Search NASA⌕ Search

SEARCH · Search NASA

Results for “hierarchical optimization”

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.

690 records · Page 39

Report on activities and findings under DOE grant “Collaborative research. An Interactive Multi-Model for Consensus on Climate Change”

The project takes a hierarchical approach. The supermodeling scheme was first studied exhaustively with simple systems of ordinary differential equations. Results were described in detail in the previous report. The principal findings were that 1) for highly non-linear systems, such as Lorenz-63, including systems which describe phenomena on very different (atmosphere/ocean) times scales, supermodeling is far superior to any form of output-averaging; 2) negative coefficients can be used to advantage in situations where all models err in the same way, but to different degrees; 3) an interesting variant of supermodeling, “weighted supermodeling”, is the limiting case where inter-model nudging coefficients in the originally conceived “connected supermodel” become infinite, but with fixed ratios, corresponding to a direct combination of the tendencies that appear in corresponding equations for the alternative models; 4) noise is useful for avoiding local optima in training the inter-model coefficients in the supermodel. The supermodeling scheme was then investigated with simple quasigeostrophic (QG) models. As described in the previous report, it was found that QG models on a sphere can be coupled most efficaciously by working in a basis which captures the most variance, rather than the most instability, a somewhat unexpected result that still deserves scrutiny in a broader context. Further studies (since the last report) with QG channel models addressed the central question of when supermodeling is superior to output averaging in situations where nonlinearites are less extreme than with the ODEs initially studied. It was found that for realistic variations in a parameter in the QG model, output averaging is sufficient to capture all but the most subtle quantitative and qualitative behavior. Supermodeling helps when qualitative differences between the models result from unrealistically large parameter differences, or when very detailed spatial structure of the modes of variability are of interest. Therefore, the scheme may still be useful in the case of full climate models with qualitatively different parametrization schemes. A supermodel was constructed from the intermediate-complexity SPEEDO model, a primitive equation model with ocean and land. Versions defined by different parameter choices, in a realistic range, were connected and the coefficients trained. Some improvement was found as compared to output averaging. The learning algorithm used thus far gives sub-optimal, but still useful results when the CO 2 level and other parameters are varied. Spatial structure remains to be studied. The first use of supermodeling with full climate models has been with variants of the ECHAM model that use different convection schemes. As yet the models are only connected at the ocean-atmosphere interface, where weighted combinations of fluxes from the two atmospheres are passed to a common ocean, and the weights adapted during a training period. The supermodel was surprisingly successful at avoiding unrealistic features such as the double-ITCZ (Intertropical Convergence Zone), a problem that arises in both of the two models run separately. The supermodels constructed thus far have not identified dynamical regime shifts in future climate. Thus the planned connection with the work of Tsonis on the relationship between regime shifts and synchronization/de-synchronization among the major climate modes (see U. Wisconsin report) has not yet been made. However the network analysis of the climate system, in observations and models, that was done in conjunction with that study, shows that models differ strongly from one another and from observations in regard to the dynamical structure described by correlation networks [Steinhaeuser and Tsonis 2013], providing a further justification for supermodeling. Toward a general software framework for supermodeling, three versions of CAM (the Community Atmosphere Model) at NCAR were configured for inter-model nudging using the DART (Data Assimilation Research Testbed) capability to stop and re-start models in synchrony. It was clearly established that the inter-model nudging adds almost no computational burden to the runs, but there appears to be a problem with the re-initialization software that is still being debugged. Publications: Several papers were published on the basic idea of the interactive multi-model (supermodel) including demonstrations with low-order ODEs. The last of these, a semi-philosophical review paper on the relevance of synchronization generally, encountered considerable resistance but was finally published in Entropy [Duane 2015]. A paper on the ECHAM/COSMOS supermodel, containing the most promising results so far [Shen et al. 2015] is presently under review.

