Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “event”

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 127 records · Page 7
All sourcesNASA NTRSDOE OSTI
NASA NTRS1991

Element abundances in solar (He-3) rich events

The results of a survey of over 228 (He-3)-rich events, with He-3/H-4 of more than 0.1, observed by the ISEE-3 spacecraft from 1978 August through 1991 April. In these events the elements above C are enhanced relative to the corresponding abundances in the corona and the degree of enhancement increases with Z or A. The flare-to-flare variations in abundances are different in character from those seen in large proton events. For example, using Fe/C as a parameter to describe these variations, the slope of the least-squares fits to ln(X/C) vs ln(Fe/C), where X is the intensity of N, O, Ne, Mg, Si, or S. Slopes of these abundance-correlation lines are much steeper for the (He-3)-rich events than for a sample of 36 large proton events, especially for Ne and heavier ions. The event-to-event variations for the (He-3)-rich events distinguish 3 groups of elements, He-O, Ne-S, and Fe. The abundances of all of the species within each group seem to vary in unison.

Reames, D. V.↗
NASA NTRS1994

Were all extinction events caused by impacts?

Extraterrestrial impacts are firmly implicated in several of the five major Phanerozoic extinction events. A critical issue now is whether extraterrestrial events have been the only mechanism that produced physical changes of sufficient magnitude to cause major extinction events. While we believe the evidence is overwhelming that the KT extinction event was caused by an impact, we also find that an event of similar or larger size near the end of the Ordovician is best explained by terrestrial causes. The Ordovician extinction event (End-O extinction event) occurred near the end of the Ordovician, but the interval of extinction was completed prior to the newly established Ordovician-Silurian boundary. In spite of extensive field studies, a convincing signature of an associated impact has not been found. However, a prominent glaciation does coincide with the End-O extinction event.

Sheehan, P. M.↗
NASA NTRS1994

The design and implementation of EPL: An event pattern language for active databases

The growing demand for intelligent information systems requires closer coupling of rule-based reasoning engines, such as CLIPS, with advanced data base management systems (DBMS). For instance, several commercial DBMS now support the notion of triggers that monitor events and transactions occurring in the database and fire induced actions, which perform a variety of critical functions, including safeguarding the integrity of data, monitoring access, and recording volatile information needed by administrators, analysts, and expert systems to perform assorted tasks; examples of these tasks include security enforcement, market studies, knowledge discovery, and link analysis. At UCLA, we designed and implemented the event pattern language (EPL) which is capable of detecting and acting upon complex patterns of events which are temporally related to each other. For instance, a plant manager should be notified when a certain pattern of overheating repeats itself over time in a chemical process; likewise, proper notification is required when a suspicious sequence of bank transactions is executed within a certain time limit. The EPL prototype is built in CLIPS to operate on top of Sybase, a commercial relational DBMS, where actions can be triggered by events such as simple database updates, insertions, and deletions. The rule-based syntax of EPL allows the sequences of goals in rules to be interpreted as sequences of temporal events; each goal can correspond to either (1) a simple event, or (2) a (possibly negated) event/condition predicate, or (3) a complex event defined as the disjunction and repetition of other events. Various extensions have been added to CLIPS in order to tailor the interface with Sybase and its open client/server architecture.

Giuffrida, G.↗
NASA NTRS1995

Solar Particle Induced Upsets in the TDRS-1 Attitude Control System RAM During the October 1989 Solar Particle Events

The three large solar particle events, beginning on October 19, 1989 and lasting approximately six days, were characterized by high fluences of solar protons and heavy ions at 1 AU. During these events, an abnormally large number of upsets (243) were observed in the random access memory of the attitude control system (ACS) control processing electronics (CPE) on-board the geosynchronous TDRS-1 (Telemetry and Data Relay Satellite). The RAM unit affected was composed of eight Fairchild 93L422 memory chips. The Galileo spacecraft, launched on October 18, 1989 (one day prior to the solar particle events) observed the fluxes of heavy ions experienced by TDRS-1. Two solid-state detector telescopes on-board Galileo, designed to measure heavy ion species and energy, were turned on during time periods within each of the three separate events. The heavy ion data have been modeled and the time history of the events reconstructed to estimate heavy ion fluences. These fluences were converted to effective LET spectra after transport through the estimated shielding distribution around the TDRS-1 ACS system. The number of single event upsets (SEU) expected was calculated by integrating the measured cross section for the Fairchild 93L422 memory chip with average effective LET spectrum. The expected number of heavy ion induced SEU's calculated was 176. GOES-7 proton data, observed during the solar particle events, were used to estimate the number of proton-induced SEU's by integrating the proton fluence spectrum incident on the memory chips, with the two-parameter Bendel cross section for proton SEU'S. The proton fluence spectrum at the device level was gotten by transporting the protons through the estimated shielding distribution. The number of calculated proton-induced SEU's was 72, yielding a total of 248 predicted SEU'S, very dose to the 243 observed SEU'S. These calculations uniquely demonstrate the roles that solar heavy ions and protons played in the production of SEU's during the October 1989 solar particle events.

