[arpiruk]

Hi thanhphatvt,

I came here to do a similar thing you wanted to do. I’ll guess I can help you out a bit.

I did a DNS with my fourth-order FVM code using more ore less the same grid as the reference. I guess you use Lereiche’s result. I can tell you that for RE18K (mean RE=15K), you need a very long .. long .. long .. long integration time. By long, I mean the sampling time > 1500 D/U_L , for D being cavity width(&depth) and U_L being lid’s velocity. Also, Lereiche result in his DNS paper is about 700 D/U_L for RE15K, and it is only 500D/U_L for Re10K. So keep that in mind.

There is a large scale movement which is not averaged out easily. For RE=10K, you’ll be fine with about 1,000 D/U_L. But, a big but, the difference btw a statistics of one 1000 samples (sampling every time unit), and another set of the 1000 can be a bout 3-5%.

Do not expect a perfect collapsing plot like in turbulent channel flow.

[Author]

Posts