Here is an overview of the conference program:
Abstracts in alphabetical order of author. Maintenance presentation abstracts behind Operations.
Panel Discussion
Cleared for Take-Off: The Advent of Electronic Vertical Takeoff (eVTOLs)
Michael Clarke (moderator), John Paul Clarke, Jon Peterson, CAE, UT Austin / Cayman Airways, Archer
Operations PresentationsAs we inch closer to the certification and operational deployment of this new mode of transportation, the global airline industry has started to pay more attention to the potential opportunities of eVTOLs. Many airlines see them as an opportunity to further advance their engagement with their customers by providing passengers a complete experience door to door. Some of the world leading airlines have placed significant orders for these new vehicles, and some of the leading eVTOL OEMs have themselves established an airline to operate these new vehicles on behalf of their airline partners and/or major helicopter operators. While most of the attention until now has been focused on the commercial aspect of operations, the time is prime to start working through the necessary operational process and procedures for seamlessly integrating these new high frequency operations into the existing airline operations network. As part of this panel discussions, we will present an overview of the eVTOL industry including market readiness and learn from the perspective of leading experts in the field.
What’s Inside the Box? Lessons Learned to Get Users Familiar With a Disruption Management Solution
Dennis Adelhütte, M2P
Introducing a disruption management solution to operation controllers has many parallels to playing the famous mystery game "What's inside the box?" with kids. This presentation will shed some light on the familiarization process and explore the journey of users as they navigate the complexities of a novel solution covering key lessons gained from user interactions including challenges such as the black-box nature of optimization as well as the significance of user interface design.
Putting Data At the Center of Our Operations - Bringing Holistic Analytics & Decision Capabilities Into the Cathay IOC
Amir Bennegadi, Marc Pilmann, Derek Chau, Cathay Pacific Airways
Spearheading the transformation of Cathay Pacific ops with a new data-driven and customer centric approach to ops control. The re-invented team within the Integrated Operations Centre (IOC) leverages an extensive Ops Decision Model (ODM) as a basis for single-currency optimization, accurately mirroring business policies across 8 functional areas (e.g. Cargo, Fuel, Revenue, Customer, Maintenance), to enable holistic, consistent, and timely decision making.
Introducing the Next Normal: Tomorrow's Safety Is Topped by Cyber- and Information Security for Safety
Yann Berger, Gesine Varfis, Airbus
Anticipated Dispatching for Alternative Airport MobilityOctober 2025 / February 2026 EASA regulation - PART-IS - becomes mandatory for all EASA approved operators and organizations with significant impact. This new standard takes into account the new normal: IoT, digitalization, AI and connectivity. A new safety first stage targeting cyber and information security for safety.
We are sharing what is at stake, what has to be addressed and what is the impact on Operations and Maintenance.
Simulated Annealing to Solve the Integrated Airline Fleet and Crew Recovery ProblemAirport mobility impacts operational performance. Thanks to anonymous passenger tracking, Groupe ADP has developed traveller prediction models for the CDG automated rail link and taxi management. Though transportation modes are different, the core prediction problems are similar and feed dispatching applications. Results have proven successful in one case and not in the other. In this paper we go through business problems, prediction models and lessons learned from these different outcomes.
Philip de Bruin, Utrecht University
Airline operations are prone to delays and disruptions, in which case the resource or flight schedules need to be adjusted. Such adjustments need to resolve a disruption while minimizing costs and passenger impact. We look at resolving disruptions in both the aircraft and crew schedules, where we take an integrated approach. For this, we develop a fast local search algorithm that resolves these disruptions in a cost-efficient manner. This is done in collaboration with KLM Royal Dutch Airlines.
A modified MIP model for aircraft tail assignment problem
Arund Chand, IBS
The aim is to create a model for aircraft tail assignment, satisfying all the constraints. Several authors in the past have done work in the area of tail assignment optimisation which is the assignment of individual trips to aircraft. Our aim is a modified model which handles additional constraints.
In this model, we are proposing a Mixed Integer Programming (MIP) model for tail assignment optimisation intending to minimise costs. The costs considered include trip costs as well as other penalty costs. This multi-objective problem is solved as a single objective by using the weighted sum method here. The model is formulated in OPL and solved using IBM ILOG CPLEX software.
The results obtained for real-world problem instances using this model have been very promising and the run time observed have been very small. Thus the proposed model can be expected to be used in real-world problems effectively. As future work, airline disruption scenario is discussed and the changes proposed for handling it has also been mentioned.
Improving Disruption Optimization Through a Better Integration of Passenger and Aircraft Recovery
Thierry Delahaye, Amadeus
A Deep Reinforcement Learning approach for the Container Loading Problem in Air TransportWe explore the integration of passenger recovery optimization into the fleet optimization, leveraging real-time data and decision making algorithms, integrating aspects of the passenger optimization into fleet decisions, and producing a set of integrated recovery alternatives.
