In the case of airlines recovering from flight delays， resources are indirectly linked to each other through the flight plan， and it is difficult to ensure the overall feasibility and global optimization of the recovery solution if each decision-making department considers the recovery of resources in its own area. This paper proposes a multi-agent simulation method for flight recovery in order to investigate the decision models of the departments and their collaborative relationship in the airline flight recovery process. Firstly， the framework of flight recovery multi-agent system is constructed based on the actual organizational structure of the airline. Then， the whole process of delay recovery is divided into three stages： Pre-recovery， feasible solution negotiation and balanced solution negotiation， and a three-stage collaborative decision-making mechanism is constructed. Finally， the core decision model of each department agent is developed according to the recovery characteristics of different resources， and the multi-agent auto-negotiation model is established based on the collaborative decision mechanism. The simulation results show that the multi-agent based collaborative decision-making method can produce a complete delay recovery plan in 3.8 s for a flight schedule consisting of three aircrafts and twelve flights in one day， and it can balance the local interests of each decision maker to a certain extent while safeguarding the overall benefits of the airline.