CBL - Campus del Baix Llobregat

Projecte llegit

Títol: Robust Airline Routing with Quantile Buffers: Fleet CapEx vs Leasing/OPEX under Delay Propagation

Estudiants que han llegit aquest projecte:

Director/a: TRAPOTE BARREIRA, CÉSAR

Departament: FIS

Títol: Robust Airline Routing with Quantile Buffers: Fleet CapEx vs Leasing/OPEX under Delay Propagation

Data inici oferta: 27-01-2026     Data finalització oferta: 27-09-2026


Estudis d'assignació del projecte:
    GR ENG SIST AEROESP
Tipus: Individual
 
Lloc de realització: EETAC
 
Paraules clau:
Airline, routing, MLP, blocktime, buffer, capex, opex, compensation, delay
 
Descripció del contingut i pla d'activitats:
This work develops a aircraft-routing MILP into a scenario-driven study on robustness vs fleet cost. The core is a one-day tail assignment / aircraft routing model with minimum turnaround time and a fleet cap, used first as a feasibility test and then to minimize fleet size and/or connection idle time. The robustness layer is introduced by buffer allocation on flight times and turnarounds using percentiles (e.g., p50-p95) or explicit buffers derived from empirical distributions, consistent with industry practice of block-time reliability targets.
The student will generate scenarios where increasing buffers reduces delay propagation but increases the aircraft required (fixed cost / CapEx) or triggers outsourcing (wet-lease / leasing OPEX). The economic layer extends the routing model by allowing each flight to be operated in-house or outsourced, and by trading additional aircraft (daily equivalent CapEx) against wet-lease premiums.
Outputs are "paper-ready" figures: (i) required fleet vs buffer percentile, (ii) idle/slack vs fleet cap, (iii) cost frontier (CapEx vs wet-lease/OPEX) across buffer policies, and (iv) a robustness metric (e.g., probability of breaking next departure, accumulated delay proxy) linked to buffer selection. The thesis is designed so you can steer which scenarios and plots to produce for the paper, while the student executes a controlled, reproducible pipeline.
 
Overview (resum en anglès):
The schedule is the core product of an airline, and designing it means deciding where to place a limited amount of slack: in the airborne block time (SBT) or in the ground turnaround (TAT). A padded schedule protects punctuality but requires more aircraft, while a tight one improves fleet utilisation at the price of delay propagation along each rotation. This thesis studies that trade-off for Iberia Express (IBS) and addresses two questions: whether the airline currently operates close to its economic optimum under its present buffer policy, and whether its real aircraft routing can be reproduced with an integer linear programming model built exclusively on open data.
The empirical basis is a one-year dump of 41,360 Flightradar24 flights (September 2022 to September 2023), used to characterise the network, fit LogNormal delay distributions per route and turnaround distributions per airport, and measure the propagation between consecutive flights (Pearson correlation ¿ = 0.405). The same analysis is replicated on Vueling and Ryanair Group as sector references. A set-partitioning tail assignment model is then formulated, parameterised by two robustness quantiles (¿TAT and ¿block) and solved through two independent methods that validate each other: branch-and-bound on the explicit integer program and maximum bipartite matching, whose equivalence is proven from Dilworth's theorem. Each Pareto solution is monetised with a Monte Carlo simulator (500 replications per point), using an attenuation coefficient ¿ = 0.508 calibrated against the observed correlation and an annualised cost function that combines fleet CapEx, EUROCONTROL delay costs, EU261 compensations and ACMI wet-lease.
The model optimum (K = 25, ¿TAT = 0.30) costs €341M per year, whereas the reconstruction of the real IBS schedule (K = 23) yields €260M; the gap is explained by the airline's published SBT, which is more generous than the empirical 95th percentile. Keeping that real SBT and letting ¿TAT and K vary, the optimum appears at ¿TAT = 0.05 with K = 22 and €259.9M per year, less than 0.1 % away from the real operation. Iberia Express therefore operates at its local economic optimum, and the scenario analysis identifies the airborne buffer as the most efficient scheduling lever.

© CBLTIC Campus del Baix Llobregat - UPC