Fault features of large rotating machinery and diagnosis using sensor fusion

Large rotating machinery such as turbines and compressors are the key equipment in oil refineries, power plants, and chemical engineering plants. To minimize the economic loss incurred because of the defects or malfunctions of these machines, diagnosis is very important. Currently, diagnosis is carried out mainly using spectral analysis. In spite of being effective in detecting the faults (monitoring), spectral analysis is often ineffective in pin-pointing what the fault is (diagnosis). This is due to the fact that it cannot clarify the spatial and temporal features in the sensor signals that are correlated to different types of faults. In this paper, phase spectra, holospectra, purified orbit diagrams, and filtered orbit diagrams are used in searching for fault features. From the data obtained from more than 50 practical machines, distinct fault features and diagnostic indices are found for 11 different types of faults including unbalance, cracks, misalignment, rub, loose bearing caps, oil whirl, surge, fluid excitation, rotating stall, electric power supply fluctuation, and pipe excitation. Accordingly, a diagnostic procedure is proposed. (C) 1995 Academic Press Limited.