The technical program is now complete and we are no longer accepting abstracts.
Here is an overview of the conference program:
Abstracts in alphabetical order of author. Maintenance presentation abstracts behind Operations.
Understanding the U.S. Pilot Shortage through System Dynamic Modeling
Pablo Guillermo Aguirre, AV8 Analytics LLC
The current shortage of pilots in the United States poses a significant concern for the airline industry. The FAA regulatory decisions of the 1500-hour rule and the age cap for pilots flying under Part 121 have been the leading cause of this shortage. However, no economic system operates indefinitely in an exponential direction without a countervailing force that restores it to equilibrium. Therefore, system dynamics modeling can be employed to demonstrate how balancing feedback loops will eventually bring the pilot crisis back to equilibrium. Specifically, these feedback loops consist of an increase in the number of younger pilots in the overall pilot pool due to generational changes over the next decade, and an increase in overall pilot wages offered by airlines to attract more student pilots to become professional pilots. This presentation demonstrates that the policy implementation caused a deviation from equilibrium of pilot supply and demand, providing a quantifiable estimate of the time required for the system to return to equilibrium.
A simulation case study to investigate operational and financial viabilities of super-tugs for Southwest Airlines
Massoud Bazargan, Petar Grigorov, Khilna Rawal, Embry - Riddle Aeronautical University
Fuel costs are typical among the top cost driving metrics for airlines, and are subject to major fluctuation. Airlines need to enhance their operations by adopting more efficient and cost saving initiatives. Currently, many airlines use super-tugs as one of these initiatives, to save cost on aircraft burning taxi fuel.
This study relates to conducting a financial and operational feasibility of super-tug+H1s for Southwest Airlines. Southwest Airlines operates a weak hub and spoke network structure primarily from their Dallas Love Field (DAL) airport. The airline conducts many of light and heavy maintenance checks of their aircraft at the hangar in this hub. We adopted a simulation modeling to study the impact of deploying super-tugs on financial and operational performances of the airline.
This study investigates the utilization and operational metrics of acquiring multiple super-tugs. These metrics highlight how and when each super-tug is being utilized for taxiing the aircraft to and from the terminal to the maintenance hangar.
Recommendations on number of super-tugs with their financial and operational justifications are presented. The results are very encouraging and provide the airlines with the opportunity to evaluate the cost and benefits of purchasing and deploying super-tugs.
Cross-functional Manpower Planning - overcoming system and organizational silos
Franziska Burmester, WePlan
Ops Planning, and in particular Workforce Planning at Airlines has undergone a significant change. Former planning-timespans no longer exist and are being shrinked together, hence the different horizontal planning teams are to work closer together. The interdenpendencies of ops departments with each other has also significantly increased leading to workforce planning having to collaborate across the organisational silos much more than pre-pandemic.
Recently we have experienced in RFIs/RFPs a change towards a more integrated planning across the airline. This requires high system flexibility and a modular approach to cope with different forecasting algorithms. We will share our vision of a solution which vertically and horizontally integrated and outline the benefits this approach will bring to airlines.
The Future of Flight Operations Powered by Optimization and Emerging Technologies
Michael Clarke, CAE Civil Flight Services
Have you ever wondered what airline operations will look like in 2025? What will change, and what will stay the same? What fundamental changes in our approach are required to support airline operations? How will new aircraft types and alternative modes of transportation change how airlines manage their operations? As airlines fully recover from the impact of the global pandemic, they continue to face some of the same challenges, including network flow constraints, limited resource availability, and the inability to manage and expand their operations effectively. We foresee future airline operations relying on increased automation for resource allocation and disruption management. This level of automation will require advanced system scalability and the ability to support autonomous decision-making. In this presentation, we will explore emerging trends in the airline industry and identify how emerging technologies can help shape the future of airline operations.
Amadeus Operations Control - Robust planning and Operations
Thierry Delahaye, Will Turner, Amadeus IT Pacific, Qantas
The Amadeus Operations Control (OC) planning optimization is solving large operational problems every day, now in production. The combination of an exact optimization focusing on finding an optimal (feasible) solution to a large allocation problem, and a metaheuristic focusing on robustness enables finding robust solutions taking into account several aspects such as crew and maintenance.
