Detection and Diagnosis of Mechanical Seal Faults in Centrifugal Pumps Based on Acoustic Measurement

Mechanical seal is one of the crucial components used in fluid machines to prevent leakages. The performance degradation of seals is generally because of improper operations, scratching, strong force, wearing and overheating which leads to low-performance operation and environmental contamination. In this paper, the airborne sound signals are investigated to monitor the mechanical seal faces and friction conditions. Based on the operation mechanisms of seals, failures of seals often generate additional vibration and sound due to direct contact of two sliding surfaces. Theses vibro-acoustics can exhibit a broadband characteristic since asperity collisions occur in the rough seal surfaces in the form of a series of random impulses. In addition, this produced sound with broadband can be modulated by the rotational components at low-frequency range. Based on these understanding, it is possible to extract the fault signatures using modulation signal bispectrum (MSB) to analyse the noisy sound signals as it has the unique capability to demodulate the signal along with noise reduction. Experimental results show that sound signals associated with MSB analysis allow mechanical seal faults to be detected in the early stage remotely. Also, the MSB is superior to identify the seal failure than spectrum analysis. Moreover, the results showing the effectiveness of remote acoustic based monitoring and gives a good indication of full separation between fault cases under different pump operation conditions.