[jiaodanuma]

Hi, all

I want to simulate the flow filed around a cube in a uniform flow using openfoam 2.2.1. However, I can not achieve a statistically steady flow so far. The streamline of mean velocity behind the cube looks always instantaneous without steady recirculation regions, and I've run the case long enough. The predicted acoustic characters in the far field based on the calculated flow field is quite different from the experimental results.

The Reynolds number is 2*10^5, the maximum y+ value is less than 1, the CFL number is less than 2. The values of drag coefficient and lift coefficient turn to be statistically steady much more easily, while it is not the case of the velocity field.

I used the structure mesh generated by pointwise, the problem is that the maximum skewness values (about 2.5 reported by checkMesh and about 0.5 reported by pointwise) are around the cube, no matter how I adjust the topology. I'm not sure if this is the true reason for the unsteady velocity field or due to something inherent in openfoam.

One of my colleagues ran cases of square cylinders with infinite length in openfoam 2.3.0, everything is perfect. Both of us use DDES model and we use the same numerical schemes (div(phi,U) Gauss LUST unlimitedGrad(U)). I tried to change the numerical scheme to be more upwind, the problem of mean velocity field didn't be solved. I tried to reduce the Reynolds number by decreasing the flow speed. when the Reynolds number is 5*10^4, the mean velocity field seems good after about 10 round trips in the computational domain, however, after about 13.2 round trips the mean velocity field behind the cube began to lose the steady status. For Reynolds number lager than 5*10^4, the steady velocity field as expected is difficult to be observed. I ran the same case in Fluent using the same structure mesh at Reynolds number of 2*10^5. Steady mean velocity field can be obtained but the acoustic thing is still not good.

The most important thing that worries me much is that I just try different ways blindly, since I don't know how to analyse the above phenomenon to find out the true reason of the problem. Since CFD work is based on a huge system, any minor mistakes in each link could lead to critical errors. There are thousands of possibilities for the incorrect result, how can I narrow them down and eventually find out the true reason? Anyone can give me some suggestions? Thank you very very much! At present, the most suspicious reason is the high skewness around the cube. I'm trying to use the unstructure mesh such as SnappyhexMesh and trimmed mesh, but there are also some problems. I'm looking forward to hearing suggestions from you. I appreciate it very much!

Best wishes