[Anthonyyy]

Dear all:
Thanks for your valuable time for this post. I am new to CFD and I want to model the statistical stationary velocity distribution in open channel (usually with 30+ m width). Given bulk mean velocity(v_in) in fvOption and periodic boundary condition for inlet and outlet, I expect I could obtained simulated velocity distribution as I wish. But the result is somehow confusing.

Here is the three plan I used in my simulation. Mean bulk velocity is used and set to 0.8 m/s, inlet and out are set to cyclic boundary condition, the upperWall surface is treated as symmetry plane, as shown in Fig 1(model.png)

1. RANS(k-omega, k-epsilon, k-omega-SST)
   At small scale(2m width), the model works well, but explode in large scale(60m, with grid size 60 x 50 x 15 in span-width, depth-wise, stream-wise, scale modified in blockMeshDict from 1 -> 10),
   
2. RSM
   LRR is selected as simulation model. At small scale(2m width), the result is weird for suddenly increased velocity as show in red circle in Fig 2(lrr.png). Same numerical explosion when scale is changed from 1 to 10.
3. LES
   At small scale(2m width), the result is good for successful simulating the secondary current in trapezoidal channel section as shown in Fig 3(Les.png). But it is really really time-consuming.

My questions are
(1) Is my model with such mean-bulk velocity and periodic cyclic boundary condition valid for this open channel simulation?
(2) How can I simulate the case of large-scale(like 60+ or 100+ width channel section) ?