The theme of this year's conference is "Efficiency and Resiliency amidst Economic and Political Uncertainty."  Efficiency and resiliency will (and should) always be primary objectives for airline operations, but the current level of economic and political uncertainties create additional challenges for and underscore the importance of both.

Here is an overview of the conference program:

• Opening presentation & host airline keynote speaker address
• Airline update
• Technical Operations presentations 
• Sponsor presentations
• Technical Maintenance presentations
• Non-sponsor vendor sales presentations
• Speaker awards & closing presentation

Abstracts in alphabetical order of author. Maintenance presentation abstracts behind Operations.

Airline Update

Vueling Airlines

Operations Presentations

Machine Learning Driven Decision Support System For Cargo Capacity Optimization in Passenger Aircraft Operations

Furkan Ayik, Egemen Çekiç, Ela Aslan; Turkish Technology, Turkish Airlines

Efficient capacity management in passenger aircraft operations is a time-critical challenge, where inaccurate capacity estimation causes inefficient loading, increased ground handling effort, last-minute adjustments, and avoidable offloads. This study presents a machine learning driven decision support system for capacity optimization that models pax baggage container demand evolution using time-aware regression and produces actionable pre-departure loading recommendations. Validated on large-scale airline operational data with time-consistent evaluation, the approach improves capacity utilization accuracy, enables earlier planning, reduces offloads thereby improves financial contributions.

What Airlines Should Want?

R. Michael Baiada; ATH Group, Inc.

"What Should Airlines Want?" (https://greenlandings.net/wp-content/uploads/2025/12/What-Should-Airlines-Want-MTS-WINTER-2026.pdf), asked the question, "How should airlines want to fly their aircraft". Airlines had no answer to my question in the 1990s, and still no answer today - Why?

ATC Control over the airline's aircraft movement locks out the airline’s “day of” business goals. Which flight lands first, second, third, fourth, etc. is critical to the success of an airline’s "day of" operation. GreenLandings® is independently validated by FAA, Embry-Riddle University, GE Aviation (saved 8.23 minutes/flt, 25,000 kg fuel/ day), Georgia Tech, Delta Air Lines and others.

The Cathay Journey to Integrated Operations Planning and Delivery: From Schedule Assessment Simulations to Super-Typhoon Recovery

Amir Bennegadi, Kelvin Tse, Paul Van Kessel; Cathay Pacific Airways

Cathay recently merged its Operations Planning department with the Integrated Operations Centre, forming a unified Integrated Operations team. This allows us to cohesively steer the planning and delivery of our operation from schedule preparation to Day of Ops, while accelerating the development of decision support capabilities across multiple domains and time horizons.

We will deep dive into two tools that are currently in use and under development. Strategically, scenario-based simulations help assess schedules and relevant trade-offs.

Operationally, advanced disruption recovery optimization tools enabled full recovery within 48 hours after Super Typhoon Ragasa in Sep 2025.

AI in Flight Ops

Harald Berg, Pedro Aloy Macedo; Lufthansa Consulting GmbH

Artificial Intelligence is becoming pivotal in Flight Operations, offering benefits like efficient flight paths, fuel savings, climate-aware routing, improved decisions, and better crew and asset use. However, AI also introduces new operational risks. This session explores AI's impact in Flight Ops through five practical use cases and an Integrated Operations Framework, showing where AI adds value and where risks arise. Topics include core AI applications, balancing opportunity and risk, and evidence of AI mistakes. Participants gain practical insight into AI’s role, risks, and responsible use in Flight Operations.

Enhancing Baggage Handling Efficiency at Charles de Gaulle Airport through Machine Learning-Based Forecasting Models

Florian Bertosio; Groupe ADP

Bag handling processing times are key to successful bag operations and good passenger experience. They are the prime factor of mishandled bags. Our study, carried at the hub of CDG airport, shows how the understanding of Bag Handling System (BHS) journey duration enables a clearer split of responsibilities between stakeholders and more efficient bag handling operations, with better resource allocation, and reduced congestion within the BHS. We then focus on the business improvements unlocked by bag BHS journey duration prediction, and present alternative prediction methods based on machine learning models, with an analysis of critical factors to reduce prediction errors. We finally show how BHS journey duration varies with BHS load and BHS health status.