Croley, D. R.↗
NASA NTRS2010

Response of the Antarctic Stratosphere to Two Types of El Nino Events

This study is the first to identify a robust El Nino/Southern Oscillation (ENSO) signal in the Antarctic stratosphere. El Nino events are classified as either conventional "cold tongue" events (positive SST anomalies in the Nino 3 region) or "warm pool" events (positive SST anomalies in the Nino 4 region). The ERA-40, NCEP and MERRA meteorological reanalyses are used to show that the Southern Hemisphere stratosphere responds differently to these two types of El Nino events. Consistent with previous studies, "cold tongue" events do not impact temperatures in the Antarctic stratosphere. During "warm pool" El Nino events, the poleward extension and increased strength of the South Pacific Convergence Zone (SPCZ) favor an enhancement of planetary wave activity during the SON season. On average, these conditions lead to higher polar stratospheric temperatures and a weakening of the Antarctic polar jet in November and December, as compared with neutral ENSO years. The phase of the quasi-biennial oscillation (QBO) modulates the stratospheric response to "warm pool" El Nino events: the strongest planetary wave driving events are coincident with the easterly phase of the QBO.

Hurwitz, M. M.↗
NASA NTRS2011

Intensity Variation of Solar Energetic Particle Events

This paper updates the influence of environmental and source factors of shocks driven by corona) mass ejections (CMEs) that are likely to influence the intensity of solar energetic particle (SEP) events. The intensity variation due to CME interaction reported in Gopalswamy et al. (2004, JGR 109, Al2105) is confirmed by expanding the investigation to all the large SEP events of solar cycle 23. The large SEP events are separated into two groups, one associated with CMEs running into other CMEs, and the other with CMEs running into the ambient solar wind. SEP events with CME interaction generally have a higher intensity. New possibilities such as the influence of corona) holes on the SEP intensity are also discussed. For example, the presence of a large coronal hole between a well-connected eruption and the solar disk center may render the shock poorly connected because of the interaction between the CME and the coronal hole. This point is illustrated using the 2004 December 3 SEP event delayed by about 12 hours from the onset of the associated CME. There is no other event at the Sun that can be associated with the SEP onset. This event is consistent with the possibility that the coronal hole interaction influences the connectivity of the CMEs that produce SEPs, and hence the intensity of the SEP event.

Gopalswamy, Nat↗
NASA NTRS2009

Concerning the Occurrence Pattern of Flux Transfer Events on the Dayside Magnetopause

We present an analytical model for the magnetic field perturbations associated with flux transfer events (FTEs) on the dayside magnetopause as a function of the shear between the magnetosheath and magnetospheric magnetic fields and the ratio of their strengths. We assume that the events are produced by component reconnection along subsolar reconnection lines with tilts that depend upon the orientation of the interplanetary magnetic field (IMF), and show that the amplitudes of the perturbations generated during southward IMF greatly exceed those during northward IMF As a result, even if the distributions of magnetic reconnection burst durations/event dimensions are identical during periods of northward and southward IMF orientation, events occurring for southward IMF orientations must predominate in surveys of dayside events. Two factors may restore the balance between events occurring for northward and southward IMF orientations on the flanks of the magnetosphere. Events generated on the dayside magnetopause during periods of southward IMF move poleward, while those generated during periods of northward IMF slip dawnward or duskward towards the flanks. Due to differing event and magnetospheric magnetic field orientations, events that produce weak signatures on the dayside magnetopause during intervals of northward IMF orientation may produce strong signatures on the flanks.

