diff options
Diffstat (limited to 'lbm.org')
-rw-r--r-- | lbm.org | 10 |
1 files changed, 7 insertions, 3 deletions
@@ -713,7 +713,7 @@ One comparably simple model for respresenting the smaller eddies in such a LES i This model yields a expression for computing the /effective relaxation rate/ $\tau_\text{eff}$ on a per-cell basis given the global relaxation time $\tau$ and a Smagorinsky constant. As the relaxation time in BGK LBM is a function of the viscosity this translates into computing the effective viscosity using a local strain-rate tensor reconstruction based on the non-equilibrium part of each -cell's populations. +cell's populations. This follows the approach laid out by Yu et al. in cite:yuDNSDecayingIsotropic2005. #+BEGIN_SRC python :session :results none tau, smagorinsky = symbols('tau, smagorinsky') @@ -1068,8 +1068,10 @@ __device__ static void apply(descriptor::D3Q19, S f_curr[19], S f_next[19], std: <<bouzidi-config>> #+END_SRC -Given a precomputed distance factor =q= we can compute an interpolated bounce back -boundary. +Following the approach by Bouzidi et al. cite:bouzidiMomentumTransferBoltzmannlattice2001 +an improved version of plain bounce back can be formulated using the distance between cell and wall. +This /interpolated/ bounce back condition reconstructs the missing populations using a basic linear +interpolation w.r.t. a precomputed wall distance factor $q$. $$\begin{align*} f_i(x_f,t+\delta t) &= 2q f_j(x_f,t) + (1-2q) f_j(x_{f} + \delta x \xi_i,t) && q \leq \frac{1}{2} \\ @@ -5844,6 +5846,8 @@ done | 4 | 240 | 1000 | 2561.12 | | 4 | 256 | 1000 | 2511.97 | +* References +#+BIBLIOGRAPHY: sources acm option:-dl option:-nobibsource * Open tasks :properties: :unnumbered: notoc |