Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “generalizability”

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 55 records · Page 3
All sourcesNASA NTRSDOE OSTI
NASA NTRS2017

Application of Human-Autonomy Teaming to an Advanced Ground Station for Reduced Crew Operations

Within human factors there is burgeoning interest in the "human-autonomy teaming" (HAT) concept as a way to address the challenges of interacting with complex, increasingly autonomous systems. The HAT concept comes out of an aspiration to interact with increasingly autonomous systems as a team member, rather than simply use automation as a tool. The authors, and others, have proposed core tenets for HAT that include bi-directional communication, automation and system transparency, and advanced coordination between human and automated teammates via predefined, dynamic task sequences known as "plays." It is believed that, with proper implementation, HAT should foster appropriate teamwork, thus increasing trust and reliance on the system, which in turn will reduce workload, increase situation awareness, and improve performance. To this end, HAT has been demonstrated and/or studied in multiple applications including search and rescue operations, healthcare and medicine, autonomous vehicles, photography, and aviation. The current paper presents one such effort to apply HAT. It details the design of a HAT agent, developed by Human Automation Teaming Solutions, Inc., to facilitate teamwork between the automation and the human operator of an advanced ground dispatch station. This dispatch station was developed to support a NASA project investigating a concept called Reduced Crew Operations (RCO); consequently, we have named the agent R-HATS. Part of the RCO concept involves a ground operator providing enhanced support to a large number of aircraft with a single pilot on the flight deck. When assisted by R-HATS, operators can monitor and support or manage a large number of aircraft and use plays to respond in real-time to complicated, workload-intensive events (e.g., an airport closure). A play is a plan that encapsulates goals, tasks, and a task allocation strategy appropriate for a particular situation. In the current implementation, when a play is initiated by a user, R-HATS determines what tasks need to be completed and has the ability to autonomously execute them (e.g., determining diversion options and uplinking new routes to aircraft) when it is safe and appropriate. R-HATS has been designed to both support end users and researchers in RCO and HAT. Additionally, R-HATS and its underlying architecture were developed with generalizability in mind as a modular software applicable outside of RCO/aviation domains. This paper will also discuss future further development and testing of RHATS.

automation↗
NASA NTRS2016

Maps Suggest Transport and Source Processes of PM2.5 at 1 km x 1 km for the Whole San Joaquin Valley, Winter 2011 (Generalizations from DISCOVER-AQ)

We present interpreted data analysis using MAIAC (Multiangle implementation of Atmospheric Correction) retrievals and appropriate RAPid Update Cycle (RAP) meteorology to map respirable aerosol (PM2.5) for the period January and February, 2011. The San Joaquin Valley is one of the unhealthiest regions in the USA for PM2.5 and related morbidity. The methodology evaluated can be used for the entire moderate-resolution imaging spectrometer (MODIS, VIIRS) data record. Other difficult areas of the West: Riverside, CA, Salt Lake City, UT, and Doa Ana County, NM share similar difficulties and solutions. The maps of boundary layer depth for 1116 hr local time from RAP allows us to interpret aerosol optical thickness as a concentration of particles in a nearly well-mixed box capped by clean air. That mixing is demonstrated by DISCOVER-AQ data and afternoon samples from the airborne measurements, P3B (on-board) and B200 (HSRL2 lidar). This data and the PM2.5 gathered at the deployment sites allowed us to estimate and then evaluate consistency and daily variation of the AOT to PM2.5 relationship. Mixed-effects modeling allowed a refinement of that relation from day to day; RAP mixed layers explained the success of previous mixed-effects modeling. Compositional, size-distribution, and MODIS angle-of-regard effects seem to describe the need for residual daily correction beyond ML depth.We report on an extension method to the entire San Joaquin Valley for all days with MODIS imagery using the permanent PM2.5 stations, evaluated for representativeness. Resulting map movies show distinct sources, particularly Interstate-5 (at approx. 1km x 1km resolution) and the broader Bakersfield area. Accompanying winds suggest transport effects and variable pathways of pollution cleanout. Such estimates should allow morbiditymortality studies. They should be also useful for actual model assimilations, where composition and sources are uncertain. We conclude with a description of new work to extend these insights to similar regions, e.g. interior valleys of California, the Po Valley, the Mediterranean litoral, and the Ganges Plain.This work show generalizable use of remote sensing, a major goal of DISCOVER-AQ, Deriving Information on Surface Conditions from COlumn and VERtically Resolved Observations Relevant to Air Quality.