Data-Driven Decision Space Tailoring for Airline Disruption Recovery

Thierry Delahaye; Amadeus

Airline disruption management requires numerous delay, swap, and cancellation decisions, leading to large mathematical models. Solving those can take significant time, which in turn causes challenges in the context of real-time disruptions when finding fast, high-quality solutions is critical.

We train machine learning models from a large-scale dataset of past disruptions and their resolutions to tailor possible decisions for each flight, leading to significantly decreased decision space and solving time, while preserving solution quality

THOR : Task Hub Operational Research

Alexandra Deniaud; Air France

Meet THOR, a tool developed to calculate the optimal number of each shift in a given week. To do this, it uses exact resolution through linear programming, but also employs approximate methods to provide balanced schedules with breathing room for AF staff. To predict the charge we'll have to cover, we can establish rules: a flight departing will trigger tasks at check-in, jet bridge..., and we can also estimate our workload at the airport: For example, we can assume that for a flight taking off at 9 a.m., 30% of passengers will already be there at 7 a.m. All these rules will enable us to create tasks that will be carried out by agents.

Operational Excellence Through Cross-Departmental Data Integration & Optimization

Panos Dimopoulos; Aegean Airlines

Ops360 is an IOCC-Ground OPS-Flight Crew-Management suite that assists in achieving operational and financial excellence unattainable with any standalone system. Operational data have been integrated from different sources to make the decisions easier and more accurate. It improves cross-departmental optimization, controls operational complexity and supports real-time decisions. It increases our situational awareness and boosts our adaptability to changes in daily operations.

Containing real-time processes, alarms, events and archived historical data across disparate business systems, Ops360 creates a single, common information point.

Airspace Flow Programs and Airline Mitigation Strategies

Ricardo Escalante Villalta; United Airlines

Airspace Flow Programs (AFPs) create substantial delays for U.S. and Canadian departures and are used to maintain safety during periods of high traffic or reduced airspace capacity. Airlines employ strategies to mitigate AFP delays, including routing flights around constrained airspaces, a commonly used but rarely measured approach. This presentation shows how Federal Aviation Administration (FAA) real-time data can quantify AFP operational impacts and airline delay mitigation outcomes. The framework enables post-event evaluation and data-driven operational discussions across days and regions through integrated analysis of AFP delays, mitigation strategies, and arrival performance.

Seamless Aircraft Reallocation with Swaptimiser

Javier Garcia, David Garcia Heredia; Ryanair

Ryanair has more than 600 aircraft distributed across 100 bases. Maintenance, disruptions and commercial decisions require constant reallocations in the most complex network in Europe.

Swaptimiser solves complex reallocation needs in minutes, considering operational robustness, crew restrictions and passenger satisfaction.

We will show how we achieved to move hours of complex human decisions through hundreds of aircraft into a streamlined decision support tool.

Beyond On-Time Performance: Exploring the Relationship Between Operational Trends and Airline Outcomes

Gavin Grochowski, Edward Stephens; Jeppesen Foreflight

The industry emphasizes On-Time Performance (OTP) as a competitive benchmark. However, OTP has flaws as a primary performance metric: it is influenced by uncontrollable factors, incentivizes schedule padding, and advantages carriers in favourable geographies.

We are exploring the development of composite metrics shown to drive positive outcomes, such as improved financial performance and customer satisfaction. Here, we present our statistical modelling to link these outcomes to a broad range of KPIs. Using market-level operational data combined with historical airline outcome measures, we evaluate if a combination of metrics explain operational performance more consistently than OTP alone.

A Major Way to Mitigate Diversions & Delays From Weather-Caused Hub Disruption

Michael Irrgang; Airline Operations Solutions, Inc.

A number of years ago a new approach to mitigate major hub disruptions during severe weather events was tested successfully and presented to the FAA. However, there was no technology at the time to implement it without heavy use of manpower. Today, with real-time and historical flight tracking data combined with use of Big Data and AI technology, it would be possible to develop a system capable of cutting unplanned diversions by >50% and delays by ~60%. How to develop such a system is presented here.

Data-Driven Optimization of Baggage Handling Systems

Rim Jabri; Groupe ADP

This study addresses the challenges of understanding and optimizing complex governance systems, focusing on Baggage Handling System (BHS).

We apply data analytics and econometric modeling to identify determinants of baggage mishandling, although quantifying the benefits remains a challenge.

