Abstract: The post-earthquake functional recovery of water treatment plant systems is characterized by multi-component coupling, temporal evolution, and uncertainty propagation. A seismic resilience assessment method based on dynamic Bayesian network (DBN) is proposed. First, a state tree model of the plant is constructed and transformed into a DBN structure using logic gate mapping rules. Second, the damage state probabilities of structures, buildings, and connecting pipelines under different peak ground acceleration (PGA) levels are obtained from component seismic fragility models. Combined with post-earthquake state transition probability tables, the system functional recovery process is derived. Finally, the resilience index is calculated using a uniform resilience evaluation time, and then critical weak components are identified through sensitivity analysis. The method is applied to a surface water treatment plant. The results show that with increasing PGA, the system damage state gradually transitions from slight damage to severe damage and destruction, the functional recovery time is significantly prolonged, and the resilience index exhibits an overall decreasing trend. The grit chamber, main inlet well, and blower room exert significant impact on system functionality and are identified as critical nodes affecting the seismic resilience. The method can characterize the temporal evolution of component damage, functionality transfer, and post-earthquake recovery process within a unified probabilistic framework, providing a decision basis for pre-earthquake strengthening, post-earthquake emergency repair, and resilience enhancement of water treatment plant systems.