Chatfield, R.↗
NASA NTRS2017

Flight Deck Surface Trajectory-Based Operations (STBO): A Four-Dimensional Trajectory (4DT) Simulation

Within human factors there is burgeoning interest in the Human-Autonomy Teaming (HAT) concept as away to address the challenges of interacting with complex, increasingly autonomous systems. The HAT concept comes out of an aspiration to interact with increasingly autonomous automation as a team member, rather than simply use automation as a tool. The authors, and others, have proposed core tenets for HAT that include bi-directional communication, automation and system transparency, and advanced coordination between human and automated teammates via predefined, dynamic task sequences known as plays (Shively et al., 2017). It is believed that, with proper implementation, HAT should foster appropriate teamwork, thus increasing trust and reliance on the system, which in turn will reduce workload, increase situation awareness, and improve performance. To this end, HAT has been demonstrated and/or studied in multiple applications including search and rescue operations (Nourbakhsh et al., 2005), healthcare and medicine (Tsui Yanco, 2007), autonomous vehicles (Parasuraman, Barnes, Cosenzo, Mulgund, 2007), photography (Lachter, Brandt, Sadler, Shively, in press), and aviation (Shively et al., in press). The current paper presents one such effort to apply HAT. It details the design of a R-HAT Agent developed as part of a NASA Research Agreement awarded to Human-Autonomy Teaming Solutions Inc. (HATS Inc), and developed in collaboration with the Human-Autonomy Teaming Laboratory at NASA Ames Research Center. The role of this Agent is to mediate interaction between the automation and the human operator of an advanced ground dispatch station, with this mediation based upon previously mentioned core tenets for HAT and the many lessons learned from the HAT research literature. This dispatch station was developed to support a NASA project investigating a concept called Reduced Crew Operations (RCO; Lachter, Brandt, Battiste, Matessa, Johnson, in press). Part of the RCO concept involves a ground operator providing enhanced support to a large number of aircraft with a single pilot on the flight deck. When assisted by the Agent, operators can monitor and support or manage a large number of aircraft and use plays to respond in real-time to complicated, workload-intensive events (e.g., an airport closure). A play is a plan that encapsulates goals, tasks, and a task allocation strategy appropriate for a particular situation. In the current implementation, when a play is initiated by a user, the Agent determines what tasks need to be done and has the ability to autonomously execute them (e.g., determining diversion options and uplinking new routes to aircraft) when it is safe and appropriate. The R-HAT Agent has been designed to both support end users and research in RCO and HAT. Additionally, the Agent and its underlying architecture were developed with generalizability in mind as a modular piece of software applicable outside of RCO aviation in domains such as those mentioned above. This paper will also discuss future further development and testing of the R-HAT Agent.

Bakowski, Deborah L.↗
NASA NTRS2019

Schedule Factors Associated with the Use of Controlled Rest in a Long-Haul Airline

