low to explain the sputtering erosions observed in the long-duration tests of ion thrusters. The only difference between the sputtering yield measurement experiments and the ion thruster tests was that the later are conducted at high ion fluences. Hence, a study was initiated to investigate if any linkage exists between high ion fluence and an enhanced sputtering yield. The objective of this research is to gain an understanding of the causes of the discrepancies between the sputtering rates of molybdenum grids in an ion thruster and those measured from our experiments. We are developing a molecular dynamics simulation technique for studying low-energy xenon ion interactions with molybdenum. It is difficult to determine collision sequences analytically for primary ions below the 200 eV energy range where the ion energy is too low to be able to employ a random cascade model with confidence and it is too high to have to consider only single collision at or near the surface. At these low energies, the range of primary ions is about 1 to 2 nm from the surface and it takes less than 4 collisions on the average to get an ion to degrade to such an energy that it can no longer migrate. The fine details of atomic motion during the sputtering process are revealed through computer simulation schemes. By using an appropriate interatomic potential, the positions and velocities of the incident ion together with a sufficient number of target atoms are determined in small time steps. Hence, it allows one to study the evolution of damages in the target and its effect on the sputtering yield. We are at the preliminary stages of setting up the simulation program.

Ray, Pradosh K.↗

What’s That Supposed to Mean? Capturing Micro-Behaviors in Teams

Future long-duration space exploration (LDSE) crews will require extensive coordination, cooperation, and team functioning as they face a myriad of challenges rooted in both taskwork and teamwork (Bell et al., 2015; Landon et al., 2018). While exposed to extreme conditions, crew members must navigate living and working together in prolonged confinement. Moreover, astronaut teams are becoming increasingly diverse, introducing significant variability in team composition. This increasing diversity, alongside traditional constraints of LDSE, introduces additional challenges into effective team functioning. To date, most methods for capturing team functioning rely on self-report measures. Such measures are prone to several limitations, including but not limited to social desirability bias, halo effect, and leniency effects (Trull & Ebner-Priemer, 2013), which skew data and limit nuanced understandings of phenomena at play. Self-report measures broadly capture team functioning, lending the nature of such methods to identifying underlying “macro”-behaviors (i.e., behaviors that are long-standing and last over time). However, team functioning is far more complex than a series of macro-behaviors, rendering reliance on self-report data deficient for accurate measurement. Recent research demonstrates the potential of alternative methods for capturing team functioning, such as speech and physiological data (Chaffin et al., 2017; Murray & Oertel, 2018). Consequently, these methods are more suitable for capturing micro-behaviors: brief, often unconscious expressions that affect the extent to which an individual feels included by others around them (Paletz et al., 2013). Micro-behaviors can be further classified into microaggressions (i.e., subtle, negative exchanges; Keller & Galgay, 2010) or micro-affirmations (i.e., subtle, positive exchanges; Kyte et al. 2020), both of which influence team functioning. Due to the subtle nature of micro-behaviors, contextual factors have a significant impact when determining if it is aggressive or affirmative. Additionally, several iterations of microbehaviors can have lingering effects on team interactions. For example, the use of “mm-hmm” by a crew member can function as both a micro-affirmation and micro-aggression. Specifically, it can be indication of active listening (i.e., micro-affirmation) or as an expression of annoyance (i.e., aggression) depending on the context in which it occurs. Auditory features (e.g., tone, frequency) can help delineate between the two forms; however, the contextual factors (e.g., previous interactions between team members, crew demographics) add a layer of complexity that render auditory features alone as insufficient to capture micro-behaviors. Consequently, this paper seeks to provide a novel approach in which multi-modal data (i.e., auditory features and contextual features) are used in a random-forest model to better identify distinguishing characteristics between micro-affirmations and micro-aggressions. In turn, detected micro-behaviors are used to predict team performance, thereby demonstrating the value of capturing micro-behaviors as supplemental data to macro-behaviors.

