A fault diagnosis method for centrifugal compressors based on ant colony clustering algorithm

With the improvement of industrial production level, the development of centrifugal compressor follows thedirection of large scale, complication, continuity and automation, and the loss caused by equipment failure is increasinglysevere. Therefore, the study on the fault diagnosis methods is of active significance to improve the ability of on-lineequipment fault detection and prevent the emergence of failures and accidents. In this paper, the radial displacement signalat the side of the shaft coupling of centrifugal compressor was measured by the eddy current sensor. Then, Fourier transformwas conducted on the collected signal data by means of frequency domain analysis method to obtain the fault-sensitivecharacteristic value, which was taken as the input characteristic parameter of ant colony clustering algorithm to transformthe fault identification into the clustering of the output and state characteristic in the process of equipment operation. Patternrecognition of centrifugal compressors in a ethylene plant was carried out by ant colony clustering algorithm in the cases ofnormal operation, rotor imbalance, oil film vortex and surge failure. Experiment results indicate that the diagnosis resultsobtained by virtue of the fault diagnosis method based on ant colony clustering algorithm for centrifugal compressors areaccurate and have high recognition ratio.