An ensemble Synthetic Eddy Method for accurate treatment of inhomogeneous turbulence

Abstract

An ensemble approach to generating turbulent inflow boundary conditions using the Synthetic Eddy Method is proposed that improves signal accuracy in recovering target statistics for inhomogeneous turbulence while reducing the cost of generating an inflow signal. This is accomplished by eliminating typical restrictions related to the Synthetic Eddy Method such as homogeneous turbulence, limited shape functions, random eddy placement, and uniform synthetic eddy convection velocity. The approach gives flexibility and freedom to prescribe arbitrary synthetic eddy length scales, shape functions, eddy placement, and convective velocities while accurately reproducing input Reynolds stresses and maintaining the simplicity of the original method. Three examples of leveraging this flexibility to provide accurate turbulent length scale inhomogeneity and reduce the required number of generated synthetic eddies are provided. Turbulent inflow signals for a channel flow configuration are generated using six variations and the resulting signal accuracy and costs are compared. These inflow signals are applied to a set of Large Eddy Simulations of turbulent channel flow and show favorable results over the original method in terms of the skin friction coefficient convergence, Q-criterion plots, and computed Reynolds stresses. The ensemble approach is shown to be more accurate in reproducing desired turbulent characteristics compared to the original Synthetic Eddy Method, in a simpler manner compared to recent variants of the method.