Controlled Rest (CR) refers to a short, voluntary nap opportunity taken by pilots on the flight deck as a countermeasure to unanticipated sleepiness in flight. This study explores the profile of CR use in a long-haul commercial airline. Forty-four pilots wore actiwatches and filled in an application-based sleep/work diary for approximately 2 weeks. After merging sleep diary, actigraphy, and schedule data, complete records were analyzed from 240 flights. Sleep was estimated during CR intervals using the Philips Actiware 6.0.9 (Bend, OR) software with wake threshold set to medium. Timing of sleep periods and flight schedules were analyzed relative to home-base time. A mixed-effects binary logistics regression was used to analyze the impact of schedule factors on CR. Preliminary analyses revealed that CR was taken on 45% (n=107) of flights. Average sleep duration within these rest periods was estimated as 24.8 ± 16.1 minutes. CR was more frequent on return flights (arriving at home-base; 58%, n=69) vs. outbound flights (departing from home-base, 31%, n=38; p<0.001). There was no difference for direction of travel (eastbound: 49%, n= 56; westbound: 40%, n= 44; northbound/southbound: 50%, n=7; p=0.272). CR was more frequent on 2-pilot flights (62%, n=81) compared to augmented crew flights (24%, n=26; p<0.001). Of note, 21% (n=23) of augmented flights contained both CR and Bunk Rest. CR was more frequent on flights <10h duration (<10h: 59%, n=78; >10h: 27%, n=29; p<0.001). Flights departing between 12:00h-19:59h (25%, n=23; p<0.026) had the lowest frequency of CR. Data from this airline show that CR is most commonly used as a countermeasure to sleepiness on return, unaugmented, <10h duration, and nighttime flights. The direction of travel did not influence the use of CR. Future studies are required to determine generalizability of these results to other airlines.

Hilditch, Cassie↗
NASA NTRS2019

Training, Retention, and Transfer of Data Entry Perceptual and Motor Processes Over Short and Long Retention Intervals

In 2 experiments, subjects trained in a standard data entry task, which involved typing numbers (e.g., 2147) using their right hands. At an initial test (20 min or 6 months after training), subjects completed the standard task, followed by a left-hand variant (typing with their left hands) that involved the same perceptual, but different motoric, processes as the standard task. At a second test (2 days or 8 months after training), subjects completed the standard task, followed by a code variant (translating letters into digits, then typing the digits with their right hands) that involved different perceptual, but the same motoric, processes as the standard task. At test, for each of the three tasks, half the trials were trained numbers (old) and half were new. Repetition priming (faster response times to old than new numbers) was found for each task, with extended delays only slightly decreasing the magnitude of the effect. Repetition priming for the standard task reflects retention of trained numbers, for the left-hand variant reflects transfer of perceptual processes, and for the code variant reflects transfer of motoric processes. There was, thus, evidence for both specificity and generalizability of training data entry perceptual and motoric processes even over very long retention intervals.

transfer↗
NASA NTRS2020

Air Quality Monitoring Case Study Using Mobile Low-cost Sensors Mounted on Trash-Trucks: Methods Development and Lessons Learned

Air quality monitoring (AQM) is crucial for cities to develop management plans supporting population health. However, there is a dearth of measurements due to the high cost of standard reference instruments. Mobile AQM using low-cost sensors deployed on routine fleets of vehicles can enable the continuous detection of fine-scale pollutant variations in cities at a lower cost. New methods need to be developed to interpret these measurements. This paper presents three such methods. First, we propose a technique to identify aerosol hotspots. Second, we employ techniques published previously to assess the generalizable map of fine and coarse particle number concentrations, to understand qualitatively the contribution of local and regional sources across the region sampled. By using the raw number concentration of differently sized particles from the Optical Particle Counters (OPCs) instead of the noisier mass concentrations, we obtain more robust results. Third, in order to evaluate source signatures in cities, we propose another technique, in which we cluster the entire range of aerosol size-distribution measurements acquired. The properties of each cluster provide insight into the aerosol source characteristics in the sampling environment. We test these methods using a dataset we collected by mounting OPCs on two trash-trucks in Cambridge, Massachusetts.

Drive-by sensing↗
NASA NTRS2020

Air Quality Monitoring Case Study Using Mobile Low-cost Sensors mounted on Trash-Trucks: Methods Development and Lessons Learned

Air quality monitoring (AQM) is crucial for cities to develop management plans supporting population health. However, there is a dearth of measurements due to the high cost of standard reference instruments. Mobile AQM using low-cost sensors deployed on routine fleets of vehicles can enable the continuous detection of fine-scale pollutant variations in cities at a lower cost. New methods need to be developed to interpret these measurements. This paper presents three such methods. First, we propose a technique to identify aerosol hotspots. Second, we employ techniques published previously to assess the generalizable map of fine and coarse particle number concentrations, to understand qualitatively the contribution of local and regional sources across the region sampled. By using the raw number concentration of differently sized particles from the Optical Particle Counters (OPCs) instead of the noisier mass concentrations, we obtain more robust results. Third, in order to evaluate source signatures in cities, we propose another technique, in which we cluster the entire range of aerosol size-distribution measurements acquired. The properties of each cluster provide insight into the aerosol source characteristics in the sampling environment. We test these methods using a dataset we collected by mounting OPCs on two trash-trucks in Cambridge, Massachusetts.

