Diagnostic method by using vibration analysis for pump fault detection

Pump cavitation is a common problem affecting systems. It occurs when the pressure of the liquid in the pump drops below a threshold and causes the liquid to vaporize creating tiny bubbles that, when they implode or collapse, trigger intense shockwaves inside the pump determining damage. Excessive viration on the pump casing could indicate cavitation. Consequently, vibration monitoring can help the detection and prevention of the harfull and undesired phenomenon. The purpuse of the research is toanalyze the capability of vibration-based techniques to detect, monitor and prevent pump cavitation. Experimental tests were performed on a gear pump used in the lubrification circuit of internal combustion engine. The pump, installed on a dedicated test bench, was forced to cavitate by plancing a calibrated orifice on the suction side. Main working parameters, like oil flow rate, suction and delivery pressure, shaft speed, were accurately measured. Different pump operating conditions with and without cavitation occurrence were investigated through the use of a non-intrusive acelerometer, installed in proximity of the port so as to monitor the phenomenon in terms of vibration amplitude. A preliminary spectral analysis, based on the Fast Fourier Transform (FFT) of the vibrational signal, was performed in order to easily cavitation fundamental frequencies. A time-domain analysis technique was then implemented, aiming to realize an on line pump cavitation detection. Specifically, a NonLinear AutoRegressive (NLAR) approach based on the use of Artificial Neural Networks (ANN) was applied for modeling system behaviour. In the paper, the results of the vibration-based method are discussed on depth, highlighting the pros and cons of the methodology. The presented outcomes demonstrate the ability of the proposed algorithm in accurately detect the presence of cavitation phenomena and to determine its intensity in pump real time operation. Hence, it may turn out to be a powerful tool for early detection of pumps incipient faults. ACKNOWLEDGMENTS The research presented in this paper was performed as part of the European H2020 DEVILS project, funded by the European Union's Horizon 2020 research and innovation programme under grant agreement Nº 737972. The authors gratefully acknowledge the EU financial support.