For the problem of severer flight delays, based on the relative research achievements in domestic and aboard, this paper proposes a two-stage method for extracting flight delay patterns. Firstly, this method uses the phase-space reconstruction theory and Recurrence Plot to analyze the chaotic characteristics of flight delay rate. Then, after acquiring the delay time subsequences, the Recursive Quantitative Analysis and K-means Clustering technology were used to extract flight delay patterns, and the characteristics of time-varying of flight delay patterns were analyzed. Finally, the example verifies this method can not only effectively extract flight delay patterns, but also can get the time-varying characteristics of the patterns. The result provides managerial and decisive reference of prediction and warning of flight delay for airports and airlines.