Drive-by sensing↗
NASA NTRSDate not supplied

Algorithmic Detection of Elemental Biosignatures

Machine learning models that classify a sample as indicative or non-indicative of life could play an important role in life-detection missions. Their predictions result from agnostic algorithms and thereby add redundancy to judgements resulting from human expertise. Additionally, their important features can reveal the most informative measurements within the operational constraints of a life-detection mission. The Ladder of Life Detection (Neveu 2018) identifies the need for an understanding of how combinations of multiple biosignatures affect overall confidence. The present work provides a starting point to answer this need, and future work will expand the data types to obtain even more predictive combinations of features. Elemental abundance was chosen as a starting set of features due to its availability in diverse sample types, which are needed to train a generalizable model. A standardized dataset was collected, including 35 non-indicative, e.g., lunar rock, basalt; 19 indicative mixed, e.g., seawater, agricultural soil; 46 indicative non-alive, e.g., coal, chalk; and 10 indicative alive, e.g., biofilm, bacteria. This dataset could be valuable for complementary biosignature research. The samples were standardized to the same limit of detection of a simulated mission scenario. Four classification models were used: k-nearest neighbors (KNN), logistic regression (LR), linear support vector machines (SVM), and Gaussian naïve Bayes (GNB). To obtain feature importances, KNN was run on three principal components of the training data and LR and SVM were run with L1 and L2 regularization. The performances and feature importances of the six model variants on 40:60 train to validation ratios were assessed with Monte Carlo simulations. ROC AUC and mean accuracy scores ranged between 82% - 94%, with sensitivity greater than specificity. For indicative of life predictors, all models had C and Ca as strong and Cl as medium; a majority of models had N, K, and P as medium. For non-indicative of life predictors, all models had Si as strong, and a majority of models had Mg, Al, and Ti as medium. Varied elements were Fe (slightly non-indicative), H (slightly indicative), O (widely varied), Na, Mn, and S. These results serve as a proof of concept and suggest important elemental signals beyond merely the CHNOPS of Earth-based life.

Algorithmic↗
NASA NTRS2020

Mitigating fatigue on the flight deck: how is controlled rest used in practice?

Controlled Rest (CR) refers to a short, voluntary nap opportunity taken by pilots on the flight deck as a countermeasure to unanticipated sleepiness in flight. This study explores the profile of CR use in a long-haul commercial airline. Forty-four pilots wore actiwatches and filled in an application-based sleep/work diary for approximately 2 weeks resulting in complete records from 239 flights. Timing of sleep periods and flight schedules were analyzed relative to home-base time. Pearson correlations were used to assess the influence of pilot demographics on CR use. A mixed-effects logistic regression was used to analyze the impact of schedule factors on CR. CR was taken on 46% (n=110) of flights, with 80% (n=106/133) of all CR attempts estimated by actigraphy to have successfully achieved sleep. Average sleep duration during successful rest periods was estimated as 31.7 ± 12.2 minutes. CR was more frequent on return (60%, n=71) vs. outbound flights (33%, n=39); 2-pilot (69%, n=83) vs. >2-pilot flights (23%, n=27); <10h (63%, n=80) vs. >10h duration flights (27%, n=30); and night (55%, n=76) vs. day flights (34%, n=34) (all p≤0.001). There was no significant difference for direction of travel (eastbound: 51%, n= 57; westbound: 40%, n= 44; p=0.059). Of note, 22% (n=26) of augmented flights contained both CR and bunk rest. Data from this airline show that CR is most commonly used as a countermeasure to sleepiness on flights <10h duration (2-pilot crews) and home-base nighttime flights. Future studies are required to determine the generalizability of these results to other airlines.

