Evidence generating methods in fusion diagnosis of vibration isolation system

Fault diagnosis by fusing information from multi-sources can provide a quicker detection and reduce losses. When using evidence theory to diagnose faults, how to generate appropriate evidence draws much attention. To get proper evidence, features are extracted in both time domain and frequency domain, and evidence generating functions are designed to process features. Use proposed methods to generate evidence from measured data in vibration isolation system and fuse obtained evidence by Dempster's rule. Results show that both of the two methods proposed in this paper can diagnose faults in time. In comparison, method in time domain has a quicker running speed, while the one in frequency domain behaves better in sustained faults.