Application of Linear and Non-linear k-? Models in Prediction of Flow and Heat Transfer in Channel with Detached Ribs

Roughness elements or turbulence promoters have been widely used to enhance heat transfer in cooling passages of modern gas turbine blades. Although such ribs substantially enhance heat transfer, the heat transfer coefficient is reduced immediately at corner downstream of each rib, creating hot spots. To remove such hot spots some of the ribs can be detached from the channel walls. In this paper, this idea is investigated using numerical methods. In this study, turbulent flow and heat transfer through three types of channels, namely: 1) a channel with detached ribs from a wall 2) a channel with detached ribs from both walls and 3) a channel with alternative attached-detached ribs on both walls have been investigated. Computations presented in this research have been obtained with the linear and non-linear k-? models. The numerical method in this work is finite volume methodology and simple algorithm. The governing equations are discretized in a semi-staggered grid system. In all equations the convective terms are approximated using Hybrid scheme. The numerical results for the channel with detached ribs close to one principal wall showed that both the linear and non-linear k-? models are able to predict the length and width of the wake downstream of the detached ribs, although both models produce weaker wakes compared with experimental data. Both models "specially the non-linear k-? model" predict lower stream-wise velocity and turbulent intensities closed to the ribbed wall. As a result, both turbulence models "and specially the non-linear k-? model" substantially under-predict the wall heat transfer. For the channel with ribs detached close to both walls, both turbulence models produce better heat transfer predictions though still predict lower heat transfer levels. Finally, for the channel with alternative attached-detached ribs, both turbulence models fail to predict reliable heat transfer levels in first half of the channel but return acceptable Nusselt levels in the second half of the channel.