We build a delay propagation network based on Bayesian Network approach to study the complex phenomenon of delay propagation within a large network consisting of the 100 busiest airports in the United States. Through topological analysis and probability analysis, we investigate the characterization of delay propagation among airports and the impact of different types of airports on delay propagation. Results indicate that the cumulative degree distribution of the delay propagation network follows an exponential function and flight delays take at most one transhipment to go from each airport to any other airports on average. For each individual airport, the effects of delay propagation are associated with airports size (traffic flow), small airports are easily affected by other airports while large airports are more affecting downstream airports but fewer affected by upstream airports. Finally, we show how the number of affected airports changes as a function of the delayed airports based on different simulation strategies.

我们基于贝叶斯网络方法构建一个延迟传播网络，以研究由美国100个最繁忙的机场组成的大型网络中延迟传播的复杂现象。通过拓扑分析和概率分析，我们研究了机场之间延误传播的特征以及不同类型机场对延误传播的影响。结果表明，延误传播网络的累积度分布遵循指数函数，航班延误平均需要最多一个转运从每个机场到任何其他机场。对于每个单独的机场，延误传播的影响都与机场规模（交通流量）相关，小型机场很容易受到其他机场的影响，而大型机场对下游机场的影响更大，而受上游机场的影响则较小。最后，我们展示了基于不同的仿真策略，受影响机场的数量如何随延迟机场的变化而变化。