58 GEOSCIENCES↗

Valence Disproportionation of GeS in the PbS Matrix Forms Pb 5 Ge 5 S 12 Inclusions with Conduction Band Alignment Leading to High n-Type Thermoelectric Performance

Converting waste heat into useful electricity using solid-state thermoelectrics has a potential for enormous global energy savings. Lead chalcogenides are among the most prominent thermoelectric materials, whose performance decreases with an increase in chalcogen amounts (e.g., PbTe>PbSe>PbS). Herein, we demonstrate the simultaneous optimization of the electrical and thermal transport properties of PbS-based compounds by alloying with GeS. The addition of GeS triggers a complex cascade of beneficial events as follows: Ge 2+ substitution in Pb 2+ and discordant off-center behavior; formation of Pb 5 Ge 5 S 12 as stable second phase inclusions through valence disproportionation of Ge 2+ to Ge0 and Ge 4+ . PbS and Pb 5 Ge 5 S 12 exhibit good conduction band energy alignment that preserves the high electron mobility; the formation of Pb 5 Ge 5 S 12 increases the electron carrier concentration by introducing S vacancies. Sb doping as the electron donor produces a large power factor and low lattice thermal conductivity (κ lat ) of ~0.61 Wm -1 K -1 . The highest performance was obtained for the 14% GeS-alloyed samples, which exhibited an increased room temperature electron mobility of ~121 cm 2 V -1 s -1 for 3 × 10 19 cm -3 carrier density, and a ZT, of 1.32 at 923 K. This is ~ 55% greater that the corresponding Sb-doped PbS sample and is one of the highest reported for the n-type PbS system. Moreover, the average ZT (ZT avg ) of ~0.76 from 400 to 923 K is the highest for PbS-based systems.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Synergistic Computational and Microstructural Design of Next- Generation High-Temperature Austenitic Stainless Steels

The purpose of this project was to: 1) study deformation twinning, its evolution, thermal stability, and the contribution on mechanical response of the new advanced stainless steels, especially at elevated temperatures; 2) study alumina-scale formation on the surface, as an alternative for conventional chromium oxide, that shows better oxidation resistance, through alloy design; and 3) design new generation of high temperature stainless steels that form alumina scale and have thermally stable nano-twins. The work involved few baseline alloys for investigating the twin formation under tensile loading, thermal stability of these twins, and the role of deformation twins on the mechanical response of the alloys. These baseline alloys included Hadfield Steel (Fe-13Mn-1C), 316, 316L and 316N stainless steels. Another baseline alloy was studied for alumina-scale formation investigations. Hadfield steel showed twinning but undesired second phases formed at higher temperatures. 