Sibeck, D. G.↗
NASA NTRS2011

Concerning the Motion and Orientation of Flux Transfer Events Produced by Component and Antiparallel Reconnection

We employ the Cooling et al. (2001) model to predict the location, orientation, motion, and signatures of flux transfer events (FTEs) generated at the solstices and equinoxes along extended subsolar component and high ]latitude antiparallel reconnection curves for typical solar wind plasma conditions and various interplanetary magnetic field (IMF) strengths and directions. In general, events generated by the two mechanisms maintain the strikingly different orientations they begin with as they move toward the terminator in opposite pairs of magnetopause quadrants. The curves along which events generated by component reconnection form bow toward the winter cusp. Events generated by antiparallel reconnection form on the equatorial magnetopause during intervals of strongly southward IMF orientation during the equinoxes, form in the winter hemisphere and only reach the dayside equatorial magnetopause during the solstices when the IMF strength is very large and the IMF points strongly southward, never reach the equatorial dayside magnetopause when the IMF has a substantial dawnward or duskward component, and never reach the equatorial flank magnetopause during intervals of northward and dawnward or duskward IMF orientation. Magnetosheath magnetic fields typically have strong components transverse to events generated by component reconnection but only weak components transverse to the axes of events generated by antiparallel reconnection. As a result, much stronger bipolar magnetic field signatures normal to the nominal magnetopause should accompany events generated by component reconnection. The results presented in this paper suggest that events generated by component reconnection predominate on the dayside equatorial and flank magnetopause for most solar wind conditions.

Sibeck, D. G.↗
NASA NTRS2011

Factors Affecting the Intensity of Solar Energetic Particle Events

This paper updates the influence of environmental and source factors of shocks driven by coronal mass ejections (CMEs) that are likely to influence the solar energetic particle (SEP) events. The intensity variation due to CME interaction reported that is confirmed by expanding the investigation to all the large SEP events of solar cycle 23. The large SEP events are separated into two groups, one associated with CMEs running into other CMEs, and the other with CMEs running into the ambient solar wind. SEP events with CME interaction generally have a higher intensity. New possibilities such as the influence of coronal holes on the SEP intensity are also discussed. For example, the presence of a large coronal hole between a well-connected eruption and the solar disk center may render the shock poorly connected because of the interaction between the CME and the coronal hole. This point is illustrated using the 2004 December 3 SEP event delayed by about 12 hours from the onset of the associated CME. There is no other event at the Sun that can be associated with the SEP onset. This event is consistent with the possibility that the coronal hole interaction influences the connectivity of the CMEs that produce SEPs, and hence the intensity of the SEP event.

Coronal mass ejections↗
NASA NTRS2012

Deflections of Fast Coronal Mass Ejections and the Properties of Associated Solar Energetic Particle Events

The onset times and peak intensities of solar energetic particle (SEP) events at Earth have long been thought to be influenced by the open magnetic fields of coronal holes (CHs). The original idea was that a CH lying between the solar SEP source region and the magnetic footpoint of the 1 AU observer would result in a delay in onset and/or a decrease in the peak intensity of that SEP event. Recently, Gopalswamy et al. showed that CHs near coronal mass ejection (CME) source regions can deflect fast CMEs from their expected trajectories in space, explaining the appearance of driverless shocks at 1 AU from CMEs ejected near solar central meridian (CM). This suggests that SEP events originating in CME-driven shocks may show variations attributable to CH deflections of the CME trajectories. Here, we use a CH magnetic force parameter to examine possible effects of CHs on the timing and intensities of 41 observed gradual E approx 20 MeV SEP events with CME source regions within 20 deg. of CM. We find no systematic CH effects on SEP event intensity profiles. Furthermore, we find no correlation between the CME leading-edge measured position angles and SEP event properties, suggesting that the widths of CME-driven shock sources of the SEPs are much larger than the CMEs. Independently of the SEP event properties, we do find evidence for significant CME deflections by CH fields in these events

coronal mass ejections (CMEs)↗
NASA NTRS2013

Source Regions of the Interplanetary Magnetic Field and Variability in Heavy-Ion Elemental Composition in Gradual Solar Energetic Particle Events

