Stochastic simulation of delay propagation

Abstract

A large challenge for passenger airlines is the design of a profitable flight schedule, for example at Kenya Airways who operates a highly connected hub-and-spoke network. The goal of this research is create a fundamental insight into the stochastic nature of delay propagation in a passenger hub-and-spoke network which allows airlines to increase schedule stability. A model has been developed to simulate the propagation of delays through a flight network whilst incorporating passenger connectivity. The analogy that can be made is the toppling of dominoes, where the fall of the first domino illustrates the primary delay caused. The conceptual model is an activity-on-node flight schedule representation where a delay can propagate through either the lines of flight or via a passenger connection if the resource is not ready. Through a Monte-Carlo simulation per flight a relation is made and visualized between the duration of the primary delay and the delay severity, i.e. the number of flights the primary delay affects downstream. From this the Expected Delay Severity is derived, a proposed flight robustness metric, where the delay severity curve is weighed against the probability of a primary delay. The simulation has been validated using empirical data from the operations of Kenya Airways. Integrating passenger connectivity is found to be essential in representing the true delay propagation in a hub-and-spoke network. Implementation at Kenya Airways can be realized by a pro-active approach that uses the stochastic simulation of delay propagation and allows to control the system impact of disruptions by identifying factors that can improve schedule stability. Several flight-retiming improvements have been made and implemented using the proposed methodology.