316N stainless steel did not show signs of deformation twinning. Conventional 316 stainless steel demonstrated extensive deformation twinning at room temperature. Investigations on this alloy, both in single crystalline and polycrystalline forms, showed that deformation twins evolve in a hierarchical manner, consisting of micron–sized bundles of nano-twins. The width of nano-twins stays almost constant as the extent of strain increases, but the width and number of the bundles increase with increasing strain. A systematic thermomechanical cycling study showed that the twins were stable at temperatures as high as 900°C, after the dislocations are annealed out. Using such cycles, volume fraction of the thermally stable deformation twins were increased up to 40% in 316 stainless steel. Using computational thermodynamics and kinetics calculations, we designed two generations of advanced austenitic stainless steels. In the first generation, Alloy 1, which had been proposed as an alumina-forming austenitic stainless steel, is fully austenitic, but possesses carbides that were not dissolvable and could not be controlled. This alloy also did not show deformation twinning. Alloy 2 was designed based on alloy 1, but was not fully austenitic and had significant traces of uncontrollable precipitates as well. Alloy 3, also designed based on alloy 1, was mainly austenitic with evolution of a second phase along the grain boundaries, but also had precipitates that were not controllable. Based on the knowledge gained from the first generation of the designed steels, two more steels, called PGAA1 and PGAA2, were proposed using genetic algorithms that, based on the modelling, were supposed to exhibit alumina-scale formation. PGAA1, however, did not demonstrate a fully austenitic structure. PGAA2 could achieve a mostly austenitic structure through thermo-mechanical processing, and was then used for oxidation tests. The oxidation tests of PGAA2, with and without nitrogen impurities, along with alloy 1, suggested that PGAA2 can form alumina-scale similar to alloy 1, but N impurity will prevent formation of such a scale, probably through formation of aluminum nitrides. For the above mentioned genetic algorithm framework of alloy design, separate models were developed for specific design criteria. For prediction of alumina formation in stainless steels, a model was constructed based off of two criteria – effective valence and third element effect. These criteria capture the thermodynamics and kinetics of alumina formation in steels. To test the efficacy and robustness of this model, they were tested against alloys in the literature which had been experimentally verified to exhibit alumina formation and the predictions were in excellent agreement with the experiments. Another meta-model for prediction of twinning in unknown steel compositions was developed by an informatics based machine learning/data mining approach. Stacking Fault Energy data was captured from the literature for a large number of steel compositions and then this data was used to build a classifier to predict deformation mechanisms. Here a training set-test set based analysis was performed to test performance. The above genetic algorithm based optimization framework for alloy design was exhibited to be a successful methodology for accelerated materials discovery in the context of alloy design.