Gradual solar energetic particle (SEP) events are those in which ions are accelerated to their observed energies by interactions with a shock driven by a fast coronal mass-ejection (CME). Previous studies have shown that much of the observed event-to-event variability can be understood in terms of shock speed and evolution in the shock-normal angle. But an equally important factor, particularly for the elemental composition, is the origin of the suprathermal seed particles upon which the shock acts. To tackle this issue, we (1) use observed solar-wind speed, magnetograms, and the PFSS model to map the Sun-L1 interplanetary magnetic field (IMF) line back to its source region on the Sun at the time of the SEP observations; and (2) then look for correlation between SEP composition (as measured by Wind and ACE at approx. 2-30 MeV/nucleon) and characteristics of the identified IMF-source regions. The study is based on 24 SEP events, identified as a statistically-significant increase in approx. 20 MeV protons and occurring in 1998 and 2003-2006, when the rate of newly-emergent solar magnetic flux and CMEs was lower than in solar-maximum years and the field-line tracing is therefore more likely to be successful. We find that the gradual SEP Fe/O is correlated with the field strength at the IMF-source, with the largest enhancements occurring when the footpoint field is strong, due to the nearby presence of an active region. In these cases, other elemental ratios show a strong charge-to-mass (q/M) ordering, at least on average, similar to that found in impulsive events. These results lead us to suggest that magnetic reconnection in footpoint regions near active regions bias the heavy-ion composition of suprathermal seed ions by processes qualitatively similar to those that produce larger heavy-ion enhancements in impulsive SEP events. To address potential technical concerns about our analysis, we also discuss efforts to exclude impulsive SEP events from our event sample.

magnetic field↗
NASA NTRS2014

Data Albums: An Event Driven Search, Aggregation and Curation Tool for Earth Science

One of the largest continuing challenges in any Earth science investigation is the discovery and access of useful science content from the increasingly large volumes of Earth science data and related information available. Approaches used in Earth science research such as case study analysis and climatology studies involve gathering discovering and gathering diverse data sets and information to support the research goals. Research based on case studies involves a detailed description of specific weather events using data from different sources, to characterize physical processes in play for a specific event. Climatology-based research tends to focus on the representativeness of a given event, by studying the characteristics and distribution of a large number of events. This allows researchers to generalize characteristics such as spatio-temporal distribution, intensity, annual cycle, duration, etc. To gather relevant data and information for case studies and climatology analysis is both tedious and time consuming. Current Earth science data systems are designed with the assumption that researchers access data primarily by instrument or geophysical parameter. Those who know exactly the datasets of interest can obtain the specific files they need using these systems. However, in cases where researchers are interested in studying a significant event, they have to manually assemble a variety of datasets relevant to it by searching the different distributed data systems. In these cases, a search process needs to be organized around the event rather than observing instruments. In addition, the existing data systems assume users have sufficient knowledge regarding the domain vocabulary to be able to effectively utilize their catalogs. These systems do not support new or interdisciplinary researchers who may be unfamiliar with the domain terminology. This paper presents a specialized search, aggregation and curation tool for Earth science to address these existing challenges. The search tool automatically creates curated "Data Albums", aggregated collections of information related to a specific science topic or event, containing links to relevant data files (granules) from different instruments; tools and services for visualization and analysis; and information about the event contained in news reports, images or videos to supplement research analysis. Curation in the tool is driven via an ontology based relevancy ranking algorithm to filter out non-relevant information and data.

Ramachandran, Rahul↗
NASA NTRS2015

NASA's Meteoroid Environments Office's Response to Three Significant Bolide Events Over North America

Being the only U.S. Government entity charged with monitoring the meteor environment, the Meteoroid Environment Office has deployed a network of all sky and wide field meteor cameras, along with the appropriate software tools to quickly analyze data from these systems. However, the coverage of this network is still quite limited, forcing the incorporation of data from other cameras posted to the internet in analyzing many of the fireballs reported by the public and media. A procedure has been developed that determines the analysis process for a given fireball event based on the types and amount of data available. The differences between these analysis process will be explained and outlined by looking at three bolide events, all of which were large enough to produce meteorites. The first example is an ideal event - a bright meteor that occurred over NASA's All Sky Camera Network on August 2, 2014. With clear video of the event from various angles, a high-accuracy trajectory, beginning and end heights, orbit and approximate brightness/size of the event are able to be found very quickly using custom software. The bolide had the potential to have dropped meteorites, so dark flight analysis and modeling was performed, allowing potential fall locations to be mapped as a function of meteorite mass. The second case study was a bright bolide that occurred November 3, 2014 over West Virginia. This was just north of the NASA southeastern all-sky network, and just south of the Ohio-Pennsylvania network. This case study showcases the MEO's ability to use social media and various internet sources to locate videos of the event from obscure sources (including the Washington Monument) for anything that will permit a determination of a basic trajectory and fireball light curve The third case study will highlight the ability to use doppler weather radar in helping locate meteorites, which enable a definitive classification of the impactor. The input data and analysis steps differ for each case study, but the goals remain the same - a trajectory, orbit, and mass estimate for the bolide within hours of the event, and, for events with a high probability of producing meteorites, a location of the strewn field within a day.