The analysis covers three dimensions: end-to-end baggage processes, BHS flow dynamics, and maintenance strategies. We develop methods to prioritize baggage handling, optimize flows, and improve preventive maintenance policies to enhance system performance and reliability.

DispatcherTaskPlanner: An Ontology-Guided Planning Agent for Dispatcher Emergency Situation Handling

Nahyun Kim, Seonggin Lee, Keumjin Lee; Korea Aerospace University, Aeronity

The adoption of AI in aircraft operation control has been limited due to safety issues. To overcome these limitations, we propose a Dispatcher Task Planner (DTP) agent to support airline operation control teams in emergency situations. The DTP agent can generate action plans for emergency response procedures. Unlike most LLM-based AI Agents which generate probabilistic responses solely from pre-trained data, our agent generates responses by ontology-guided reasoning. It enables the agent to implement complex aviation-specific planning tasks.

NASA Center for Air Transportation Resilience (CATRes): Year 2 Update

Max Z. Li; University of Michigan

The Center for Air Transportation Resilience (CATRes), a NASA University Leadership Initiative, conducts data-driven research to strengthen the resilience of the air transportation system. This presentation provides a technical overview of ongoing CATRes research, including cluster analyses of historical disruptions, generative models for synthetic disruption training, and airline–air traffic collaborative optimization. Now in its second year, CATRes has made substantial technical progress; we will dive deep into results most relevant to airline operations and system resilience.

Validating Integrated Airline Recovery Decisions Using Time-Travel Benchmarking

Cristina Núñez, Marco Manzoni; Big Blue Analytics

Integrated airline recovery solvers aim to generate globally consistent decisions across aircraft, crew, and passenger domains during disruptions. While much effort is devoted to building such solvers, a fundamental OR challenge remains: validating that decisions are good in real, time-critical operations.

This presentation introduces a validation methodology based on time-travel benchmarking, in which historical disruption scenarios are reconstructed by rewinding the operational state to the exact moment a disruption occurred. This allows the solver to be evaluated under the same information set faced by the OCC, comparing solver decisions with actual OCC actions.

From On-Time Performance to P&L: Quantifying the True Cost of Airline Disruptions

Satya Sahu, John Szatkowski; Seabury Airline Strategy Group

Airlines track OTP and cancellations but often lack a finance-aligned way to value disruption. ASGcostIQ is a defect-cost framework that prices disruptions such as delays, cancellations, and diversions using context-sensitive, non-linear cost curves calibrated to the airline's own P&L and operational data. We show how it consumes schedule-simulation outputs and supports day-of-ops trade-offs (hold vs go, delay vs cancel) and investment ROI. An anonymized client airline operation example shows how modeled disruption profiles guided schedule, firewalling and spare decisions and quantified net benefit.

Optimized for Whom? The Hidden Conflict in Flight Scheduling

Jeanette Schmidt; M2P Consulting GmbH

Flight schedules are often optimized from a single functional perspective, although they serve multiple stakeholders with different objectives. Operations Control uses the schedule to stabilize the network and protect passenger flows. Maintenance relies on the same schedule to plan resources and ensure technical reliability. These perspectives lead to different definitions of optimality and frequently to conflicting decisions. This talk shows how isolated schedule changes may locally improve one area while shifting the problem to another, and argues for a holistic optimization approach based on shared goals, common constraints, and joint decision-making.

From Simulation to Optimization: Building Robust Airline Schedules at Scale

Narges Sereshti, Cecilia Chen; Air Canada

At Air Canada, we built a Routing Optimizer that turns disruption-aware insights into robust aircraft routings at network scale. Signals from machine learning models and a simulation engine—vulnerable rotations, tight turns, and high delay‑propagation risk—feed a time–space network optimization model that improves aircraft routings, optionally adjusting departure times, while enforcing maintenance and operational constraints. The mixed‑integer program minimizes expected delay, towing and gate pressure, and passenger misconnections, while increasing targeted “whitespace” where it matters most. The formulation is made tractable through graph-based constraint reduction and time decomposition.