36 MATERIALS SCIENCE↗

X-ray Optics for BES Light Source Facilities

Each new generation of synchrotron radiation sources has delivered an increase in average brightness 2 to 3 orders of magnitude over the previous generation. The next evolution toward diffraction-limited storage rings will deliver another 3 orders of magnitude increase. For ultrafast experiments, free electron lasers (FELs) deliver 10 orders of magnitude higher peak brightness than storage rings. Our ability to utilize these ultrabright sources, however, is limited by our ability to focus, monochromate, and manipulate these beams with X-ray optics. X-ray optics technology unfortunately lags behind source technology and limits our ability to maximally utilize even today’s X-ray sources. With ever more powerful X-ray sources on the horizon, a new generation of X-ray optics must be developed that will allow us to fully utilize these beams of unprecedented brightness. The increasing brightness of X-ray sources will enable a new generation of measurements that could have revolutionary impact across a broad area of science, if optical systems necessary for transporting and analyzing X-rays can be perfected. The high coherent flux will facilitate new science utilizing techniques in imaging, dynamics, and ultrahigh-resolution spectroscopy. For example, zone-plate-based hard X-ray microscopes are presently used to look deeply into materials, but today’s resolution and contrast are restricted by limitations of the current lithography used to manufacture nanodiffractive optics. The large penetration length, combined in principle with very high spatial resolution, is an ideal probe of hierarchically ordered mesoscale materials, if zone-plate focusing systems can be improved. Resonant inelastic X-ray scattering (RIXS) probes a wide range of excitations in materials, from charge-transfer processes to the very soft excitations that cause the collective phenomena in correlated electronic systems. However, although RIXS can probe high-energy excitations, the most exciting and potentially revolutionary science involves soft excitations such as magnons and phonons; in general, these are well below the resolution that can be probed by today’s optical systems. The study of these low-energy excitations will only move forward if advances are made in high-resolution gratings for the soft X-ray energy region, and higher-resolution crystal analyzers for the hard X-ray region. In almost all the forefront areas of X-ray science today, the main limitation is our ability to focus, monochromate, and manipulate X-rays at the level required for these advanced measurements. To address these issues, the U.S. Department of Energy (DOE) Office of Basic Energy Sciences (BES) sponsored a workshop, X-ray Optics for BES Light Source Facilities, which was held March 27–29, 2013, near Washington, D.C. The workshop addressed a wide range of technical and organizational issues. Eleven working groups were formed in advance of the meeting and sought over several months to define the most pressing problems and emerging opportunities and to propose the best routes forward for a focused R&D program to solve these problems. The workshop participants identified eight principal research directions (PRDs), as follows: Development of advanced grating lithography and manufacturing for high-energy resolution techniques such as soft X-ray inelastic scattering. Development of higher-precision mirrors for brightness preservation through the use of advanced metrology in manufacturing, improvements in manufacturing techniques, and in mechanical mounting and cooling. Development of higher-accuracy optical metrology that can be used in manufacturing, verification, and testing of optomechanical systems, as well as at wavelength metrology that can be used for quantification of individual optics and alignment and testing of beamlines. Development of an integrated optical modeling and design framework that is designed and maintained specifically for X-ray optics. Development of nanolithographic techniques for improved spatial resolution and efficiency of zone plates. Development of large, perfect single crystals of materials other than silicon for use as beam splitters, seeding monochromators, and high-resolution analyzers. Development of improved thin-film deposition methods for fabrication of multilayer Laue lenses and high-spectral-resolution multilayer gratings. Development of supports, actuator technologies, algorithms, and controls to provide fully integrated and robust adaptive X-ray optic systems. Development of fabrication processes for refractive lenses in materials other than silicon. The workshop participants also addressed two important nontechnical areas: our relationship with industry and organization of optics within the light source facilities. Optimization of activities within these two areas could have an important effect on the effectiveness and efficiency of our overall endeavor. These are crosscutting managerial issues that we identified as areas that needed further in-depth study, but they need to be coordinated above the individual facilities. Finally, an issue that cuts across many of the optics improvements listed above is routine access to beamlines that ideally are fully dedicated to optics research and/or development. The success of the BES X-ray user facilities in serving a rapidly increasing user community has led to a squeezing of beam time for vital instrumentation activities. Dedicated development beamlines could be shared with other R&D activities, such as detector programs and novel instrument development. In summary, to meet the challenges of providing the highest-quality X-ray beams for users and to fully utilize the high-brightness sources of today and those that are on the horizon, it will be critical to make strategic investments in X-ray optics R&D. This report can provide guidance and direction for effective use of investments in the field of X-ray optics and potential approaches to develop a better-coordinated program of X-ray optics development within the suite of BES synchrotron radiation facilities. Due to the importance and complexity of the field, the need for tight coordination between BES light source facilities and with industry, as well as the rapid evolution of light source capabilities, the workshop participants recommend holding similar workshops at least biannually.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