sleepiness↗
NASA NTRSDate not supplied

Knowledge Discovery for Early Failure Assessment of Complex Engineered Systems Using Natural Language Processing

Emerging complex engineered systems may have unexpected safety issues due to novel operational environments, increasing autonomy, human-machine interaction, and other factors. To prevent failures in operation or testing that necessitate costly redesign, it is desirable to predict likely failure modes early in the design process. Text-based information about past engineering failures presents one possible solution by facilitating the retrieval of information that can inform new designs. However, identifying documents containing relevant information and extracting required information can be prohibitively time-consuming when implemented at scale. In this research, an automated natural language processing-based framework is proposed to discover relevant knowledge from documents containing failure-related design information. Documents containing usable information are filtered using sentiment analysis based on a custom lexicon specialized for engineering design and by filtering out documents containing only irrelevant topics. Next, from the identified usable documents, information relating to engineering failures, contributing factors that can be controlled at design time (“risk factors”), and recommended preventative actions are extracted. Semantic similarity is then used to group similar pieces of extracted information for improved generalizability. The proposed framework is applied to NASA’s Lessons Learned Information System (LLIS). The framework can be used to identify documents containing usable failure-related design information from other databases, extract relevant information from these documents, and generalize the acquired knowledge such that it can be applied to novel systems.

Sequoia R. Andrade↗
NASA NTRS2021

The Causes of Propeller Pitching Moment and the Conditions for its Significance

Recent development of vertical takeoff and landing (VTOL) aircraft has renewed interest in the study of propellers. One metric in particular, the propeller pitching moment, has been observed to be important to VTOL aircraft stability and control in the past. Propellers at angles of attack could not be accurately modeled in generations past due to a lack of computational power, but even with advances in computer technology, modern designers seem to possess insufficient knowledge in this area. In this dissertation, we study the physics behind propeller pitching moment in the context of an isolated propeller and a propeller upstream of a wing. An unsteady 3D vortex lattice method is developed specifically to model propellers at angles of attack and is validated by comparing to high-fidelity CFD analyses. We then use the model to isolate velocity influences to show that the propeller pitching moment is largely caused by two effects: a skewed wake and the presence of wing circulation. Generated maps of propeller pitching moment over a range of operational parameters corresponding to VTOL transition show that the low flight speeds and high angles of attack encountered during transition lead to significant magnitudes of propeller pitching moment that would be difficult to trim using passive methods. Also, derivation of a generalizable metric of significance shows that the peak contribution of propeller pitching moment to aircraft stability is comparable to a longitudinal displacement of the center of gravity by several percent of the wing chord. Finally, we give a concluding discussion on the impact of propeller pitching moment on VTOL aircraft design.

Xiaofan Fei↗
NASA NTRS2021

Exploration Mission Tasks: A Technical Manual

This technical manual is an abridgement of the Generalizable Skills and Knowledge for Exploration Missions (NASA/CR-2018-22045) report (Stuster et al., 2019), describing research conducted under Cooperative Agreement 80NSSC18K0042 for the Human Factors and Behavioral Performance Element, Human Research Program, located at the National Aeronautics and Space Administration’s (NASA) Johnson Space Center. The research identified tasks that will be conducted by human crew during an expedition to Mars, and the abilities, skills, and knowledge that will be required of crew members. The 3-year study uses research methods that were developed to analyze the work performed by a variety of civilian and military occupational specialties and is consistent with Human Factors methods. The work began by developing a comprehensive inventory of 1,125 tasks that are likely to be performed during the 12 phases of the first human expeditions to Mars, from launch to landing 30 months later. Sixty subject matter experts (SMEs) rated expedition tasks in terms of (likely) frequency of performance, difficulty to learn, and importance to mission success; a fourth metric (criticality), was derived by summing the mean ratings of the three dimensions. Seventy-two SMEs placed the physical, cognitive, and social abilities necessary to perform the tasks in order of importance for specialist domains identified by the task analysis. The research team then identified: 1) Abilities, skills, and knowledge that can be retained and generalized across tasks and 2) Implications for crew size and composition. Study results also led to recommendations concerning equipment, habitats, and procedures for exploration-class space missions. Note: The full-mission task inventory was developed during a comprehensive review of documentation and concepts of operations. It was understood by the study team that the tasks were based on currently available information, and that the tools, equipment, propulsion methods, and/or other aspects of actual human expeditions to Mars might be different from those described here, as a consequence of technological development and evolving Mars Design Reference Missions. The purpose and scope of this technical manual is to present the core, actionable information that resulted from this research. The abridged format is intended to address the needs of development and research teams to quickly access, discern, and use the information in the course of their exploration-related work.