We expect our work to enhance the quality of recovery solutions, especially passenger experience.
In the future, we expect to refine our model with data, exploring their application to other aspects of disruptions.
This study applies a variant of the Container Loading Problem with stability constraints to air transport, a sector heavily relying on this process. We build upon an existing Deep Reinforcement Learning model for 3DBPP, which considered one infinite-height container and ignored stability. Our Python-based DRL approach was tested on many artificial instances of different sizes with identical containers and appears to quickly provide highly packed solutions for realistic large-scale instances.Learning From Other Industries – Leveraging Reinforcement Learning for Improved Dispatching In OperationsLearning From Other Industries – Leveraging Reinforcement Learning for Improved Dispatching In OperationsLearning From Other Industries – Leveraging Reinforcement Learning for Improved Dispatching In Operations
Bagroom Operation Scheduling System: Reformulating Mature Optimization Solutions
Jesse Fowers, Sajia Afrin, American Airlines
A reformulation of a mature schedule optimizer built in-house at American Airlines that assigns which bagroom pier/make-up unit is to collect bags for each outbound flight. We incorporated two machine learning models that provided key stochastic inputs and ensure that each crew delivering bags planeside is assigned a set of flights that are within their capacity to reliably sort and pull from the bag room. We also show a tool that demonstrates the operational improvement to station leadership.
Alert to Action – Recovery Decisions Revealed
Tomas Gustafsson, Boeing
Costing of Irregular Operations Events (Currently an Unsolved Problem)In the Airline OCC, Ops Controllers and Crew Trackers spend the days managing disruptions, which on the detailed level manifest themselves as for example not enough ground time, exceeding Max FDP or crew with wrong qualifications. We were interested in better quantifying these issues for general understanding, but even more so to see if there are patterns in what recovery actions are applied. Could we even predict these actions to a reasonable fidelity? In this presentation we’ll share results from a few airlines, and discuss future practical applications.
A major reason why Irregular Operations (IROPS) events have been so devastating to airlines is that their true costs are very difficult to measure. Aircraft, pilots, cabin attendants, and passengers that are together for one flight might all go in different directions for the next, which means that a delay in one flight could lead to multiple downline delays. This cascading effect should be considered part of the cost and included in the calculation of the best go-forward plan.
Most airlines can track the cost of tangible IROPS expenses such as hotel and meal vouchers, customer refunds (for cancelled trips), or rebooking. However, they do not track what may be the dominant cost: the ability of frequent fliers to transfer some portion of their business to the competition. This is the major negative long-term effect of IROPS on an airline’s customers, and it will result in a much more significant cost of market share shifts over time.
Airlines should focus on defining a broader set of cost components and working across department lines to define appropriate metrics, which will drive improved responsiveness in IROPS situations and ultimately lower costs.
Large-Scale Airline Recovery Using Mixed-Integer Optimization and Supervised Learning
We combine mixed-integer optimization and supervised machine learning techniques to find better solutions to large-scale integrated airline recovery problems than those found with other exact and heuristic approaches. We reduce the solution space for a given disruption by adding constraints based on the patterns discovered in the solutions to historical disruptions. Our experiments on networks with >2,500 daily flights generate solutions of significantly higher quality than benchmark methods.
Achieving Sky-High Punctuality: Data-Driven Strategies for Qatar Airways' On-Time Success in 2023
Guy Shipton, Svetlana Ozova, Qatar Airways
Exploring Qatar Airways' 2023 OTP success through data-driven strategies, this presentation highlights the 'Q Not Delay' initiative and real-time monitoring tactics. We'll discuss goal setting and analytics' role in enhancing punctuality and operational efficiency.
Air Cargo Ground Service Vehicles Coordination: A Responsive Algorithm and Some Numerical Evidence
Jenny Tonka, Universite de Liege
Since suboptimal ground operations are a major cause of flight delays, we study the complex problem of coordinating air cargo ground service vehicles:
Descent Simulations for More Fuel-Efficient Arrival Procedures-We model it as a set of rich VRPs linked by multiple synchronization constraints.
-We propose an aircraft-centric heuristic that uses recursive procedures to optimize service vehicle paths.
-We prove its efficiency through tests on a real use case and numerical evidence.
-We explore how it can contribute to the digital transformation of airports.
Learning From Other Industries – Leveraging Reinforcement Learning for Improved Dispatching In OperationsArrival procedures affect airlines operating costs, but fuel-efficiency teams lack tools to precisely identify and quantify their effects at scale. Using historic QAR data, we train ML models that represent current aircraft performances to simulate descent profiles, considering their lateral paths, speed limits, vertical constraints and flight modes. Scenarios can also be integrated inside flight contexts like actual segments or conditions, which enable collaboration between airlines and ATC.
