A CFD and experimental study on cavitation in positive displacement pumps: Benefits and drawbacks of the 'full' cavitation model. Issue 1 (1st January 2016)

• Record Type: Journal Article
• Authors: Iannetti, Aldo; Stickland, Matthew T.; Dempster, William M.
• Is Part Of: Engineering applications of computational fluid mechanics. Volume 10:Issue 1(2016)
• Page Start: 57
• Page End: 71
• Publication Date: 2016-01-01
• Subjects: cavitation in PD pumps -- air fraction and cavitation -- experimental validation of cavitation CFD model -- full cavitation model; Computational fluid dynamics -- Periodicals; 620.10640285
• DOI: 10.1080/19942060.2015.1110535
• Languages: English
• ISSNs: 1994-2060

Abstract

ABSTRACT: To fill the gap in the literature in terms of numerical studies of positive displacement (PD) pumps in a cavitating condition, a comprehensive and transient computational fluid dynamics (CFD) model of a PD pump, simulating the cavitation arising during the suction stroke, was created. The 'full' cavitation model was utilized to study its capability on PD pump cavitation. A set of three plunger speeds were simulated. Using the highest plunger speed, an assessment was made of the effect of 1.5, 3, 4.5 and 15 parts per million (ppm) of air mass fraction on pump performance and cavitation. An experimental test rig, replicating the CFD model, was designed and built in order to validate the numerical model and find its weaknesses. CFD modeled, in a consistent way, the fluid dynamics phenomena related to cavitation (the chamber pressure approaching the vapor pressure, the vaporization/condensation and the pressure spike occurrence at the end of the suction stroke marking the end of cavitation). On the other hand the CFD pressure trends calculated appeared stretched along the time axis with respect to the experimental data, and this highlighted issues in the multiphase and cavitation models: the vaporization/condensation rate calculated by CFD did not follow the real dynamics correctly because the non-condensable gas expansion was overestimated. This was seen when comparing the CFD/experimental results where the simulated pressure drop gradient at the beginning of the suction stroke and the pressure peaks as the valve closed exhibited a delay in their occurrence. The simulation results were sensitive to the dissolved air mass fraction as the delay depended on the amount of air dissolved in the water. Although the influence of the air mass fraction was considered consistent, the 3 ppm CFD case was the closest to the experimental results, whereas the analyst expected the 15 ppm case to be more accurate.