With the core optimization logic planning in place, we are now focusing on building an integrated, reworked model for operations, where existing considerations of aircraft allocations and crew are joined by operational decisions such as delays, cancellations and ferry flights.
In this presentation, we analyze the impact of the robustness focus of the planning optimizer on the operational window, as well as our iterative development model for the operational optimization, prioritizing features, and building and testing the algorithm on a large range of representative instances in close collaboration with users, with a core focus on producing solutions which are publishable in real life.
A Long-Term Approach to Irregular Operations Recovery
Ira Gershkoff, T2RL
Recovery from waves of long delays and flight cancellations (known in the industry as “Irregular Operations” or “IROPS”) has always been a problem for airlines, but recent events seem to have been more severe than ever before. Improving performance in the recovery fundamentally requires aligning aircraft and crew flows to re-accommodate passengers as quickly and reliably as possible. Existing tools do a good job of optimizing the aircraft, crew, or passenger flows independently but cannot look at all of them together. Even if they could, the dynamic nature of IROPS would likely render any composite solution to be non-viable just a few minutes after it was implemented.
We will present an architectural approach to developing a more robust environment for managing these events. It represents a strategy that will take some time to fully implement, but will produce significant incremental improvements along the way in terms of:
• Managing cancellations and delays to maximize accommodation of disrupted passengers
• Maintaining utilization of flight crews, while minimizing reserve usage
• Realigning aircraft flows to return to normal operations once the IROPS event is over
• Improved training and certification of Operations Controllers and Crew Schedulers
The net result of these enhancements will be better performance in IROPS events in terms of aircraft and crew utilization, customer accommodation, and airline profitability.
Applying Modern Technology to the Turnaround Process
Nancy Dolores Hernandez, Francisco Javier España, Luis Eduardo Sanchez, AeroMexico
Leveraging AI to Improve Operational Efficiency and Stability in Airline Operations
Aeromexico will present why a new technological solution was needed to provide information, control, and analytics for the flight turnaround process. An airline’s below-the-wing processes are an area where having proper monitoring, data gathering and more effective control can be used to better refine the operation in order to reduce delays and increase operational efficiency. The different elements of the solution they implemented in conjunction with Boeing are reviewed with a focus on illustrating the potential to improve the airline’s operations, reduce delays, and reliably reduce turn times, in conjunction with the analytics that naturally can be derived from their solution.
Flight operations controllers face complex decisions every day, requiring quick analysis of multiple sources of information to minimize the impact of schedule disruptions. In response to this challenge, we have developed aiOCC, an AI-powered system that translates information from multiple sources into actionable recommendations. aiOCC monitors events around aircraft, rotation, passengers and crews to proactively identify potential delay risks and suggest adaptations to the schedule. By leveraging AI, we aim to improve operational efficiency and stability, ultimately leading to higher on-time performance. In this presentation, we will discuss the business case for aiOCC and provide high-level technical insights into its development and implementation, including the challenges we faced and overcame.The Impact of COVID-19 on U.S. Domestic Market Air Freight Carriers Efficiency With Regards to Freight and Fuel Burn
Jack Troutt, Nicholas Riggs, Utah Valley University
The COVID-19 pandemic has had a significant impact on the global economy, but few industries were fundamentally impacted more than aviation markets. As a result of the pandemic, air cargo freight saw significant growth as governments, economies, and societies relied on air freight to keep the world moving forward as ports and borders were closing around the globe. Utilizing data from the U.S. Department of Transportation’s Bureau of Transportation Statistics, a study was undertaken to analyze the impact that COVID-19 had on two major cargo operators based in the U.S. Focus was given to efficiency of operations vs. freight carried, with efficiency measured as a product of fuel burn and total cargo carried. The study investigated operations from 2019 to 2022, which includes the period during the height of the pandemic and into the current ‘recovery’ years. Results from the study is helping to create a robust picture on the changes air cargo carriers made as a result to the 8-11% increase in air freight carried during 2020. Based on these results, steps can be taken to reduce the effects of future market disruptions on air freight operations.
