Efficient and Optimally Accurate Numerical Algorithms for Stochastic Turbulent Flow Problems

In this paper, we first propose a filter-based continuous Ensemble Eddy Viscosity (EEV) model for stochastic turbulent flow problems. We then propose a generic algorithm for a family of fully discrete, grad-div regularized, efficient ensemble parameterized schemes for this model. The linearized Implicit-Explicit (IMEX) EEV generic algorithm shares a common coefficient matrix for each realization per time-step, but with different right-hand-side vectors, which reduces the computational cost and memory requirements to the order of solving deterministic flow problems. Two family members of the proposed time-stepping algorithm are analyzed and proven to be stable. It is found that one is first-order and the other is second-order accurate in time for any stable finite element pairs. Avoiding the discrete inverse inequality, the optimal convergence of both schemes is proven rigorously for both 2D and 3D problems. For appropriately large grad-div parameters, both schemes are unconditionally stable and allow weakly divergence-free elements. Several numerical tests are given for high expected Reynolds number ($\textbf{E}[Re]$) problems. The convergence rates are verified using manufactured solutions with $\textbf{E}[Re]=10^{3},10^{4},\;\text{and}\; 10^{5}$. For various high $\textbf{E}[Re]$, the schemes are implemented on benchmark problems which includes: A 2D channel flow over a unit step problem, a non-intrusive Stochastic Collocation Method (SCM) is used to examine the performance of the schemes on a 2D Regularized Lid Driven Cavity (RLDC) problem, and a 3D RLDC problem, and found them perform well.