NASA System Safety Handbook: System Safety Framework and Concepts for Implementation - Volume 1

System safety assessment is defined in NPR 8715.3C, NASA General Safety Program Requirements as a disciplined, systematic approach to the analysis of risks resulting from hazards that can affect humans, the environment, and mission assets. Achievement of the highest practicable degree of system safety is one of NASA's highest priorities. Traditionally, system safety assessment at NASA and elsewhere has focused on the application of a set of safety analysis tools to identify safety risks and formulate effective controls.1 Familiar tools used for this purpose include various forms of hazard analyses, failure modes and effects analyses, and probabilistic safety assessment (commonly also referred to as probabilistic risk assessment (PRA)). In the past, it has been assumed that to show that a system is safe, it is sufficient to provide assurance that the process for identifying the hazards has been as comprehensive as possible and that each identified hazard has one or more associated controls. The NASA Aerospace Safety Advisory Panel (ASAP) has made several statements in its annual reports supporting a more holistic approach. In 2006, it recommended that "... a comprehensive risk assessment, communication and acceptance process be implemented to ensure that overall launch risk is considered in an integrated and consistent manner." In 2009, it advocated for "... a process for using a risk-informed design approach to produce a design that is optimally and sufficiently safe." As a rationale for the latter advocacy, it stated that "... the ASAP applauds switching to a performance-based approach because it emphasizes early risk identification to guide designs, thus enabling creative design approaches that might be more efficient, safer, or both." For purposes of this preface, it is worth mentioning three areas where the handbook emphasizes a more holistic type of thinking. First, the handbook takes the position that it is important to not just focus on risk on an individual basis but to consider measures of aggregate safety risk and to ensure wherever possible that there be quantitative measures for evaluating how effective the controls are in reducing these aggregate risks. The term aggregate risk, when used in this handbook, refers to the accumulation of risks from individual scenarios that lead to a shortfall in safety performance at a high level: e.g., an excessively high probability of loss of crew, loss of mission, planetary contamination, etc. Without aggregated quantitative measures such as these, it is not reasonable to expect that safety has been optimized with respect to other technical and programmatic objectives. At the same time, it is fully recognized that not all sources of risk are amenable to precise quantitative analysis and that the use of qualitative approaches and bounding estimates may be appropriate for those risk sources. Second, the handbook stresses the necessity of developing confidence that the controls derived for the purpose of achieving system safety not only handle risks that have been identified and properly characterized but also provide a general, more holistic means for protecting against unidentified or uncharacterized risks. For example, while it is not possible to be assured that all credible causes of risk have been identified, there are defenses that can provide protection against broad categories of risks and thereby increase the chances that individual causes are contained. Third, the handbook strives at all times to treat uncertainties as an integral aspect of risk and as a part of making decisions. The term "uncertainty" here does not refer to an actuarial type of data analysis, but rather to a characterization of our state of knowledge regarding results from logical and physical models that approximate reality. Uncertainty analysis finds how the output parameters of the models are related to plausible variations in the input parameters and in the modeling assumptions. The evaluation of unrtainties represents a method of probabilistic thinking wherein the analyst and decision makers recognize possible outcomes other than the outcome perceived to be "most likely." Without this type of analysis, it is not possible to determine the worth of an analysis product as a basis for making decisions related to safety and mission success. In line with these considerations the handbook does not take a hazard-analysis-centric approach to system safety. Hazard analysis remains a useful tool to facilitate brainstorming but does not substitute for a more holistic approach geared to a comprehensive identification and understanding of individual risk issues and their contributions to aggregate safety risks. The handbook strives to emphasize the importance of identifying the most critical scenarios that contribute to the risk of not meeting the agreed-upon safety objectives and requirements using all appropriate tools (including but not limited to hazard analysis). Thereafter, emphasis shifts to identifying the risk drivers that cause these scenarios to be critical and ensuring that there are controls directed toward preventing or mitigating the risk drivers. To address these and other areas, the handbook advocates a proactive, analytic-deliberative, risk-informed approach to system safety, enabling the integration of system safety activities with systems engineering and risk management processes. It emphasizes how one can systematically provide the necessary evidence to substantiate the claim that a system is safe to within an acceptable risk tolerance, and that safety has been achieved in a cost-effective manner. The methodology discussed in this handbook is part of a systems engineering process and is intended to be integral to the system safety practices being conducted by the NASA safety and mission assurance and systems engineering organizations. The handbook posits that to conclude that a system is adequately safe, it is necessary to consider a set of safety claims that derive from the safety objectives of the organization. The safety claims are developed from a hierarchy of safety objectives and are therefore hierarchical themselves. Assurance that all the claims are true within acceptable risk tolerance limits implies that all of the safety objectives have been satisfied, and therefore that the system is safe. The acceptable risk tolerance limits are provided by the authority who must make the decision whether or not to proceed to the next step in the life cycle. These tolerances are therefore referred to as the decision maker's risk tolerances. In general, the safety claims address two fundamental facets of safety: 1) whether required safety thresholds or goals have been achieved, and 2) whether the safety risk is as low as possible within reasonable impacts on cost, schedule, and performance. The latter facet includes consideration of controls that are collective in nature (i.e., apply generically to broad categories of risks) and thereby provide protection against unidentified or uncharacterized risks.