Transforming Airline Irregular Operations Management with Operations Research
Billy Wang, American Airlines
Since the 50’s, Operations Research, as a technique, has been a stable driving force that enables the airline industry to automate the decision making process, optimize efficiency and operate at scale. As the airlines keep growing their schedules and face new challenges such as resource shortage, Operations Research is playing an increasingly critical role in the airline operations. When it comes to Irregular Operations (IROPs), from ATC management, to rea-time rescheduling, to crew and aircraft recovery, Operations Research models are at the front and center providing the much needed power of automation and intelligence to the Operations Control Center teams to manage these challenging IROPs in a most optimal way possible. In this talk, we will examine closely some of specific challenges the Airlines face when dealing with IROPs, followed with an explanation on why OR models are so critical to the management of these events. We will conclude the talk by going over a couple of advanced OR models deployed at American Airlines that transformed the way how the airline manages IROPs.
Piece by Piece – Avoiding Baggage-Driven Disruption
Jörn Weerts & Niels Knak, Deutsche Lufthansa AG
Last year revealed how crucial baggage processes are for the smooth operation of any airline. Halls filled with piled up, left behind suitcases became the symbol of our industry. At Lufthansa we launched, relaunched or intensified many measures to mitigate the risk of baggage-driven disruptions. We focus on supporting with data driven approaches. We chose three examples to share.
As adding new features is an ongoing process this positive impact will even grow in the future.
Improved Decision Making for Disruptive Weather Events using Ensemble Forecasts
John Williams, Mike Robinson & Tim Niznik, IBM, MITRE & American Airlines
Decisions made in the absence of probability-based analytics will not just usually be wrong, they will consistently be sub-optimal, especially when those decisions involve weather. An increasingly important need for both airline operation centers (AOC’s) and air traffic management (ATM) is for improved response to weather events, specifically in terms of quantifying the level of risk around potential courses of action. What is needed is a practical, data-driven process that explicitly considers weather uncertainty in balance with the costs and risks for decision options. Such a capability would yield more consistent and efficient ATM initiatives and AOC responses that reduce avoidable delay, flight cancellations, costs, and passenger disruption. This talk brings together the perspectives of The MITRE Corporation, The Weather Company, and American Airlines who will present a framework to “bridge the gap” from deterministic to probabilistic decision-making using ensemble weather forecasts and multiple “trajectories” of possible weather constraint scenarios.
A deep neural network for the dynamic pathfinding problem applied to the optimization of aircraft trajectories
Dominik Zurek, CAE
Flight planning is a challenging problem that airlines face every day, constituting an integral part of their operations. At its core, it involves solving a form of dynamic shortest path problem, in which it is required to determine a route in a graph of airspace connections. The cost of the selected route depends heavily on weather conditions (with wind being the predominant factor), the performance capabilities of the aircraft (with the weight of fuel and its burn rate changing significantly throughout the entire route), as well as on other factors, such as payload weight and overflight fees. The leading approaches are currently based on adapting the A* algorithm to the dynamic time-dependent graphs. Consequently, the main challenge of the optimal usage of the A*-based algorithm is the choice of a heuristic function, which would evaluate the remaining cost of the flight from any given vertex in a graph. In the considered setting, the heuristic functions depend on estimating the time of reaching a given airspace way point (graph vertex). However, with the need to factor in continuously changing weather conditions and the state of the aircraft itself, this estimation becomes a complex task. To address this issue, we propose the usage of deep neural network to improve heuristics. The neural network is trained based on historical data which includes aircraft parameters and weather conditions. Due to the specification of this data we propose the concept of combining fully connected, convolution and recurrent neural networks. Experimental results show that our proposed solution is a very promising direction for developing efficient heuristics for trajectory optimization problems.