Brandin Munson↗
NASA NTRSDate not supplied

Controlled Rest: Investigating the Use of an In-Flight Sleepiness Countermeasure

INTRODUCTION: Sleepiness is commonly reported amongst commercial airline pilots and is recognized as a safety risk due to its impact on performance. Controlled Rest (CR) refers to a short, voluntary nap opportunity taken by pilots on the flight deck as a countermeasure to unanticipated sleepiness in flight. This study explores the profile of CR use in a long-haul commercial airline. METHODS: Forty-four pilots filled in an application-based sleep/work diary and wore actiwatches for approximately 2 weeks. Complete data sets from 239 flights including sleep diaries, actigraphy, and schedules were merged and analyzed. Sleep diary entries were used to set CR intervals in the actigraphy software, which was then used to predict sleep within these intervals. All time-stamps of sleep periods and flight schedules were adjusted for home-base time of the pilots. Pearson Correlations were used to assess the influence of pilot demographics on CR use. A mixed-effects logistic regression was used to analyze the impact of schedule factors on CR. RESULTS: Pilots reported taking CR on 46% (n=110) of observed flights. Average CR attempt duration was 43.1 ± 11.0 minutes. Eighty-percent (n=106/133) of all CR attempts were estimated by actigraphy to have successfully achieved sleep with an average sleep duration during successful rest periods of 31.7 ± 12.2 minutes. Captains reported taking CR on 38% of flights (n=39/102), compared to First Officers reporting 52% (n=71/137) of flights with CR (p=0.131). Age, experience, BMI, and sleep need were not associated with the percentage of flights with CR (all p>0.244). The following schedule factors were associated with a higher frequency of CR: night (55%, n=76) vs. day flights (34%, n=34); <10h (63%, n=80) vs. >10h duration flights (27%, n=30); return (60%, n=71) vs. outbound flights (33%, n=39); and 2-pilot (69%, n=83) vs. >2-pilot flights (23%, n=27) (all p≤0.001). There was a trend for more CR on eastbound flights, but this was not significant (eastbound: 51%, n= 57; westbound: 40%, n= 44; p=0.059). Of note, 22% (n=26) of augmented flights (>2-pilots) contained both CR and Bunk Rest (in a designated rest facility). DISCUSSION: Data from this airline show that pilots commonly use CR to mitigate sleepiness in-flight, especially on flights <10h duration and during home-base nighttime flights. Future studies are required to determine generalizability of these results to other airlines.

sleepiness↗
NASA NTRSDate not supplied

Attitude Reconstruction of Free-Flight CFD Generated Trajectories Using Non-Linear Pitch Damping Coefficient Curves

Attitude history reconstruction of Free-flight CFD generated trajectories with non-linear pitch damping coefficient curves is investigated. Free-flight CFD simulations of the capsule shape used for the Genesis sample return mission and the upcoming Dragonfly mission to Titan are conducted for 1-, 2-, and 3-degree-of-freedom cases. Two different data reduction methodologies are employed to derive a pitch damping curve as a function of instantaneous angle of attack. These curves are then used to reconstruct the attitude history of the body which is compared to the raw simulation results. While both data reduction methods produce pitch damping curves that can generally reconstruct the motion seen in the Free-flight simulations, it is found that optimization of the pitch damping curve using an inverse estimation process yields superior and more generalizable results. Further refinement of this technique could allow pitch damping curves derived using CFD to serve as a capability complementary to existing techniques for dynamic stability characterization.