Dezfuli, Homayoon↗

A Machine Learning Approach to Improve Air Traffic Management Initiatives

Collaborating closely with commercial air carriers and related organizations, the Federal Aviation Administration(FAA) regulates air traffic and ensures the safety and efficiency of air operations. Air traffic controllers make strategic decisions, such as delaying, rerouting, or canceling flights, partly based on guidance provided by the FAA’s Air TrafficControl System Command Center (ATCSCC). The guidance includes, among other things, control measures known asTraffic Management Initiatives (TMIs) designed to enhance safety and improve operational efficiency. TMIs play a crucial role in managing the demand and capacity within the U.S. National Airspace System (NAS). Two major TMIs that are routinely used (primarily to mitigate the adverse effects of bad weather) are Ground Delay Programs (GDPs) andGround Stops (GSs). In a GDP, flights destined for airports facing thunderstorm activity experience delays at their origin airports. This proactive approach minimizes the risk of routing aircraft through hazardous weather conditions and also replaces (fuel burning) airborne delays with ground delays. In a GS, a temporary restriction is imposed on the departure or arrival of aircraft at a specific airport or within a designated airspace. Although other TMIs (e.g., miles-in-trail) are also implemented as part of (air) traffic flow management in the NAS, the focus of this work is on GDPs and GSs. Since TMIs, by design, lead to flight delays or cancellations, it is crucial to put in place the right set of parameters(e.g., scope and duration of the GDP). For example, when the end time of a GDP extends beyond what is necessary, it imposes unnecessary delays on departing flights. This situation could occur as a result of inaccurate prediction of the(required) duration of the GDP based on the weather forecast. On the other hand, if a GDP ends prematurely before the underlying capacity constraints are resolved at the destination airport, it may result in airborne holding. The delicate balance lies in matching the termination of the GDP precisely with the resolution of capacity constraints, avoiding both the imposition of unnecessary ground delays and the need for airborne holding due to premature program termination.Failing to specify the right parameters for TMIs also leads to flight delays, creating a significant obstacle in managing the increasing traffic volumes causing increased work load for the controllers. To address this issue, we propose the integration of Machine Learning (ML) models in the traffic flow management(TFM) pipeline. In current operations, decisions are made by human experts based on extensive training, historical patterns, available traffic and weather data. Since we have an abundance of data from past events that tell us the likely impact of various TMIs, by ingesting historical data, properly trained ML models can offer valuable insights and aid human decision-making. With the FAA increasingly exploring advanced analytics, ML emerges as a focal point for enhancing TFM within the National Airspace System (NAS). As a first step, this study aims to provide traffic controllers with decision-making support for the issuance and adjustment of TMIs. Data analytics and machine learning have been previously employed to address some of the challenges associated with TMIs. Numerous studies have concentrated on various facets of TMI issuance, exploring factors influencing TMI parameters, including arrival rate, airport capacity, and delay prediction. For example, using weather forecasts, several statistical methods were used to produce probabilistic capacity profiles which in conjunction with deterministic models provided insights into the GDP planning process [1–4]. The downside of using deterministic models is that they rely on fixed inputs and predetermined rules, which lack the ability to account for the inherent uncertainty and variability present in real-world scenarios. In a separate series of studies, researchers aimed to predict the occurrences of GDPs and GSs. The majority of these studies utilized various supervised learning methods, including Decision Trees, Naive Bayes, Support VectorMachines, and Random Forests to analyze the influence of weather conditions and arrival demand on TMI incidents[5–8]. However, these studies primarily focused on predicting the incidence of TMIs without explicitly addressing the scope of TMIs, including their duration and their geographical coverage. Furthermore, the emphasis of these studies was largely on GDPs, given their higher frequency and longer duration when compared to GSs. A limited number of studies focused on predicting the parameters of TMIs, specifically addressing their duration and extent. In one such study focusing on optimizing the TMI parameters at San Francisco International Airport (SFO),the authors utilized a probabilistic forecast of fog [9]. They simulated various capacity scenarios based on the (fog)burn-off forecasts, selecting GDP parameters that minimized airborne and overall ground delays. However, this approach exclusively emphasizes stratus (fog) burn-off as the primary determinant of GDP and GS, neglecting other influential factors like severe weather events, runway closures, lower capacity than traffic demand, and other important variables. Given the complexity of predicting the TMI and determining its scope, we seek a more holistic approach. We aim to consider all significant factors that could impact TMIs and their parameters. What sets this research apart is the fusion of all data sources relevant to the