AIR INJECTION IN VERTICAL WATER COLUMN: EXPERIMENTAL TEST AND NUMERICAL SIMULATION USING VOLUME OF FLUID AND TWO-FLUID METHODS

SARACHE PIÑA, JOHAN; GODINO DARÍO; CORZO SANTIAGO; DAMIAN RAMAJO

CHEMICAL ENGINEERING SCIENCE

PERGAMON-ELSEVIER SCIENCE LTD

Año: 2022 vol. 255 p. 1 - 16

0009-2509

A novel test consisting on fast air injection into a vertical water column was experimentally and numerically studied. Measurements were focused on capturing the air-water interface as well as the volume of water evicted by the air, for a wide range of air flow rates. The numericalsimulations, were performed with the Eulerian Two-Fluid (TF) and the Volume of Fluid (VOF) methods.For the TF method three topologies were considered: bubbly flow, drop flow, and blending. The last is a more smart methodology to automatically handle with the different flow regimes. For the VOF method, the standar VOF (SVOF), the Adaptive Mesh Refinement (AMR), and the high order Piecewise Linear Interface Calculation (PLIC) methods were assessed. SVOF and TF blending methods were choice to study the mesh convergence and turbulence modeling. Two-dimensional (2D) meshes of 2, 1 and 0.5 mm, and three-dimensional (3D) meshes of 2, 1 and 0.75 mm were considered. In all cases the standard VOF (SVOF) formulation showed mesh convergenceand good agreement with experiments. The error for the finest 3D mesh was around 1%, but increased up to 20% for the finest 2D mesh. On the other hand, for the TF method mesh convergence was only evidenced for the 2D meshes. Regard the TF method, the bubbly and drop topology cases led to unacceptable solutions both in terms of inteface capturingas well as liquid evicted. On the other hand, the blending methodology clearly improved the estimations, although the interface was only partially captured because of the numerical diffusion. On the other hand, all the VOF methods were in relative good agreement both in terms of the liquid evicted as well as interface capturing. Errors were were reduced by refining the mesh. The initial swelling, with small bubblesand large slugs around the air injector was well captured, and the ligaments and drops spilled for high air flow rates were quite well estimated. The SVOF method showed low computational cost for coarse grids, but the computing time increased more than linearly with the mesh size. The AMR method showed accurate solutions, but the computing cost was largely increased with the refining level. Finally, the use of PLIC method for the coarser mesh reduced the error from 15% (SVOF) to 3.5% keeping low the computing cost.