Application of a near-wall domain decomposition method to turbulent flows with heat transfer

• A non-overlapping domain decomposition method is applied to three RANS models.
• The algorithm does not include any free parameters.
• It significantly reduces the computational time whilst retaining sufficient accuracy.
• The approach is tested on several tasks with heat transfer.
• The solutions show little sensitivity to the size of the near-wall region.

A near-wall domain decomposition method for use in turbulence modelling is applied to the k-ωSST, Spalart–Allmaras and BL-v2‾/k turbulence models. The near-wall region is excluded from the main computational mesh. This eliminates the expense of computing the solution in the viscous sub layer and reduces the total computation time.A one-dimensional boundary layer equation is used to transfer the wall boundary condition from the wall to an interface located within the flow domain. The boundary conditions imposed on the interface are of Robin type and are written in mesh-independent form. The boundary layer equation can contain source terms such as the pressure gradient or near-wall damping terms. Scalar boundary conditions can be calculated using the same formalism as the boundary conditions for the velocity.The implementation of the boundary conditions is tested on a channel flow, two heated annulus flows and a two-dimensional, asymmetric diffuser. For each case, different locations of the interface boundary are tested. The results are not sensitive to the location of the interface. Friction factors and heat transfer data calculated with the domain decomposition approach compare well to the results obtained with the fully-resolved forms of the respective turbulence models.