A vibration cavitation sensitivity parameter based on spectral and statistical methods

• Octave band analysis and PCA used on RMS velocity to obtain key indicators.
• Mahalanobis distance is used to set thresholds since data is normally distributed.
• Pump health separated into no cavitation, incipient, and fully formed condition.
• Method works on a range of types and sizes of centrifugal pumps.

Cavitation is one of the main problems reducing the longevity of centrifugal pumps in industry today. If the pump operation is unable to maintain operating conditions around the best efficiency point, it can be subject to conditions that may lead to vaporisation or flashing in the pipes upstream of the pump. The implosion of these vapour bubbles in the impeller or volute causes damaging effects to the pump. A new method of vibration cavitation detection is proposed in this paper, based on adaptive octave band analysis, principal component analysis and statistical metrics. Full scale industrial pump efficiency testing data was used to determine the initial cavitation parameters for the analysis. The method was then tested using vibration measured from a number of industry pumps used in the water industry. Results were compared to knowledge known about the state of the pump, and the classification of the pump according to ISO 10816.