A Multi-Criteria Optimization Approach for Reserve Crew Assignment in Day-of-Operations Disruption Recovery

Narendiran Sivanesan, Dr. Max Barkhausen, Dr. Andreas Andresen; tulanÄ

We address the problem of assigning reserve pilots to open trips during disruptions, balancing immediate coverage against future flexibility. We present a scoring-based optimization framework evaluating candidate assignments across standby utilization, roster fragmentation, regulatory buffers, and coverage preservation. The approach models crew availability as time-dependent paths subject to fatigue regulations and contractual constraints. The system has been deployed at a large European carrier, supporting daily operational decisions. Validation against historical data shows improved assignment consistency and reserve utilization under high-disruption conditions with scarce resources.

Human-Centric Optimization for High-Quality Large-Scale Schedule Recovery

Kenichi Tsutsui; All Nippon Airways

Managing large-scale disruptions is challenging due to complex constraints and reliance on manual expertise. We present a Decision Support System (DSS) that integrates Rule-Based Heuristics with Mathematical Optimization. By evaluating a comprehensive set of KPIs—from operational stability to minute passenger impacts—the system replicates the high-standard decision-making characteristic of Japanese airline operations with high repeatability. Users can prioritize objectives to generate four optimized scenarios. Results show that this approach reduces decision-making time by 70% during large-scale disruptions, significantly enhancing operational reliability and customer satisfaction.

Simulation Tool to Stress-Test Day-of-Ops Performance of Strategic Flight Schedules Based on Historical Delay Data

Dennis Ulbrich, Christian Kumpf, Andreas Hottenrott, Till Rasche, Bernhard von Mutius; Lufthansa Group

Flight schedule design is evolving from purely commercial planning to anticipating operational impact. Our tool, RobustOps, uses historic data and Monte-Carlo methods to simulate aircraft rotations, delays, and smart mitigations like aircraft swaps. It translates complex operations into clear metrics - punctuality, passenger disruption, and costs - allowing airlines to stress-test schedules and optimize reserves. Case studies at Lufthansa Group show high predictive accuracy and benefits enabled. Our simulation core is now being embedded into an integrated optimizer to build schedules that are both commercially attractive and operationally robust.

On-Time Departure Performance Prediction With Machine Learning Methods

Mehmet Yalcin; Turkish Technology

This study employs machine learning models to predict whether flights scheduled for the following day will depart on time. The main objective is to forecast flights' on-time performance in advance and identify which flights are likely to experience delays, enabling preventive interventions and improving overall punctuality. Tree-based models, particularly CatBoost, are used with 2023-2025 flight data. Data sources include temporal variables, historical on-time performance, flight scheduling information, maintenance intervals and weather conditions.

Maintenance Presentations

How NOT to do Data Science - Focus on Engineering & Maintenance / Supply Chain

Kieran Brady; TUI Airline

The application of Data Science in airlines often fail due to fundamental missteps in approach, execution or post-release. That’s why we will talk about common pitfalls that we have faced in the last years, such as real-world data being too messy to do an entirely data-driven decision-making process, the need for pragmatic strategies, the temptation to experiment with hyped technologies or solution-first thinking, and neglect of post-release maintenance.

Success requires asking the right questions, stepping away from the sunk cost fallacy, having engaged stakeholders, enforcing coding standards, and adopting flexible, business-driven approaches to deliver sustainable value.

Optimizing Aircraft Maintenance Planning at Ryanair

David Garcia Heredia; Ryanair

Aircraft Maintenance is a vital and complex part for any airline. Vital because aircraft cannot fly without the corresponding checks. Complex because it involves the coordination of a lot of people and steps (planning, validating, adjusting, executing).

In this presentation, we will show CheckMate: a tool that generates optimized maintenance plans for Ryanair. The plan respects any business rules specified by the user, and outputs, for each aircraft, in which hangar to do the check, when, and how to reach the hangar via swaps, so that the minimum operational network disruption is produced.

Predictive Tool for Anticipating Non-Routine Findings in Heavy Maintenance Checks

Alejandro Guemes; The Boeing Company

Non-routine findings during heavy maintenance checks (HMCs) can significantly increase the burden and costs for airlines, especially for aging fleets. This work introduces a predictive maintenance-planning tool that allows airlines to anticipate frequent non-routine findings, thereby optimizing resource allocation and reducing operational disruptions. By leveraging maintenance records submitted by Boeing operators from 1978 to the present, the proposed model applies machine-learning techniques to identify non-routine patterns and forecast upcoming HMCs. Preliminary results, obtained in collaboration with Boeing customers, show high confidence in detecting non-routine findings.