Sydney R. Begerowski↗

What’s That Supposed to Mean? Capturing Micro-Behaviors in Teams

Future long-duration space exploration (LDSE) crews will require extensive coordination, cooperation, and team functioning as they face a myriad of challenges rooted in both taskwork and teamwork (Bell et al., 2015; Landon et al., 2018). While exposed to extreme conditions, crew members must navigate living and working together in prolonged confinement. Moreover, astronaut teams are becoming increasingly diverse, introducing significant variability in team composition. This increasing diversity, alongside traditional constraints of LDSE, introduces additional challenges into effective team functioning. To date, most methods for capturing team functioning rely on self-report measures. Such measures are prone to several limitations, including but not limited to social desirability bias, halo effect, and leniency effects (Trull & Ebner-Priemer, 2013), which skew data and limit nuanced understandings of phenomena at play. Self-report measures broadly capture team functioning, lending the nature of such methods to identifying underlying “macro”-behaviors (i.e., behaviors that are long-standing and last over time). However, team functioning is far more complex than a series of macro-behaviors, rendering reliance on self-report data deficient for accurate measurement. Recent research demonstrates the potential of alternative methods for capturing team functioning, such as speech and physiological data (Chaffin et al., 2017; Murray & Oertel, 2018). Consequently, these methods are more suitable for capturing micro-behaviors: brief, often unconscious expressions that affect the extent to which an individual feels included by others around them (Paletz et al., 2013). Micro-behaviors can be further classified into micro-aggressions (i.e., subtle, negative exchanges; Keller & Galgay, 2010) or micro-affirmations (i.e., subtle, positive exchanges; Kyte et al. 2020), both of which influence team functioning. Due to the subtle nature of micro-behaviors, contextual factors have a significant impact when determining if it is aggressive or affirmative. Additionally, several iterations of micro-behaviors can have lingering effects on team interactions. For example, the use of “mm-hmm” by a crew member can function as both a micro-affirmation and micro-aggression. Specifically, it can be indication of active listening (i.e., micro-affirmation) or as an expression of annoyance (i.e., aggression) depending on the context in which it occurs. Auditory features (e.g., tone, frequency) can help delineate between the two forms; however, the contextual factors (e.g., previous interactions between team members, crew demographics) add a layer of complexity that render auditory features alone as insufficient to capture micro-behaviors. Consequently, this paper seeks to provide a novel approach in which multi-modal data (i.e., auditory features and contextual features) are used in a random-forest model to better identify distinguishing characteristics between micro-affirmations and micro-aggressions. In turn, detected micro-behaviors are used to predict team performance, thereby demonstrating the value of capturing micro-behaviors as supplemental data to macro-behaviors.

Sydney Begerowski↗

Modeling of Electromagnetic Scattering by Discrete and Discretely Heterogeneous Random Media by Using Numerically Exact Solutions of the Maxwell Equations

In this paper, we discuss some aspects of numerical modeling of electromagnetic scattering by discrete random medium by using numerically exact solutions of the macroscopic Maxwell equations. Typical examples of such media are clouds of interstellar dust, clouds of interplanetary dust in the Solar system, dusty atmospheres of comets, particulate planetary rings, clouds in planetary atmospheres, aerosol particles with numerous inclusions and so on. Our study is based on the results of extensive computations of different characteristics of electromagnetic scattering obtained by using the superposition T-matrix method which represents a direct computer solver of the macroscopic Maxwell equations for an arbitrary multisphere configuration. As a result, in particular, we clarify the range of applicability of the low-density theories of radiative transfer and coherent backscattering as well as of widely used effective-medium approximations.

Dlugach, Janna M.↗

Probabilistic Assessment of Radiation Risk for Astronauts in Space Missions

Accurate predictions of the health risks to astronauts from space radiation exposure are necessary for enabling future lunar and Mars missions. Space radiation consists of solar particle events (SPEs), comprised largely of medium energy protons, (less than 100 MeV); and galactic cosmic rays (GCR), which include protons and heavy ions of higher energies. While the expected frequency of SPEs is strongly influenced by the solar activity cycle, SPE occurrences themselves are random in nature. A solar modulation model has been developed for the temporal characterization of the GCR environment, which is represented by the deceleration potential, phi. The risk of radiation exposure from SPEs during extra-vehicular activities (EVAs) or in lightly shielded vehicles is a major concern for radiation protection, including determining the shielding and operational requirements for astronauts and hardware. To support the probabilistic risk assessment for EVAs, which would be up to 15% of crew time on lunar missions, we estimated the probability of SPE occurrence as a function of time within a solar cycle using a nonhomogeneous Poisson model to fit the historical database of measurements of protons with energy > 30 MeV, (phi)30. The resultant organ doses and dose equivalents, as well as effective whole body doses for acute and cancer risk estimations are analyzed for a conceptual habitat module and a lunar rover during defined space mission periods. This probabilistic approach to radiation risk assessment from SPE and GCR is in support of mission design and operational planning to manage radiation risks for space exploration.

