Enhancing the survivability of a new data-driven robust airline hub network with risk-averse criterion

The widely adopted hub-and-spoke architecture in airline network designs can trigger cascading effects during disruptions and result in further losses. This paper aims to enhance the survivability of airline hub networks in the design phase by optimizing reliability, sustainability, and efficiency. However, it is challenging to account for reliability under unpredictable disruptions such as interdiction and natural disasters. Under the stimulation of available event information, it is a promising solution to incorporate uncertain disruption scenarios into reliable airline hub network design through a data-driven robust approach. This paper develops a bi-level multi-objective optimization framework, and builds new risk-neutral and risk-averse models under the worst-case mean and conditional-value-at-risk criteria, where data-driven ambiguity sets are constructed through statistical hypothesis testing, and empirical probability distribution is determined by fault tree analysis. The constructed ambiguity sets have probabilistic guarantee, which help us transform the proposed models into mixed-integer second-order cone programming models, for which an effective branch-and-cut algorithm is designed. Numerical experiments on a real case demonstrate the robustness and reliability of our location-routing decisions. The results also illustrate that our data-driven approach outperforms stochastic optimization approach in out-of-sample performance and that the proposed branch-and-cut algorithm surpasses the Gurobi solver in computational efficiency.