The theme of this year's conference is "Navigating Through Disruptions with Analytics," which highlights the pivotal role of analytics in addressing the challenges posed by political, natural, and other disruptions.

Here is an overview of the conference program:

• 19 technical Operations presentations 
• 6 technical Maintenance presentations

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

Operations Presentations -- Still under review

Proactive vs. Reactive Disruption Management: Enhance Trust and Efficiency

Julius Bauß; M2P

Disruptions in airline operations demand quick decisions. While recovering the schedule via a solver is a powerful reactive method, users often struggle to understand and trust its solutions, leading to hesitancy in applying them.

We propose a proactive approach: pre-identifying good cancellation options. Using a simplified mathematical model, these options are generated within seconds, providing understandable solutions. This enhances user confidence and trust in the solver as a support tool.

A Density-Based Traffic Following Algorithm for Air Traffic Control

John Paul Clarke; UT Austin / Cayman Airways

--

Modeling Airline Disruptions In Order To Predict Optimization Complexity And Solution Shape

Thierry Delahaye; Amadeus

We explore models for predicting the complexity and potential solutions for airline disruptions.

Key features such as disruption nature, location, and timing are extracted and used to predict the complexity of solving the instance as well as the likely solution shape, such as whether delays or cancellations are required.

Our predictions can be used to determine which profiles should be run on the optimizer, and to inform users of likely decisions to be taken by the optimizer.

Two-Stage Stochastic Optimization for Resilient Airline Operations: Cost Efficiency and On-Time Performance Under Ground Time Uncertainties via CI Adjustments and Hub-Centric Swaps

Mehmet Ertem; Turkish Airlines

This study proposes a two-stage stochastic optimization model to minimize total operational costs while improving OTP under ground uncertainties. First-stage decisions at the initial hub cycle optimize pre-departure adjustments: aircraft swaps, departure delays, and cruise speeds controlled via CI adjustments. Second-stage decisions, applied in the subsequent hub cycle, re-optimize CI and swaps to mitigate realized disruptions.

Real-Time Recovery: A Duty Manager’s Daily Experience with Automated Disruption Management

Horacio Garcia Esparza, Pau Collellmir Cardenal; Viva Aerobus, BIGBLUE Analytics

Horacio Garcia Esparza, Viva Aerobus’s duty manager, shares his daily routine using an automated disruption management solution. He explains how he monitors real-time data, rapidly evaluates recovery options, and makes swift decisions to keep flights on schedule—achieving a 40% reduction in disruption costs. This technical presentation offers practical, day-to-day insights into managing airline disruptions using an automated solution.

Passenger Recovery in Lufthansa Group’s OPS Suite

Kadir Göcer, Daniel Kappeler; Lufthansa

Lufthansa Group, in collaboration with Google Cloud and Google Research, builds its Operations Decision Support Suite (OPSD) to enhance efficiency, customer experience, and sustainability. OPSD integrates diverse solvers and data into a flexible framework. This presentation highlights the Cancellation Impact Analyzer (CIA), an early module optimizing passenger rebooking post-cancellation, showcasing production examples and its planned integration into disruption recovery.

Operations Centric Robust Airline Scheduling

Pranav Gupta; Amadeus

Traditional scheduling methods with little consideration for operational disruptions often lead to discrepancies between planned and actual metrics. This work highlights the importance of modeling a digital twin of day of ops and evaluating schedules for reliability. Using simulation guided optimization and ML on historical data, we predict delays and optimize schedules to maximize on-time performance and minimize propagation delays.

Machine Learning-Driven Classification of Schedule Interruption Data

Sean Huynh, Gevik Sardarbegians; Boeing

Boeing Data Science teams are developing ML models to classify Schedule Interruption (SI) data across seven critical attributes. The project aims to improve airline dispatch reliability through automated classification, targeting 70% accuracy in Phase I and 90% in Phase II. Implementation involves iterative model refinement with SME validation of 10,000 records. This cloud-based solution will enhance Boeing's ability to analyze maintenance data and improve aircraft reliability globally.

Quantifying Baggage Mishandling : Analyzing Key Factors and Operational Improvement

Rim Jabri; Groupe ADP

Mishandled Bags are a major challenge to the aviation industry, affecting profitability and passenger satisfaction. An econometric model has been developed to identify and quantify key factors influencing baggage loss, including maintenance, flight delays, and airport infrastructure.