Kim, Myung-Hee↗

A Multilevel Approach For SolvingLarge-Scale QUBO Problems With Noisy Hybrid Quantum Approximate Optimization

Quantum approximate optimization is one ofthe promising candidates for useful quantum computation,particularly in the context of finding approximate solutionsto Quadratic Unconstrained Binary Optimization (QUBO)problems. However, the existing quantum processing units(QPUs) are of relatively small size, and canonical mappingsof QUBO via the Ising model require one qubit per vari-able, rendering direct large-scale optimization infeasible.In classical optimization, a general strategy for addressingmany large-scale problems is via multilevel/multigrid meth-ods, where the large target problem is iteratively coarsenedand the global solution is constructed from multiple small-scale optimization runs. In this work, we experimentallytest how existing QPUs perform when used as a sub-solverwithin such a multilevel strategy. To this aim, we com-bine and extend (via additional classical processing steps)the recently proposed Noise-Directed Adaptive Remapping(NDAR) and Quantum Relax&Round (QRR) algorithms.We first demonstrate the effectiveness of our heuristicextensions on Rigetti’s superconducting transmon deviceAnkaa-2. We find approximate solutions to10instances offully connected82-qubit Sherrington-Kirkpatrick graphswith random integer-valued coefficients obtaining normal-ized approximation ratios (ARs) in the range∼0.98−1.0,and the same class with real-valued coefficients (ARs∼0.94−1.0). Then, we implement the extended NDAR andQRR algorithms as subsolvers in the multilevel algorithmfor6large-scale graphs with at most∼27,000variables.In practice, the QPU (with classical post-processing steps)is used to find approximate solutions to dozens of at most82-qubit problems, which are iteratively used to constructthe global solution. We observe that quantum optimizationresults are competitive in terms of the quality of solutionswhen compared to classical heuristics used as subsolverswithin the multilevel approach.Reproducibility: source code and data are available at[TBA upon acceptance]

quantum computing↗

Enhanced orbit determination filter sensitivity analysis: Error budget development

An error budget analysis is presented which quantifies the effects of different error sources in the orbit determination process when the enhanced orbit determination filter, recently developed, is used to reduce radio metric data. The enhanced filter strategy differs from more traditional filtering methods in that nearly all of the principal ground system calibration errors affecting the data are represented as filter parameters. Error budget computations were performed for a Mars Observer interplanetary cruise scenario for cases in which only X-band (8.4-GHz) Doppler data were used to determine the spacecraft's orbit, X-band ranging data were used exclusively, and a combined set in which the ranging data were used in addition to the Doppler data. In all three cases, the filter model was assumed to be a correct representation of the physical world. Random nongravitational accelerations were found to be the largest source of error contributing to the individual error budgets. Other significant contributors, depending on the data strategy used, were solar-radiation pressure coefficient uncertainty, random earth-orientation calibration errors, and Deep Space Network (DSN) station location uncertainty.

Estefan, J. A.↗

A High Performance Computing Approach to Tree Cover Delineation in 1-m NAIP Imagery Using a Probabilistic Learning Framework

Tree cover delineation is a useful instrument in deriving Above Ground Biomass (AGB) density estimates from Very High Resolution (VHR) airborne imagery data. Numerous algorithms have been designed to address this problem, but most of them do not scale to these datasets, which are of the order of terabytes. In this paper, we present a semi-automated probabilistic framework for the segmentation and classification of 1-m National Agriculture Imagery Program (NAIP) for tree-cover delineation for the whole of Continental United States, using a High Performance Computing Architecture. Classification is performed using a multi-layer Feedforward Backpropagation Neural Network and segmentation is performed using a Statistical Region Merging algorithm. The results from the classification and segmentation algorithms are then consolidated into a structured prediction framework using a discriminative undirected probabilistic graphical model based on Conditional Random Field, which helps in capturing the higher order contextual dependencies between neighboring pixels. Once the final probability maps are generated, the framework is updated and re-trained by relabeling misclassified image patches. This leads to a significant improvement in the true positive rates and reduction in false positive rates. The tree cover maps were generated for the whole state of California, spanning a total of 11,095 NAIP tiles covering a total geographical area of 163,696 sq. miles. The framework produced true positive rates of around 88% for fragmented forests and 74% for urban tree cover areas, with false positive rates lower than 2% for both landscapes. Comparative studies with the National Land Cover Data (NLCD) algorithm and the LiDAR canopy height model (CHM) showed the effectiveness of our framework for generating accurate high-resolution tree-cover maps.