Deep Reinforcement Learning Ensemble Method for Aircraft Recovery ProblemTo accommodate for disruptions in train operations, dispatchers monitor the network to take fast countermeasures. To support the dispatchers in decision-making, we applied a coupled reinforcement learning and real-time simulation approach that provides the dispatchers with optimal suggestions for actions – such as early turnarounds or priority changes. Airlines face similar problems in flight scheduling, IROPs or crew management and can leverage the methodological approach and lessons learned.
Efficient flight scheduling is crucial to properly allocate airline resources, but even the best flight schedule has to face unexpected delays and disruptions. Machine learning-based methods can identify suitable recovery methods as unexpected events occur. Reinforcement learning approaches are especially promising since they extract suitable solutions much more efficiently than conventional optimization and meta-heuristics methods and provide timely rescheduling capabilities for airlines.
Maintenance Presentations
Optimizing Air Canada's Operational Maintenance Schedule
Rahim Akhavan & Andres Robayo Romero, Air Canada
Federal and Safety regulations specify requirements to maintain different aircraft components on specific thresholds of miles, hours, or cycles. We present a sophisticated optimization system to generate an optimized maintenance schedule for the 0-10 days. It is optimizing the yield remaining on routine/non-routine maintenance tasks while minimizing the number of work orders. It incorporates an extensive list of constraints including hangar, staff, parts, and ground time availabilities.
Proactive Aircraft Engine Maintenance: Predictive Modeling for Fuel Nozzle Replacement and Scheduling Optimization
Nahid Parvez Farazi, Shuang Ling, United Airlines
The study aims to proactively schedule aircraft engines for fuel nozzle replacement before reaching a failure phase, enhancing operational reliability and minimizing disruptions. A statistical thresholding based predictive model is proposed leveraging the engine sensor data from a fleet of almost 200 aircrafts for the early detection of fuel nozzle coking. A multicriteria optimization ranking is proposed to prioritize engines for component replacement considering the replacement capacity.
Statistical Prediction Model for Aircraft Reliability Repairable System Based on Cost Indicators
Adel Ghobbar, SORT Engineering
Heavy Check Demand Forecasting: Using Algorithms to Streamline Long-Term Strategic Decision MakingReliability analysis with dynamic failure rates is a revolutionary idea for MRO organizations. Reliability information is conventionally derived from exponential distribution with (MTBF). Weibull distribution with dynamic failure rates in this research replaces exponential distribution with constant failure rates, which provide a more robust and vigorous reliability analysis, a reliability-monitoring model is created, which helps to analyze component reliability in a cost-effective way.
Justin Lee, United Airlines
With over 800 new aircraft on order, United faces unique challenges when optimizing resources to complete heavy maintenance. To replace a time-intensive, manual solution, the TechOps Data Analytics team built a greedy algorithm for forecasting long-term heavy check demand that allows for flexibility in fleet plan changes, age-dependent spans, maintenance program changes, and other features. This approach provides accuracy and agility while informing multi-million-dollar strategic decisions.
Better Prognostic Performance Metrics
Shashvat Prakash, Collins Aerospace
While prognostic health management practices are increasingly viable, there remains a gap between their theoretical validation and their practical performance. Metrics like model precision or remaining life estimation are not as important as the ability to minimize lost time while maximizing detection with as many supporting examples as possible. Operational effectiveness has two objectives: time-correlation against remaining life and cost justification for lost remaining life.
Analytics for Servicing Innovation: Integrating Technician Expertise and Data Analytics for Enhanced Tire Condition Monitoring
Michael Quann, United Airlines
Unscheduled tire removals frequently cause flight delays. While worn tires are typically caught at service checks, several factors lead to a high rate of worn tires being flagged at the gate by pilots. We present the newly developed Tire Condition Monitoring (TCM) tool, which addresses this challenge in two ways: technician-driven proactive identification of near-to-worn tires, and data-driven flagging of tires for removal with a predictive tire wear model.
An Engine Shop-visit Optimization Model for Cost Reduction and Efficiency Improvement
Alejandro Vicente Amigo, Mohammad Kouhi, Vueling
Leveraging Enhanced Maintenance Logbook Information to Gain Heightened Insights into Line MaintenanceIn Vueling, the engine shop visit planning is a manual process prone to errors and inefficiencies. To improve this, a digital tool is proposed for the power plant team. Using a LP optimization model, the tool minimizes costs, considering LLP, performance restoration, and End-Of-Lease compensation costs. Key constraints include LLP life limits, redelivery conditions, spare engine availability, hangar slots, and engine drop instances. This model promises enhanced planning efficiency and cost.
Maintenance logbooks have long been used for detailing maintenance history. Leveraging electronic logbooks and data science, we have enhanced our understanding of line maintenance and ability to identify key issues. With Logbook Activity Allocation Model, a maintenance item’s life cycle and moving path can be clearly identified and allocated. This data fuels research on improving maintenance success. Integration with ML models further enables problem-solving capabilities previously inaccessible.