Flow and heat transfer in cross-stream type T-junctions: A computational study

Abstract

The present computational study is concerned with the thermal mixing of flow-crossing streams in a T-shaped junction, focussing primarily on a configuration subjected to temperature-dependent fluid property conditions. The reference experimental investigation is conducted by Hirota et al. (2010). Preliminary, a quasi two-dimensional configuration with constant fluid properties, for which the reference DNS (Direct Numerical Simulation) database is made available by Hattori et al. (2014), is simulated. The presently applied computational model is based on a VLES (Very Large Eddy Simulation) formulation of Chang et al. (2014). The residual turbulence is modeled employing the appropriately modified RANS-based (Reynolds-Averaged Navier–Stokes) elliptic-relaxation eddy-viscosity model of Hanjalić et al. (2004). In addition to the VLES, both flow configurations are computed applying the background RANS model representing the constituent of the present VLES method. Whereas the eddy viscosity model describes fully-modeled turbulence in the RANS framework, it relates to the unresolved sub-scale turbulence within the VLES methodology. Unlike the RANS method, the VLES method is capable of capturing the spectral dynamics of turbulence to an extent complying with the underlying grid resolution. The latter model feature contributes decisively to an appropriately intensified turbulence activity in the separated shear layer regions and, consequently, to an enhanced mixing process. The results obtained with the present VLES model follow closely the reference DNS data in the Hattori et al. (2014) case with respect to velocity and temperature fields as well as the fields of associated turbulent quantities in all characteristic regions of the flow domain: main and branch streams’ merging zone, flow-reversal and post-reattachment regions. In the more complex Hirota et al. (2010) configuration, the flow field is captured reasonably well, while the computationally obtained thermal fields suggest a somewhat more intensive mixing relative to the reference experiment.