In order to assess effectively the passenger evacuation performance in ship fires, a quantitative method combining the computational fluid dynamics (CFD)-derived fire data and AnyLogic evacuation simulation is proposed in this paper. In this developed method, the speed reduction factor, instantaneous untenable conditions and fractional effective dose (FED) model are utilized to consider the effects of smoke visibility, heat and toxic gases in ship fires on the passenger walking speed, activity space and human health conditions. In order to demonstrate this proposed method, passenger evacuation analysis is performed in a hypothetical zone of the ship with obstacles for fast t-squared fires. The results indicate that the relationship between the required safe egress time (RSET) and fire heat release rate (HRR) is non-monotonic. However, the total FED increases with the in-crease of the fire HRR and the sequence of survivors. Compared with two existing methods, this proposed method is able to assess more rigorously the passenger evacuation performance in ship fires. The existing ASET/RSET method is more conservative, and the existing FED method with the Pathfinder underestimates the FED in ship fires. This work will provide a more accurate method for ship fire safety assessment and performance-based fire protection design.