The current bridge maintenance sy stem in China primarily relies on bridge inspection data to prioritize the state of bridges and allocate a limited budget accordingly.This strategy focuses on repairing bridges with low performance，concentrating on restorative maintenance but ignoring preventive maintenance.In view of the total lifecycle cost，the repair-oriented method may require a lower initial investment but leads to significant resource consumption in the mid to late stages.Based on existing highway bridge maintenance systems，an intelligent maintenance system relying on Chinese standards and bridge inspection data and combining repair and prevention was developed.The system model mainly consisted of two modules：the bridge performance degradation model and the intelligent maintenance decision model.Firstly，the long-term performance of the bridge was predicted based on dynamic Markov and Weibull models using long-term inspection data and mathematical statistics.On this basis，a mathematical model following preventive maintenance concepts was established，and it was optimized and solved by using swarm intelligence algorithms.Finally，a prestressed concrete box girder bridge on the Banyin Passage in Shenzhen was studied，and the model was tested using actual bridge inspection data.The results show that compared with traditional maintenance methods，the intelligent maintenance system can reduce maintenance costs by up to 10%.