Data-Driven Maintenance Prioritization at United: Maintenance Priority Score
Greg Jackson, United Airlines
With the resurgence of customer demand, pandemic-era personnel retirements, and additions of new aircraft, Maintenance Planning must ensure the optimal allocation of resources to protect the operation from maintenance-related interruptions. United’s Tech Ops Data Analytics team, in partnership with Aircraft Routing, Maintenance Planning and Execution, has created and deployed a data-driven Maintenance Priority Score and data pipeline from many data sources to produce one easily understood metric that empowers Maintenance Planning to easily identify the maintenance tasks that present the most risk to the operation, thereby allowing the opportunity to proactively address the issue. This shifts the paradigm from primarily prioritizing maintenance based on time remaining to a holistic risk-based approach that focuses on time remaining, probability of operational interruption, impact to the customer, and workload recoverability.
Strategic Maintenance Policy Simulator
Paul van Kessel, KLM
To shift from decisions based on personal expertise towards data driven decision-making, the Strategic Maintenance Policy Simulator has been developed to help answer questions regarding maintenance hours, locations, time in hanger, etc. It simulates the upcoming airline season with the goal of optimizing earning potential, including: minimizing maintenance ground time, maximizing potential revenue, maximizing operational reliability, and minimize delay & cancelations costs. This leads to a Pareto front with operational revenue and operational costs on either side. The Strategic Maintenance Policy Simulator is used to find the optimal operational point by enabling exploration of the effects of different maintenance policies. The simulator consists of a schedule optimizer based on MILP formulation combined with a delay prediction module, trained on historical data. Key Performance Indicators include: workforce utilization, unscheduled ground time, and ground time.
Eliminating manual coding of ATA codes using Natural Language Processing
Varun Khemani, Jose Ramirez-Hernandez, American Airlines
ATA codes are utilized for indexing aircraft systems and can also be used to observe trends in aircraft reliability as mechanics enter a free-text description of maintenance work and assign specific codes to such work. As thousands of records of maintenance work and corresponding ATA codes are generated daily, addressing inaccuracy in code assignment requires significant resources for manual re-coding. To automate the re-coding process, we take advantage of Natural Language Processing (NLP) techniques to build classification models able to link descriptions of maintenance work to the correct ATA codes. Pretrained NLP models are not a perfect match to the language used by aircraft mechanics to describe their maintenance work. Thus, we first adapt our models to the mechanics vocabulary before using pre-trained models for the classification process. As a result, our models can achieve above 90% accuracy in ATA code classification, thereby substantially reducing the need for manual recoding.
Augmenting Predictive Maintenance Models: Simulation Techniques to Improve Airline Decision Making
Paul Lowe, Emily Haider-Hawkinson, Boeing
Boeing Global Services is developing a simulation framework for Advance Fault Notification (AFN) with an associated prognostic for specific aircraft parts to enable reduction of airplane life-cycle costs and increase platform availability. Data-driven failure prognostics for commercial airplane parts and subsystems by identify signals in the data stream of sensor information that are associated with an upcoming fault. These early warnings can then be integrated into airline analytics products like the Boeing Airplane Health Management portal to enable the predictive maintenance.
The simulation goal is to identify at what threshold of the prognostic should a maintenance be performed to avoid a failure while accounting for false alarms/missed alarms, maintenance costs, MEL requirements, and maintenance opportunities. The simulation enables understanding of schedule risks, costs, aircraft availability, based on the what-if scenarios and demonstrates the utility of using simulation to better understand implementation scenarios for prognostic alerts.
Applying Artificial Intelligence to Reveal the Pathways and Lifecycles of Aircraft Components
Anne Marie Newman, zeroG GmbH / Lufthansa Group
Unlocking the pathways, usage patterns, and lifecycles of aircraft components as they experience multiple removals and installations throughout years of use presents challenges in making value-oriented process- and condition-based decisions. Maintenance, Repair & Overhaul (MRO) data sources typically originate from very different systems, services, and environments and are therefore highly heterogeneous, making data fusion and data analysis a central challenge. Without a fully transparent, representative, and connected history of component provenance, true value-oriented decisions are impossible to make. In this talk we present insights from a joint research initiative conducted by Lufthansa Technik and zeroG, Lufthansa’s data science subsidiary, which uses artificial intelligence and machine learning methods to reconstruct and make predictions about the pathways and lifecycles of aircraft components. We illustrate how incomplete data can be used in machine learning algorithms to reconstruct the likeliest real-world scenarios that are implied by the observed data. We present theoretical insights of the underlying models, as well as the benefits of deploying user-centric solutions to improve day-to-day decision making through front-end applications, thereby simplifying complex operational processes.