Blaauw, Rhiannon C.↗
NASA NTRS2015

ANTARES Constrains a Blazar Origin of Two IceCube PeV Neutrino Events

Context. The source(s) of the neutrino excess reported by the IceCube Collaboration is unknown. The TANAMI Collaboration recently reported on the multiwavelength emission of six bright, variable blazars which are positionally coincident with two of the most energetic IceCube events. Objects like these are prime candidates to be the source of the highest-energy cosmic rays, and thus of associated neutrino emission. Aims. We present an analysis of neutrino emission from the six blazars using observations with the ANTARES neutrino telescope. Methods. The standard methods of the ANTARES candidate list search are applied to six years of data to search for an excess of muons - and hence their neutrino progenitors - from the directions of the six blazars described by the TANAMI Collaboration, and which are possibly associated with two IceCube events. Monte Carlo simulations of the detector response to both signal and background particle fluxes are used to estimate the sensitivity of this analysis for di erent possible source neutrino spectra. A maximum-likelihood approach, using the reconstructed energies and arrival directions of through-going muons, is used to identify events with properties consistent with a blazar origin. Results. Both blazars predicted to be the most neutrino-bright in the TANAMI sample (1653-329 and 1714-336) have a signal flux fitted by the likelihood analysis corresponding to approximately one event. This observation is consistent with the blazar-origin hypothesis of the IceCube event IC 14 for a broad range of blazar spectra, although an atmospheric origin cannot be excluded. No ANTARES events are observed from any of the other four blazars, including the three associated with IceCube event IC20. This excludes at a 90% confidence level the possibility that this event was produced by these blazars unless the neutrino spectrum is flatter than -2.4.

general↗
NASA NTRS2016

Landslides in West Coast Metropolitan Areas: The Role of Extreme Weather Events

Rainfall-induced landslides represent a pervasive issue in areas where extreme rainfall intersects complex terrain. A farsighted management of landslide risk requires assessing how landslide hazard will change in coming decades and thus requires, inter alia, that we understand what rainfall events are most likely to trigger landslides and how global warming will affect the frequency of such weather events. We take advantage of 9 years of landslide occurrence data compiled by collating Google news reports and of a high-resolution satellite-based daily rainfall data to investigate what weather triggers landslide along the West Coast US. We show that, while this landslide compilation cannot provide consistent and widespread monitoring everywhere, it captures enough of the events in the major urban areas that it can be used to identify the relevant relationships between landslides and rainfall events in Puget Sound, the Bay Area, and greater Los Angeles. In all these regions, days that recorded landslides have rainfall distributions that are skewed away from dry and low-rainfall accumulations and towards heavy intensities. However, large daily accumulation is the main driver of enhanced hazard of landslides only in Puget Sound. There, landslide are often clustered in space and time and major events are primarily driven by synoptic scale variability, namely "atmospheric rivers" of high humidity air hitting anywhere along the West Coast, and the interaction of frontal system with the coastal orography. The relationship between landslide occurrences and daily rainfall is less robust in California, where antecedent precipitation (in the case of the Bay area) and the peak intensity of localized downpours at sub-daily time scales (in the case of Los Angeles) are key factors not captured by the same-day accumulations. Accordingly, we suggest that the assessment of future changes in landslide hazard for the entire the West Coast requires consideration of future changes in the occurrence and intensity of atmospheric rivers, in their duration and clustering, and in the occurrence of short-duration (sub-daily) extreme rainfall as well. Major regional landslide events, in which multiple occurrences are recorded in the catalog for the same day, are too rare to allow a statistical characterization of their triggering events, but a case study analysis indicates that a variety of synoptic-scale events can be involved, including not only atmospheric rivers but also broader cold- and warm-front precipitation. That a news-based catalog of landslides is accurate enough to allow the identification of different landslide/ rainfall relationships in the major urban areas along the US West Coast suggests that this technology can potentially be used for other English-language cities and could become an even more powerful tool if expanded to other languages and non-traditional news sources, such as social media.

landslides↗
NASA NTRS2017

A Hierarchical Relationship Between the Fluence Spectra and CME Kinematics in Large Solar Energetic Particle Events: A Radio Perspective

