Abstract: As the commercial air transport increasing in China, the problem of flight delays is increasingly serious. Wake separation (the minimon separation between the leading and following aircraft to avoid wake encounter) limits the capacity of airports. Aimed at this problem, this paper study the evolution of wake vortex with large eddy simulation (LES). The self-adaptive grid method is applied to the LES of wake vortex to improve the computation efficiency and a lift-drag model is applied to the wake vortex generation process to simulate the roll up phase. Based on the LES wake vortex evolution database, a fast-time wake separation prediction system is established. Given the real time ambient wind field and the aircraft parameters, the prediction system can output the suggested wake separation. The results show that under the average wind condition in Beijing Capital International Airport in 2014, the current wake separation can be reduced by 7%~50% with the established system, thus the airport capacity can be considerably improved.