entry↗
NASA NTRSDate not supplied

Attitude Reconstruction of Free-Flight CFD Generated Trajectories Using Non-Linear Pitch Damping Coefficient Curves

Attitude history reconstruction of Free-flight CFD generated trajectories with non-linear pitch damping coefficient curves is investigated. Free-flight CFD simulations of the capsule shape used for the Genesis sample return mission and the upcoming Dragonfly mission to Titan are conducted for 1-, 2-, and 3-degree-of-freedom cases. Two different data reduction methodologies are employed to derive a pitch damping curve as a function of instantaneous angle of attack. These curves are then used to reconstruct the attitude history of the body which is compared to the raw simulation results. While both data reduction methods produce pitch damping curves that can generally reconstruct the motion seen in the Free-flight simulations, it is found that optimization of the pitch damping curve using an inverse estimation process yields superior and more generalizable results. Further refinement of this technique could allow pitch damping curves derived using CFD to serve as a capability complementary to existing techniques for dynamic stability characterization.

entry↗
NASA NTRSDate not supplied

Bayesian Model Selection for Reducing Bloat and Overfitting in Genetic Programming for Symbolic Regression

When performing symbolic regression using genetic programming, overfitting and bloat can negatively impact generalizability and interpretability of the resulting equations as well as increase computation times. A Bayesian fitness metric is introduced and its impact on bloat and overfitting during population evolution is studied and compared to common alternatives in the literature. The proposed approach was found to be more robust to noise and data sparsity in numerical experiments, guiding evolution to a level of complexity appropriate to the dataset. Further evolution of the population resulted not in overfitting or bloat, but rather in slight simplifications in model form. The ability to identify an equation of complexity appropriate to the scale of noise in the training data was also demonstrated. In general, the Bayesian model selection algorithm was shown to be an effective means of regularization which resulted in less bloat and overfitting when any amount of noise was present in the training data.

G F Bomarito↗
NASA NTRSDate not supplied

What Went Wrong: A Survey of Wildfire UAS Mishaps through Named Entity Recognition

Increasingly, unmanned aircraft systems (UAS) are being applied to wildfire incidents for tasks such as mapping, aerial ignition, and delivery. As a result, incident reporting systems for wildfires are beginning to accumulate data related to UAS mishaps in wildfire response. In this research, we apply state-of-the-art natural language processing (NLP) techniques to develop a custom Named Entity Recognition (NER) model which extracts a Failure Modes and Effects Analysis (FMEA)-style survey of wildfire UAS mishaps reported in SAFECOM. The custom NER model is built by fine-tuning an existing (BERT) model, resulting in a generalizable NER model that can extract engineering relevant entities including failure modes, causes, effects, control processes, and recommendations from any failure-relevant text. Similar mishaps are clustered and reported as single rows within the FMEA. For each cluster, frequency, severity, and overall risk are computed. The methodology can be applied as part of a broader safety management system to track trends in mishaps and discover knowledge that can be utilized to improve safety outcomes and system performance.

Machine Learning↗
NASA NTRSDate not supplied

Automated Probabilistic Finite Element Model Calibration Tool Based on Uncertainty Quantification and Machine Learning

Qualification and certification of safety critical parts is a hurdle to the adoption of metallic additively manufactured components for aerospace vehicle applications. Challenges include variability in part properties due to inconsistent defect distribution and microstructure. Understanding of the process through finite element modeling (FEM), and process control through in-situ monitoring, may result in significant improvements; however, solutions useful to manufacturers will require large volumes of data and automated data utilization. Toward this end, a generalizable automated FEM calibration paradigm is developed. This paradigm leverages existing and novel tools from machine learning and uncertainty quantification to enable the automatic calibration of FEMs without requiring prior knowledge of the model performance across input parameter space, including meshing and solver settings, which can require time consuming manual model probing or cause noisy and inconsistent predictions. The result is a probabilistic distribution of calibrated and validated FEM input parameters targeting measured data.

Additive manufacturing model calibration finite el↗