We report on further evidence that solar energetic particles are organized by the kinematic properties of coronal mass ejections (CMEs). In particular, we focus on the starting frequency of type II bursts, which is related to the distance from the Sun where the radio emission starts. We find that the three groups of solar energetic particle (SEP) events known to have distinct values of CME initial acceleration, also have distinct average starting frequencies of the associated type II bursts. SEP events with ground level enhancement (GLE) have the highest starting frequency (107 MHz), while those associated with filament eruption (FE) in quiescent regions have the lowest starting frequency (22 MHz); regular SEP events have intermediate starting frequency (81 MHz). Taking the onset time of type II bursts as the time of shock formation, we determine the shock formation heights measured from the Sun center. We find that the shocks form on average closest to the Sun (1.51 Rs) in GLE events, farthest from the Sun in FE SEP events (5.38 Rs), and at intermediate distances in regular SEP events (1.72 Rs). Finally, we present the results of a case study of a CME with high initial acceleration (approx.3 km s-2) and a type II radio burst with high starting frequency (approx. 200 MHz) but associated with a minor SEP event. We find that the relation between the fluence spectral index and CME initial acceleration continues to hold even for this minor SEP event.

Gopalswamy, N.↗
NASA NTRS2017

A Hierarchical Relationship Between the Fluence Spectra and CME Kinematics in Large Solar Energetic Particle Events: A Radio Perspective

We report on further evidence that solar energetic particles are organized by the kinematic properties of coronal mass ejections (CMEs). In particular, we focus on the starting frequency of type II bursts, which is related to the distance from the Sun where the radio emission starts. We find that the three groups of solar energetic particle (SEP) events known to have distinct values of CME initial acceleration, also have distinct average starting frequencies of the associated type II bursts. SEP events with ground level enhancement (GLE) have the highest starting frequency (107 MHz), while those associated with filament eruption (FE) in quiescent regions have the lowest starting frequency (22 MHz); regular SEP events have intermediate starting frequency (81 MHz). Taking the onset time of type II bursts as the time of shock formation, we determine the shock formation heights measured from the Sun center. We find that the shocks form on average closest to the Sun (1.51 Rs) in GLE events, farthest from the Sun in FE SEP events (5.38 Rs), and at intermediate distances in regular SEP events (1.72 Rs). Finally, we present the results of a case study of a CME with high initial acceleration (approx. 3 km s-2) and a type II radio burst with high starting frequency (~200 MHz) but associated with a minor SEP event. We find that the relation between the fluence spectral index and CME initial acceleration continues to hold even for this minor SEP event.

Gopalswamy, N.↗
NASA NTRS2019

Deep Learning Method for Detecting Precursors to Adverse Events

With the recent advancements in Deep Learning methods, the ability to model large complex heterogeneous data sets are fundamentally changing industry and research. Coupled with hardware improvements, and ease of implementation, a wide variety of deep neural network architectures can quickly be developed to solve a sweeping range of problems such as: object detection in images, automatic healthcare diagnosis using heterogenous data sources, real time language translating and sentence prediction, upscaling low resolution images, and forecasting of multivariate timeseries. Generally, many of these architectures outperform classical machine learning approaches in their respective tasks, however, this typically comes at a cost of interpretability. These black box algorithms generally suffer from lack of transparency in both model complexity as well as the rationale behind the prediction. This lack of comprehension, is driving an emerging area of interest in “Explainable AI”. An algorithm called: “Deep Temporal Multiple Instance Learning”1 was a recently developed to identify precursors to adverse events and has been applied in the aviation domain. The deep learning architecture is designed to capture the evolution of the probability of the outcome over the time preceding the adverse event using a multiple instance learning approach as illustrated in Figure 1. Precursors are defined when the probability of the event has exceeded a threshold at some point in the timeseries, at which point, a sensitivity analysis is performed to determine contributing factors. The contributing factors are used to explain and define the precursor during the periods where the probability score is high. The identified contributing factors are then presented to subject matter experts to provide objective insights into the leading factors associated with the particular adverse event. The algorithm has been tested on flight data from a commercial airline and has the ability to discover precursors to known adverse events that take the form of safety critical operations, such as unstable approach events on final approach. Apart from detecting precursors to adverse events, the converse can also be leveraged to discover corrective actions. These positive actions manifest themselves as periods in the timeseries when the precursor score has been lowered from an elevated state; meaning that if the system had been left uncorrected, it would have eventually reached the adverse event state. Characterizing these state changes can help identify successful interventions that may not have been known before. Policy makers and procedure designers can use this additional knowledge to craft more safety and efficient resilient procedures for future operations and therefore improve the overall performance of the National Airspace.

Matthews, Bryan L.↗