Abstract: Analyzing and extracting fault feature information from equipment condition monitoring data to accurately and quickly identify equipment health status is very important for carrying out predictive maintenance and ensuring equipment operation reliability and safety. Taking the centrifugal compressor as the research object and taking the vibration monitoring data of the normal state of the equipment as the reference data,the correlation health index model,the coherence health index model,and the spectral distance health index model applying with the normal state original signal and the real-time monitoring signal are constructed. Based on a large number of failure case studies for the three health index models and statistical analysis of health degree distribution,equipment health rating criteria are developed,a data-driven mechanical equipment health evaluation method is formed,and the mapping relationship between centrifugal compressor health degree characterization and running status is revealed. The bearing experiment data and centrifugal compressor rotor imbalance failure data are used to verify the accuracy and applicability of the constructed equipment health index model and health rating criteria,respectively. The results show that the constructed health index models can better characterize the operating state of the equipment. Compared with the fixed threshold alarm methods of effective value and peak-peak,the constructed health evaluation criteria for mechanical equipment is more practical for guiding predictive maintenance.

Key words: health evaluation, case study, health index model, health rating criteria, predictive maintenance