03.1 - Aerodynamics – CFD Methods and Validation
TRANSITION MODELING IN SUPPORT OF CFD VISION 2030 - HIGHLIGHTS OF RECENT EFFORTS AT THE NASA LANGLEY RESEARCH CENTER
M. Choudhari¹, E. Beyak¹, N. Hildebrand¹, F. Li¹, E. Vogel¹, P. Paredes, National Institute of Aerospace, United States; V. Srivastava², B. Venkatachari²; ¹NASA Langley Research Center, United States ;²Analytical Mechanics Associates, United States
According to CFD Vision 2030, the most critical area in CFD simulation capability that will remain a pacing item is the ability to adequately predict viscous turbulent flows with possible transition and flow separation. Established methodology for transition prediction correlates the onset of transition with the linear amplification of instability waves, but automated calculations of the amplification factors pose a major challenge for routine CFD analyses, especially in the presence of strong viscous-inviscid interaction. The more recent approach based on the recasting of empirical transition correlations in the form of auxiliary transport equations is computationally robust, but cannot be easily generalized to capture the physical complexity of the transition process across the speed regime. This paper provides an overview of the recent advancements in CFD-integrated transition modeling at the NASA Langley Research Center. The dual-pronged strategy is aimed at simultaneous improvements in the robustness of automated stability computations coupled with RANS flow solvers along with the accuracy and the physical basis of transport-equations-based transition models. The paper outlines the lessons learned from the implementation and assessment of both approaches in NASA flow solvers, including their assessment in the context of various canonical configurations, encompassing both 2D and 3D boundary layers with a variety of pressure gradients, edge Mach numbers, and multiple instability mechanisms that may lead to transition either in isolation and/or in concert with one another. Furthermore, the development of machine-learning-based stability models as an effective surrogate for the direct computation of instability characteristics will be outlined. Several shortcomings of popular transport-equations-based transition models have also been identified in the course of this work. These include the lack of adequate physics to capture the cumulative history of disturbance