Sensitivity analyses in a buoyancy-driven closed system with high resolution CFD using Boussinesq approximation and variable density models

Jonathan K. Lai, Elia Merzari, Yassin A. Hassan

Research output: Contribution to journalArticle

2 Scopus citations

Abstract

The influences of the Schmidt number and fluid models of the Boussinesq approximation and variable density (low-Mach formulation) are investigated using large eddy simulation (LES) where fluid naturally advects between two tanks connected by a pipe. The geometry resembles a reactor-like vessel which comprises of a cold leg-downcomer region. The closed system consists of two miscible fluids with different densities and viscosities separated by a valve. Analysis with Reynolds-Averaged Navier–Stokes (RANS) is used for calculating Taylor length scales to establish minimum grid resolution for LES. Using particle image velocimetry (PIV) data, validation is conducted for four different magnitudes of Schmidt numbers. Evidence indicates that capturing the effects from the highest Schmidt number is essential for properly predicting fluctuations and instabilities in the flow. Although downcomer comparisons with PIV consist of larger errors, they are explained through additional sensitivity analyses involving initial conditions, geometry, and residual energy. Lastly, the influence of larger density differences is demonstrated using simulations with a higher Atwood number, temporally scaled using the fluid front velocity. Recommendations are made regarding future experiments and simulations for design of similar buoyant mixing problems.

Original languageEnglish (US)
Pages (from-to)1-13
Number of pages13
JournalInternational Journal of Heat and Fluid Flow
Volume75
DOIs
StatePublished - Feb 1 2019

All Science Journal Classification (ASJC) codes

  • Condensed Matter Physics
  • Mechanical Engineering
  • Fluid Flow and Transfer Processes

Fingerprint Dive into the research topics of 'Sensitivity analyses in a buoyancy-driven closed system with high resolution CFD using Boussinesq approximation and variable density models'. Together they form a unique fingerprint.

  • Cite this