Segments↗

Salvaging Data Records with Missing Data: Data Imputation using the Multivariate t Distribution

When doing multivariate data analysis, one commonobstacle is the presence of incomplete observations, i.e., observationsfor which one or more key fields are blank. Missing datais often countered by deleting entire observations that containmissing data. The negative effects of deleting entire observationsare multiple: deleting observations reduces sample size andcan also result in biased inferences even if data is missing atrandom. In addition, knowledge contained within incompleteobservations is knowledge lost when they are deleted– and theeffort spent collecting that knowledge is effort wasted. Data imputationmethods, or methods of statistically “filling-in” missingdata, can help combat small sample sizes by using the existinginformation in partially complete observations with the end goalof producing less biased and higher confidence inferences. Whena sample from a multivariate normal population is only partiallycomplete, and the missing data meets appropriate assumptions(missing at random), robust data imputation of the missing datacan be implemented with monotone data augmentation (MDA)using the multivariate t distribution.Missing data imputation is applied to data from the NASA InstrumentCost Model (NICM) using the MDA algorithm underthe assumption of having a multivariate t distribution with fixeddegrees of freedom. A sensitivity analysis to the degrees offreedom parameter is presented to demonstrate robustness ofthe multivariate t distribution when dealing with small samplesas compared to the multivariate normal distribution.

DiNicola, Michael↗

Dynamical evolution of the Oort cometary cloud

The dynamical evolution of comets in the Oort cloud under the influence of stellar perturbations has been modeled using Monte Carlo techniques. It is shown that the cloud has been depleted over the history of the solar system. Comets are lost from the cloud by direct ejection due to close stellar encounters, diffusion of aphelia to distances beyond the sun's sphere of influence, or diffusion of perihelia into the planetary region where Jupiter and Saturn perturbations either eject them on hyperbolic trajectories or capture them to short-period orbits. The population of the cloud is estimated to be 1.0 - 1.5 x 10 to the 12th comets and the total mass is on the order of 1.9 earth masses. In addition to random passing stars, less frequent encounters with giant molecular clouds may play a significant role in randomizing the orbits of comets in the cloud and reducing the effective radius of the sun's sphere of influence.

Weissman, P. R.↗

A Systematic Error Correction Method for TOVS Radiances

Treatment of systematic errors is crucial for the successful use of satellite data in a data assimilation system. Systematic errors in TOVS radiance measurements and radiative transfer calculations can be as large or larger than random instrument errors. The usual assumption in data assimilation is that observational errors are unbiased. If biases are not effectively removed prior to assimilation, the impact of satellite data will be lessened and can even be detrimental. Treatment of systematic errors is important for short-term forecast skill as well as the creation of climate data sets. A systematic error correction algorithm has been developed as part of a 1D radiance assimilation. This scheme corrects for spectroscopic errors, errors in the instrument response function, and other biases in the forward radiance calculation for TOVS. Such algorithms are often referred to as tuning of the radiances. The scheme is able to account for the complex, air-mass dependent biases that are seen in the differences between TOVS radiance observations and forward model calculations. We will show results of systematic error correction applied to the NOAA 15 Advanced TOVS as well as its predecessors. We will also discuss the ramifications of inter-instrument bias with a focus on stratospheric measurements.

Joiner, Joanna↗

Soluble TNF-alpha receptor 1 and IL-6 plasma levels in humans subjected to the sleep deprivation model of spaceflight

BACKGROUND: The extent to which sleep loss may predispose astronauts to a state of altered immunity during extended space travel prompts evaluation with ground-based models. OBJECTIVE: We sought to measure plasma levels of selected cytokines and their receptors, including the putative sleep-regulation proteins soluble TNF-alpha receptor (sTNF-alpha R) I and IL-6, in human subjects undergoing 2 types of sleep deprivation during environmental confinement with performance demands. METHODS: Healthy adult men (n = 42) were randomized to schedules that varied in severity of sleep loss: 4 days (88 hours) of partial sleep deprivation (PSD) involving two 2-hour naps per day or 4 days of total sleep deprivation (TSD). Plasma samples were obtained every 6 hours across 5 days and analyzed by using enzyme-linked immunoassays for sTNF-alpha RI, sTNF-alpha RII, IL-6, soluble IL-2 receptor, IL-10, and TNF-alpha. RESULTS: Interactions between the effects of time and sleep deprivation level were detected for sTNF-alpha RI and IL-6 but not for sTNF-alpha RII, soluble IL-2 receptor, IL-10, and TNF-alpha. Relative to the PSD condition, subjects in the TSD condition had elevated plasma levels of sTNF-alpha RI on day 2 (P =.04), day 3 (P =.01), and across days 2 to 4 of sleep loss (P =.01) and elevated levels of IL-6 on day 4 (P =.04). CONCLUSIONS: Total sleep loss produced significant increases in plasma levels of sTNF-alpha RI and IL-6, messengers that connect the nervous, endocrine, and immune systems. These changes appeared to reflect elevations of the homeostatic drive for sleep because they occurred in TSD but not PSD, suggesting that naps may serve as the basis for a countermeasures approach to prolonged spaceflight.