Data-Driven Maintenance Prioritization at United: An Updated Framework

Brian Ho; United Airlines

United's maintenance team faces unprecedented demands as its fleet grows rapidly and global maintenance footprint expands, requiring reinvented approaches to maintenance functions. Maintenance Planning must optimally allocate resources to prevent operational disruptions, but complex maintenance tasks make prioritization difficult. United previously presented a solution at AGIFORS, but since then has completely reworked the framework through Tech Ops collaboration. This presentation explains how the pipeline uses historical and forecasted data to produce risk scores across dimensions, combines them into a consumable output, and integrates it into planning and execution tools.

Cluster-Based Monitoring for Predictive Maintenance Alerts with Graph Embeddings

Shuang Ling; United Airlines

Predictive maintenance alerts must be monitored continuously as equipment behavior and operating conditions change. Because alerts span different assets and failure modes with distinct patterns and impacts, fleet-wide metrics can hide degradation and drift. We model relationships between alerts, assets, and defect context as a graph, learn graph embeddings, and cluster alerts into comparable groups. Performance and drift are benchmarked among alerts within the same cluster, enabling fair peer comparisons and clearer prioritization. Cluster-level monitoring identifies where retraining, threshold tuning, or feature updates will deliver the most operational value across heterogeneous fleets.

Simplified Technical English Validation

Melanie Lorang; The Boeing Company

Boeing created STEVE (Simplified Technical English Validator Engine), a Python ML/AI platform enforcing ASD STE100 for maintenance documents. STEVE combines rule based natural language processing, lexicon and syntactic checks, and document consistency to provide line level feedback, prescriptive corrections, API integration, telemetry for iterative refinement, and traceable decisions -- reducing rework, accelerating publishing, and improving compliance and clarity.

Smart Overtime Management with an Analytical Staffing Framework

Mihir Nevpurkar; American Airlines

The Tactical Staffing Support Tool uses a task‑based, data‑driven approach to improve maintenance staffing efficiency and manage overtime within airline operations. Traditional overtime methods relied on fixed seasonal thresholds that did not reflect daily workload. TSST analyzes real maintenance task activity to understand technician utilization and recommend staffing levels that match operational needs. This has reduced overtime headcount by more than 30 percent and supports consistent, efficient maintenance operations.

The Truth About Power-by-the-Hour l What Every Airline Maintenance OR Specialist Should Know

Jeff Oboy; PA

Power-by-the-hour (PBH) is one of the most recognised and adopted concepts in aircraft engine maintenance. For years PBH agreements have alleviated significant operational burdens traditionally faced by airline powerplant managers while finance teams have welcomed the benefits of improved risk and budget management. However, PBH has come to foster a degree of complacency and contributed to misconceptions about an airline's influence and decision-making authority. This talk aims to provide airline OR teams with a clearer understanding of PBH and such's implications for decisions about when, where, and how engines can be most optimally serviced.

Predictive Maintenance With Unstructured Operational Data: An Alternative to Sensor-Based Approaches

Faizan Patankar; Amygda

Airlines generate vast data in systems like Maximo and pilot reports, but struggle to identify patterns in these unstructured datasets. Traditional predictive maintenance relies on OEM sensor data, but data sharing limitations constrain ease of use.

We demonstrate an alternative, AI-led pattern recognition, that discovers hidden correlations across disparate log entries. Validated on aviation ground equipment, it identified 80% of at-risk assets 60+ minutes before deployment.

This works with existing CMMS and complements OEM systems. By monitoring asset-risk, it eliminates per-fault model development, offering airlines an accessible path to predictive capability despite sensor limitations.

Data-Driven Operational Risk Forecasting with Decision-Oriented Outputs

Timothy Sham; American Airlines

Unscheduled Aircraft Out-of-Service (AOS) events disrupt airline operations, driving delays, cancellations, and cost. To mitigate this impact, the Tech Ops Performance & Analytics team developed machine learning models to forecast AOS risk. The models use schedule-derived and operational inputs capturing operating patterns and system conditions, augmented with indicators of evolving maintenance demand. To support both near-term operations and long-range planning, the framework provides two horizons: a 90-day forecast using richer signals, and a one-year directional outlook based on fewer inputs. These forecasts translate risk into decision‑ready guidance to reduce operational disruption.