Modeling of Natural Transition in Properly Three-Dimensional Flows

Abstract

The paper describes a method for viscous flow prediction which includes the prediction of transition onset location as an integral part of the calculation. The method is similar to others in that it is based on stability analysis of the boundary layers but acknowledges that solutions of the Reynolds-averaged Navier-Stokes (RANS) equations yield boundary layers of inadequate accuracy when conventional mesh sizes are used. The boundary layers are therefore recomputed to high accuracy by solving the laminar boundary layer equations. Transition onset locations determined by the stability analysis are passed back to the RANS solver for further iterations. The method described in this paper differs from others in that the laminar boundary layer calculation is fully three-dimensional and the stability analysis is through the fully three-dimensional parabolised stability equations for compressible flow. Results are presented to illustrate the capability of the laminar boundary layer method and the effectiveness of the method for solving the parabolised stability equations. In addition, preliminary validation results are presented from calculations of the viscous flow over the AFRL 1303 UCAV concept to demonstrate that the fully three-dimensional laminar boundary layer and stability analysis components can be combined successfully with a RANS method for calculating flows with natural transition.= components of the metric tensor associated with the ξ and ς directions (often written g 11 and g 33 ) = velocity components at outer edge of boundary layer in ξ,ς directions n = kinematic viscosity ω = specific turbulent dissipation rate, or the angle between surface grid lines in the boundary layer grid v = frequency of component of J

A Survey of Factors Affecting Blunt-Leading-Edge Separation for Swept and Semi-Slender Wings

“…The natural transition results showed altered flow patterns in the insipient separation and subsequent part-span vortex flow, as compared to fully turbulent computations, and the transitional computations agreed better with experiment. Subsequent publications addressed further modeling considerations and numerical effects (e.g., grid resolution) with application to six additional 1303 test conditions [Arthur 41 , 2007] as well as a more advanced computational approach based upon coupling RANS and three-dimensional boundary layer methods [Arthur 42 , 2009] with application to the 1303 configuration at a very low angle of attack.…”