Non-NASA Center↗

Orientation and length of mammalian skeletal myocytes in response to a unidirectional stretch

Effects of mechanical forces exerted on mammalian skeletal muscle cells during development were studied using an in vitro model to unidirectionally stretch cultured C2C12 cells grown on silastic membrane. Previous models to date have not studied these responses of the mammalian system specifically. The silastic membrane upon which these cells were grown exhibited linear strain behavior over the range of 3.6-14.6% strain, with a Poisson's ratio of approximately 0.5. To mimic murine in utero long bone growth, cell substrates were stretched at an average strain rate of 2.36%/day for 4 days or 1.77%/day for 6 days with an overall membrane strain of 9.5% and 10.6%, respectively. Both control and stretched fibers stained positively for the contractile protein, alpha-actinin, demonstrating muscle fiber development. An effect of stretch on orientation and length of myofibers was observed. At both strain rates, stretched fibers aligned at a smaller angle relative to the direction of stretch and were significantly longer compared to randomly oriented control fibers. There was no effect of duration of stretch on orientation or length, suggesting the cellular responses are independent of strain rate for the range tested. These results demonstrate that, under conditions simulating mammalian long bone growth, cultured myocytes respond to mechanical forces by lengthening and orienting along the direction of stretch.

Non-NASA Center↗

Towards an Aviation Large Language Model by Fine-tuning and Evaluating Transformers

In the aviation domain, there are many applications for machine learning and artificial intelligence tools that utilize natural language. For example, there is a desire to know the commonalities in written safety reports such as voluntary post incidents reports or aerial wildfire operations reports to better understand the risks present. Another use-case is the possibility of extracting airspace procedures and constraints currently written in documents such as Letters of Agreement. These applications can benefit from the use of state-of-the-art natural language processing techniques when adapted to the language/phraseology specific to the aviation domain. This paper evaluates the viability of adaptation of NLP tools to the aviation domain by fine-tuning transformer based models using aviation data sets. In 2018, a novel language model based on neural units (also called transformers) was created and became known as “Bidirectional Encoder Representations from Transformers” or BERT. This architecture combined with large amounts of English training data and innovative semi-supervised training tasks set the standard for what would later emerge as Large Language Models. The performance of these models was further improved by hyperparameter tuning and refinement of the semi-supervised training task and resulted in “Robustly Optimized BERT Pre-training Approach through hyperparameter tuning” or RoBERTa models. These pre-trained Large Language Models proved to be useful for a wide variety of natural language processing tasks such as text classification and question answering through a process called fine-tuning. The transformer architecture with pre-trained weights served as the basis with the last few layers replaced with layers fine-tuned to perform a new task e.g., a layer that provides a label for the entire input text. This process of fine-tuning can also be used to adapt the models to new domains; e.g., BioBERT started with the pre-trained BERT model and was completed by additional fine-tuning and training on biomedical documents. Transformer-based architectures can also be used to create rich representations of text called embeddings which can serve as the input to other machine learning models. This allows simpler algorithms such as logistic regression to use context-rich representations of the text while still remaining quick to train and evaluate. In the world of aviation, there is a growing demand for natural language processing and understanding but the domain presents unique challenges. Due to the technical content (and specialized language) of most aviation documents, fine-tuning pre-trained Large Language Models to specific tasks has not met the benchmark on natural language processing tasks set by simpler models trained from scratch on the data. To address this deficiency, this paper evaluates the improvements from fine-tuning a Large Language Model on a large set of aviation documents using the original semi-supervised training tasks before performing specific natural language tasks. In fine-tuning, a domain-specific dataset is used on the original training task but with the pre-trained Large Language Model instead of starting from a random initialization. This approach allows the model to be adapted to the specific domain language without discarding the information gained from training on general English data. This paper utilized two major dataset types to train and assess the RoBERTa fine-tuning performance. The first are 7,057 Letters of Agreement which are Federal Aviation Administration (FAA) documents that formalize airspace operations across the national airspace system. They contain many examples of ‘aviation English’ using domain specific terminology and phrasing which serves as a representative basis to perform the semi-supervised fine-tuning. The second type is the 494 document classification labels to be used for evaluation. This down-stream evaluation aims to show the performance of the fine-tuned model, better understand how much data is needed for an effective fine-tuning, and how fine-tuning can be adapted for different applications in-the domain. After semi-supervised training, evaluation begins by encoding the documents for classification using the fine-tuned RoBERTa model. Then a logistic regression classifier is trained to label the document type and compared against our ground truth labels. This currently leads to a 82.8% accuracy on 10-fold cross validation showing improvement over baseline RoBERTa which achieved 81.0%. We plan to measure the improvements on additional tasks and it is expected that these improvements will lead to more robust models that can tackle the natural language processing challenges present in aviation datasets.