By analyzing these factors and using counterfactual analysis, the potential reduction in mishandling rates under alternative scenarios is estimated, providing insights for operational improvements.

Closing the Loop: Turning Flight Delay Predictions into Actionable Strategies

Anand Krishna; ISA

Flight delays impact costs, customer experience, and airline reputation. While ML-driven delay predictions help, their effectiveness depends on integration with operations. Accuracy declines over time, requiring a balance between early intervention and incorrect forecasts. Our framework aligns predictions with structured workflows, enabling risk-based decisions and automated responses. By embedding predictive insights into OCC operations, we enhance efficiency, reduce disruptions, and improve airline performance.

Simulating Flight Disruptions to Improve Recovery Strategies of a Disruption Management Tool

Markus Kühlen, Shayekh Hassan; German Aerospace Center (DLR)

A flight disruption management tool offers data-driven decision support in an airline’s Operations Control Center. This presentation showcases how simulating potential disruptions, such as flight delays and aircraft on ground events, can be used to test and improve the recovery strategies applied by a flight disruption management tool. The presented approach helps to minimize the impact of real disruptions, resulting in better cost efficiency and improved on-time performance.

Data-Driven Insights to Maximize Air Traffic Efficiency

Jennifer Leslie, Emma Thompson; United Airlines

With increased aviation demand, traffic flows need to be effectively managed, while ensuring a safe operation. A new analytics team focuses on air traffic business rules while understanding the data. This specialized team provides data insights to maximize efficiency and support our customers. This presentation discusses the need for this team, the challenges, plus our vision for relevant analyses and data driven support across three timeframes: post event, real-time, and eventually, predictive.

Nasa Center for Air Transportation Resilience: Data-Driven Research and Analytics to Identify, Mitigate Disruptions and Create Foundations for a More Resilient Air Transportation System

Max Z. Li; University of Michigan

The Center for Air Transportation Resilience (CATRes), a NASA University Leadership Initiative, conducts research to improve the resilience of the air transportation system. In this presentation, we will provide an overview of the CATRes initiative and detail our current research activities, such as: cluster analyses to identify historical disruptions; generative models to create synthetic disruptions for training; airline-air traffic collaborative optimization, and passenger delay estimation.

Improved Regret Bounds for Online Decisions in Airline Recovery

Lavanya Marla; University of Illinois at Urbana-Champaign

In this work, we present a novel methodology to construct tight performance bounds for airline recovery problems. Our approach expands upon the theory of penalized information-relaxation bounds developed by Brown, Smith and Sun (2010). We expand this approach to airline recovery by presenting a novel way to avoid enumerating all system states. Our new bounding approach improves upon existing bounds and helps better evaluation of predictive-prescriptive models for airline recovery.

A Genetic Algorithm–Based Optimizer for Balancing Planned Cask and Schedule Quality

Evert Meyer, Alan Sagan, Ben Hinton-Lever; Virgin Australia

Airlines struggle balancing planned CASK vs. operational robustness in short-term scheduling. We present a genetic algorithm (GA) based optimizer that improves schedule quality by including trade-offs between turn times, ops spares, and crew productivity. By encoding aircraft rotations as “genes,” it explores a broad solution space and builds an efficient frontier of key metrics. We'll share design, deployment, and present early results around resource utilization and on-time performance.

Augmenting Operations Research with Large Language Models

Sumaya Abdul Rahman, Sanjay Chawla; Hamad Bin Khalifa University, Qatar

We will showcase ways in which Large Language Models (LLMs) can augment Operations Research solvers. The first scenario is in problem formulation, i.e., using an LLM to map a problem into a well-specified Integer Program that can be solved by an OR solver. The second scenario is solution verification and "what if scenarios" where we can use Chain of Thought abilities of LLMs to reason with the output of OR solvers. The work will highlight the increasing use of LLMs in Operations Research.

Predictive and Prescriptive Analytics: Proactive Decision-Making in Airline Control Centers

Sadaqat Ali Soomo, Abdallah Kourri; Qatar Airways

Our presentation outlines a proactive strategy for an airline control center using predictive and prescriptive analytics. Predictive analytics anticipates events like delays, while prescriptive analytics recommends optimal actions. The key aspect of this strategy is the Network Stress Index, which is a composite measure derived from Delay prediction and propagation models and stress indicators like weather condition. Monitoring this index improves decision-making and enhances network resilience.

