In order to address the problem that the existing susceptible-infected-recovered(SIR) model does not consider the impact of the topology of the airport network on the flight delay propagation, the topological characteristic index of the airport network was calculated based on the complex network theory.The comprehensive importance of each airport was evaluated by using the entropy weighted technique for order preference by similarity to ideal solution(TOPSIS) method, and all airports were divided into several classes, thus the improved SIR model for the flight delay propagation of airports in different classes was established and the propagation thresholds of the model were solved. Taking the Chinese airport network in 2023 as an example, all airports were classified into five categories, and an improved SIR model was used to simulate the flight delay propagation of airports under different initial conditions. The simulation results show that the improved SIR model is more suitable for simulating delay propagation.The higher the comprehensive importance of an airport, the wider the delay propagation range becomes, the faster the speed is, and the longer the recovery period lasts.The delay propagation ability of E-class airports is significantly improved, when flights from multiple airports are delayed.The topology of the airport network has an inhibitory effect on the delay propagation, and it is difficult for delays to spread to all the airports. The improved SIR model demonstrates the differential effects of airport importance on delay propagation and provides reference for the formulation of flight delay management strategies.