ATM↗

Environmental Testing of the NEXT PM1 Ion Engine

The NEXT propulsion system is an advanced ion propulsion system presently under development that is oriented towards robotic exploration of the solar system using solar electric power. The Prototype Model engine PM1 was subjected to qualification-level environmental testing to demonstrate compatibility with environments representative of anticipated mission requirements. Random vibration testing, conducted with the thruster mated to the breadboard gimbal, was executed at 10.0 Grms for 2 minutes in each of three axes. Thermal-vacuum testing included a deep cold soak of the engine to temperatures of -168 C and thermal cycling from -120 to 203 C. Although the testing was largely successful, several issues were identified including the fragmentation of potting cement on the discharge and neutralizer cathode heater terminations during vibration which led to abbreviated thermal testing, and generation of particulate contamination from manufacturing processes and engine materials. Thruster performance was nominal throughout the test program, with minor variations in some engine operating parameters likely caused by facility effects. In general, the NEXT PM1 engine and the breadboard gimbal were found to be well-designed against environmental requirements based on the results reported herein. After resolution of the findings from this test program the hardware environmental qualification program can proceed with confidence.

Synder, John S.↗

Tortuosity Computations of Porous Materials using the Direct Simulation Monte Carlo

Low-density carbon fiber preforms, used as thermal protection systems (TPS) materials for planetary entry systems, have permeable, highly porous microstructures consisting of interlaced fibers. Internal gas transport in TPS is important in modeling the penetration of hot boundary-layer gases and the in-depth transport of pyrolysis and ablation products. The gas effective diffusion coefficient of a porous material must be known before the gas transport can be modeled in material response solvers; however, there are very little available data for rigid fibrous insulators used in heritage TPS.The tortuosity factor, which reflects the efficiency of the percolation paths, can be computed from the effective diffusion coefficient of a gas inside a porous material and is based on the micro-structure of the material. It is well known, that the tortuosity factor is a strong function of the Knudsen number. Due to the small characteristic scales of porous media used in TPS applications (typical pore size of the order of 50 micron), the transport of gases can occur in the rarefied and transitional regimes, at Knudsen numbers above 1. A proper way to model the gas dynamics at these conditions consists in solving the Boltzmann equation using particle-based methods that account for movement and collisions of atoms and molecules.In this work we adopt, for the first time, the Direct Simulation Monte Carlo (DSMC) method to compute the tortuosity factor of fibrous media in the rarefied regime. To enable realistic simulations of the actual transport of gases in the porous medium, digitized computational grids are obtained from X-ray micro-tomography imaging of real TPS materials. The SPARTA DSMC solver is used for simulations. Effective diffusion coefficients and tortuosity factors are obtained by computing the mean-square displacement of diffusing particles.We first apply the method to compute the tortuosity factors as a function of the Knudsen number for computationally designed materials such as random cylindrical fibers and packed bed of spheres with prescribed porosity. Results are compared to literature values obtained using random walk methods in the rarefied and transitional regime and a finite-volume method for the continuum regime. We then compute tortuosity factors for a real carbon fiber material with a transverse isotropic structure (FiberForm), quantifying differences between through-thickness and in-plain tortuosities at various Knudsen regimes.

Tortuosity↗