issuance and adjustment of TMIs and it represents the first comprehensive attempt to optimize TMIs in this manner. Since this comprehensive solution involves various aspects, we break down the problem into smaller components and input all parameters into a unified model called the “TMI Adjuster”. Figure 1 shows the overall framework and the list of datasets used in each model. The objective of the TMI Adjuster module is to deliver reliable, consistent and expedited recommendations for the progression, adjustment, and termination of TMIs. The ML solution entails developing a pipeline capable of predicting the necessity of a TMI (e.g., GS or GDP) along with its various parameters. For example, in the case of a GS, this includes the scope of the GS either in terms of distance from the destination airport or based on pre-defined airspace sectors. Here, scope refers to those regions and departing airports that are subject to the GS. In this paper, we concentrate on the issuance of GSs in the three major airports in the New York area — LaGuardia(LGA), John F. Kennedy International (JFK), and Newark Liberty International (EWR). We fuse traffic, weather and other relevant aviation data from years 2017 to 2019 to train and validate the ML models. In particular, we use the following datasets: •Terminal Aerodrome Forecast (TAF): meteorological forecasts specific to each airport, issued four times a day, covering predefined time periods. •TMI data: includes all GSs and GDPs along with their respective parameters. •Aviation System Performance Metrics (ASPM): includes traffic related data such as aircraft delays, arrival, and departure rates. •Notices to Airmen (NOTAMs): utilized to extract runway closure data and manage interdependencies between terminals in close proximity. •Flight cancellation data •Airspace Flow Programs (AFP): includes information on flight airborne holdings caused by TMIs. The data preprocessing entails transforming ASPM, TMI, AFP, NOTAMs, and weather data into an hourly format and consolidating all datasets by merging them based on date and time as the primary key. The TMI Adjuster framework comprises two parallel models: one dedicated to GS and a second model focused on GDP. As previously mentioned, our specific focus is on the GS model as a multi-classification problem. In this framework, each data point of the GS model input summarizes ten hours of data. Specifically, the data loader for the GS model generates the input and output of the model as follows: at a given time step, the input includes the actual traffic, weather, and TMI data from the two-hour window before the time step, alongside the weather forecast and scheduled traffic for the next 8 hours starting from the time step. Based on this information, the output of the GS model for each time interval consists of three dimensions. The first dimension represents a binary decision on whether there should be a GS in place for the next hour or not. The second dimension is related to the scope of the GS in the United States, and the third dimension is related to the scope of the GS in Canada (i.e., to determine if the GS impacts airports in Canada).One of the challenges with TMI modeling is the sparsity of TMI events, particularly regarding its scope. To address this challenge in the scope of the GS model output, we implement grouping. The GS scope for the US region is defined based on a list of centers that should be included when the GS is in place. With 20 centers in the US, we utilized historical data to group them into 4 categories. In particular, we summarized our historical data in a graph format where nodes represent centers, and link weights are defined based on the co-occurrence of centers in the scope parameter ofTMIs. By identified strongly connected components in this graph, we were able to partition the centers into four groups. We consider two model structures for the GS Model. Firstly, a hierarchical classification model [10], where the human decision-making for a GS is of hierarchical nature. The decision-maker first decides whether there is a need fora GS, and if the answer is yes, determines the scope. A hierarchical classification model organizes the problem into a class hierarchy, typically a tree or a Directed Acyclic Graph (DAG) structure, and considers the dependency of the decision in the previous step to the next component [10]. Here, we employ the local classifier per level approach, which involves training one multi-class classifier for each level of the class hierarchy. The second structure is the independent structure. In this setting, as the name suggests, we do not consider the dependency of the decisions in the different dimensions of the output of the model. Instead, for each dimension, we train a multi-class classifier independently. Table 1 summarizes GS model statistics for training, validation and testing. The table documents the effect of limiting data to the time steps when there was actually a TMI in place or when a TMI had just terminated. This resulted in a more balanced distribution of the GS class(GS positive class)versus “No GS”(GS negative class), which might help the training process. While JFK and LGA follow very similar distributions, with 40% and 42% GS positive class respectively, EWR has proportionally fewer GS incidents at 28%. Our subsequent phase involves evaluating the performance of both hierarchical structure and independent structure using different state-of-the-art multi-class classifier models such as Random Forest, Decision Trees, K-nearest Neighbors, and Logistic Regression and forecast the duration and scope of the GSs.

Farzan Masrour Shalmani↗