Due to the development of novel Internet applications, Fault-tolerant needs to be improved and QoS (Quality of Service) routing capability is necessary in the network. Fault-tolerant QoS routing is one of the effective mechanisms to solve this problem. Considering the dynamic characteristic of network, we introduce the knowledge of fuzzy mathematics and probability theory to describe network state and adopting a preventive shared-path protective strategy, we proposed an intelligent fault-tolerant QoS routing mechanism based on the PSO (Particle Swarm Optimization) and SA (Simulated Annealing) hybrid optimization in the case of single SRLG (Shared Path Risk Link Group) failure. The goal is to find a work path and a backup path to satisfy the user QoS constraints, which optimizes the end-to-end reliability, cost and QoS parameters of the path. At last, simulation results in NS2 reveals that compared with previous approaches, the percentage of route validity and fault-restoring ratio is increased. It shows that our mechanism is feasible and effective.