Prediction of Smoke, Odors, and Fumes Events Related to Auxiliary Power Units
Timothy Sham, Varun Khemani, Jose Ramirez-Hernandez, American Airlines
Smoke, Odors, and fumes (SOF) events related to the operation of Auxiliary Power Units (APUs) can cause a negative impact in the operation of commercial aircraft, which could become out of service and then generating expensive delays and cancellations. As part of our initiatives to mitigate such negative impact, the Power Plant Engineering group in collaboration with the Operations Research & Advanced Analytics group at American Airlines, have developed Machine Learning (ML) models that are utilized to predict in advance the occurrence of APU-related SOF events. Our models utilize key APU operational and performance data to generate daily predictions across one of our narrow-body fleets. Moreover, these predictions allow the ranking of the aircraft based on the risk of occurrence of an APU-related SOF event within a pre-specified period, which in turns, allows for better maintenance planning and a more pro-active actioning of corrective actions on identified APUs with high-risk of SOF events.
Streamlining Aircraft Engine Maintenance: Using Artificial Intelligence for Efficient and Accurate Inspection
Fabian Vogel, zeroG GmbH
Line Maintenance Team Scheduling
Maintenance of aircraft engines is critical to ensuring the safety of passengers and crew onboard and avoiding costly engine failures. Currently, engine maintenance is performed by highly skilled technicians who manually inspect each blade for defects and judge which defects are out of serviceable limits. This is a time-consuming and demanding task. Due to the large number of undamaged blades the risk of missing actual defects increases. Huge advancements in artificial intelligence and computer vision have made it possible to analyze hours of maintenance videos for defects in seconds. Our solution automatically pre-selects potential blade defects in real-time from borescope videos using deep learning segmentation models and can significantly reduce inspection time while maintaining high accuracy. This enables technicians to focus on the pre-selected potential blade defects, rather than examining each blade, thus freeing up their time, reducing nuisance effects, and increasing efficiency.
Every day, more than 200 flights arrive or depart from KLM’s main hub, Amsterdam. A group of sixty ground engineers keeps the fleet in an airworthy and healthy state, allowing the network to run smoothly. They are supported by a team of five planners, which solve the daily complex puzzle of matching the engineers’ teams to aircraft.
To support Line Maintenance in an increasingly dynamic and versatile operation, LiMiTS (Line Maintenance Team Scheduling) was developed. Originally, a planning used to be made by paper which results in the fact that operational feasibility was hard to oversee.
Moving from a paper driven towards a data driven environment results in two benefits:
1. Data overview: It provides a single source of truth regarding all the data they need to execute their activities:
a. Workforce availability
b. Maintenance scheduling status
c. Network scheduling status; where originally this data used to be displayed into separated legacy systems, they are now brought together into one coherent overview
2. Schedule optimization: LiMiTS consist of a schedule optimizer based on a MILP optimization model. It takes into account a hierarchical objective function with the aim of distributing workload over the teams and throughout the day. Optimal schedules are created whilst taking into account ground engineer certifications and the latest flight schedule.
An optimization approach to solve the hangar maintenance scheduling problem
Max Witteman, TUI Group (Tui Fly)
This presentation addresses the scheduling of aircraft maintenance checks that require hangar space. The allocation of hangar space to different aircraft is a heavily constrained combinatorial problem that needs to be solved frequently by our maintenance planners. We propose a novel two-stage framework for scheduling the hangar maintenance checks, delivering value for both short-term and long-term planning. The problem is formulated as a linear programming formulation and can solve real-life scenarios with a heterogeneous fleet of aircraft and various hangar facilities across the world. Our model aims to minimize the outsourcing of our maintenance considering end-of-lease conditions, allowed hangar configurations, fleet availability and man-hour limitations. Our approach is validated within all five airlines of the TUI group using a heterogeneous fleet of 140 aircraft. The presentation will feature a live demo that demonstrates the user experience enabled by our front-end interface, while also providing an explanation of the robust and scalable back-end processes that power our solution