Smart Gating at American Airlines: Enhancing Efficiency and Sustainability

Guodong Zhu; American Airlines

Smart Gating is a patent-pending gate optimization engine developed at American Airlines. It optimizes gate assignments to improve separations, balance terminal-runway operations, and reduce congestion, creating a resilient plan that minimizes pushback conflicts and taxi times, thereby enhancing on-time performance, cutting fuel use, and lowering emissions. Deployed at American Airlines' five largest hubs, it cuts taxi time by 17 hours daily and CO₂ emissions by over 13,000 metric tons annually.

Reducing Distributional Impact Of Commercial Space Launches

Roberta Zimmerman, William Steinberg; United Airlines, JetBlue

U.S. carriers reported significant disruptions during one scrubbed launch; 99,000 delay minutes, 83 cancelled flights, 9,578 cancelled customers, 2,640 delayed flights, 60 gate returns, and over 485,000 passengers. Revised airspace protection and launch window refinement has enabled air carrier access to most of the Atlantic Routes offshore of Florida. We will analyze the impacts caused by commercial space launches and the mitigations that have reduced delays and effects on commercial carriers.

Maintenance Presentations -- Still under review

SRM Damage Type Classification

Sean Huynh, Gevik Sardarbegians; Boeing

Aircraft structures face diverse damage types including corrosion, fatigue cracks, impact damage and delamination. This project uses ML algorithms to automate damage classification from Logbook and BCS data, replacing time-consuming manual methods. The model analyzes maintenance records to identify 10 distinct damage types, with capability to add more. Early results show high accuracy and significant time savings, while maintaining consistent damage classification across the fleet.

Predictive Tool Development For Heavy Maintenance Non-Routine Findings

Sean Huynh, Gevik Sardarbegians; Boeing

Aviation heavy maintenance faces critical challenges from Non-Routine Findings during D Checks, with aircraft downtime averaging 24-30 days. These unexpected structural issues significantly impact operator costs. This study proposes a predictive maintenance planning tool enabling support teams to anticipate common findings based on historical data. The tool aims to optimize resource allocation and reduce operational disruptions.

Enhancing Leak Detection and Proactive Management in the 787 Integrated Cooling System (ICS) for Operational Efficiency

Aman Misri; United Airlines

In the complex realm of airline operations, traditional optimization-based scheduling models face significant limitations when addressing real-world challenges, such as uncertainties and nonlinearities inherent in dynamic environments. This research explores the integration of Reinforcement Learning (RL) into aircraft line maintenance scheduling—a task characterized by its dynamic and unpredictable nature—and highlights the transformative potential of AI-driven methodologies in airline scheduling. Extensive testing demonstrates that the RL-based approach may outperform conventional optimization models, delivering notable improvements in operational agility, cost efficiency, and flexibility.

Prognostic Health Management with Reliability
Shashvat Prakash; Collins Aerospace

A prognostic health management (PHM) strategy works when the failure is mode is wear-driven and detectable. PHM's promise is to reduce unscheduled interruptions at the cost of useful life. This lost life is largely a function of the uncertainty around remaining life. Merging reliability data with a prognostic estimate can both improve the detection and reduce the lost life.

An Automated Solution For E2E Heavy Maintenance Scheduling With Tail-Specific Check Creation

Ignacio Riego, Jaime Saez; LATAM Airlines

We present an automated solution using three integrated optimization models which interact in a feedback loop until convergence is achieved. First, maintenance tasks are grouped into checks; second, the duration of these checks is estimated; and finally, a scheduling model constructs an optimal long-term plan. Our approach has been shown to reduce strain on key resources, increase fleet availability and reduce overall cost compared to a schedule built with standardized checks.

Agent Based Reinforcement Learning Approach To Airline Maintenance Planning and Optimisation

Syed Shaukat; University of New South Wales

In the complex realm of airline operations, traditional optimization-based scheduling models face significant limitations when addressing real-world challenges, such as uncertainties and nonlinearities inherent in dynamic environments. This research explores the integration of Reinforcement Learning (RL) into aircraft line maintenance scheduling—a task characterized by its dynamic and unpredictable nature—and highlights the transformative potential of AI-driven methodologies in airline scheduling. Extensive testing demonstrates that the RL-based approach may outperform conventional optimization models, delivering notable improvements in operational agility, cost efficiency, and flexibility.