Abstract: Alarm association rules mining is an important task in system fault diagnosis and localization. Once the system fails, it will produce a large number of alarm information. By analyzing the characteristics of the booking system alarm data, this study puts forward alarm association rules mining algorithm based on sliding time window model to find the fault source and the correlation between fault factors in a large number of alarm information. The experiments show that the valuable alarm association rules can be acquired from the alarm data accurately and rapidly. These rules can provide support decision for the system maintenance personnel.