自然出口如何设置边界条件

river.hu

请问LBM中如何设置类似fluent中outFlow这种出口条件

举例：一段水平管，左边端面入口，只在距离圆心0.5倍半径范围内Vx=1, Vy=0,

右边端面出口，是流体自然发展的流动，请问如何设置边界？

是否必须右边设置Vx=0.25?理论上如果长度不够，流动还未充分发展，右边端面应该是不均匀的

请问大牛该如何设置？Re<1000

nanosmile

充分发展流动
你可以试试 f[NX][j][k]=f[NX-1][j][k]

river.hu

谢谢nanosmile的指点
有没有相关书籍或者文献说明这么做的合理性？

血色蔷薇

第一点，看你的意思是想用速度进口条件...我个人认为，你可以考虑周期边界条件。相关的理论可以参见西交大，何雅玲或者华中科技，郭照立有关格子BOLTZMANN方法的书。如果我没记错的话，何雅玲的书应该在第六章。
第二点，我觉得你在出口边界是不可能再设置一个速度边界条件的，对于一个管道而言，进口速度给定了，要是再设置出口速度的话很有可能会造成质量或者动量的不守恒。这一点是与压力边界条件不同的。而周期边界条件正是为了保证质量守恒而提出的...fluent里面OUTFLOW也不需要设置出口速度。
最后一点，用LBM方法的话Re都是很小的，你要是能吧Re算到100以上都很不容易了...我还没在文献里见过...何况1000...

[ 本帖最后由 血色蔷薇 于 2011-4-13 18:24 编辑 ]

luo@odu.edu

Why "用LBM方法的话Re都是很小的"?! The key is the "grid Reynolds number"

Re* = U \delta_x / \nu = U / \nu

where \delta_x is the grid spacing and is assume to be 1. Thus, if you want to have large Re, you need a large mesh. There are plenty of work using LBE to simulate high-Re flows.

nanosmile

我记得在哪看到过，不过现在一下还想不出来。
我也是初学者，说的不一定对，
我的理解是：f[NX][j][k]=f[NX-1][j][k]出口条件，认为流动已经充分发展，出口处的速度流型已经稳定，不会再随时间改变。就像Poiseuille 流动的结论一样，充分发展后，任何截面的速度分布都是一样的。
哈哈，仅供参考啊~~

river.hu

谢谢罗老师重要的指点，我再多读点文献，bow

river.hu

我看了何老师的书，充分发展可以用速度更新方法
U[Nx][j]=U[Nx-1][j]; 再根据U[Nx][J]计算Feq，还是利用非平衡外推法
结果看起来是正确的，出口中间速度大，两边速度小点
直接用f[NX][j][k]=f[NX-1][j][k]算出的流场不对称，也许是我程序不对

血色蔷薇

this is a privilege I am lucky to have, thank you for your tutoring Prof. Luo.

luo@odu.edu

It would be a great pleasure if I can help. But sometimes I felt powerless.

There are several things one needs to understand before get into, say, the LB method. For those in this discussion board, some basic knowledge of fluid mechanics/physics is necessary, and then some basics of computational and numerical methods. Only then, we can talk about "computational fluid dynamics" or CFD. The LB method is just a specific technique of CFD.

What's unusual about the LBE is the fact that it is derived from kinetic equation, as opposed to the conventional CFD methods based on direct discretization of the Navier-Stokes equations. Thus, to master the LB method, some basic knowledge of kinetic theory becomes imperative.

BTW, when I responded to your comment about the Reynolds number, the premise is that we are doing direct numerical simulations (DNS), i.e., no turbulence modeling of any sort is involved. Otherwise my response is meaningless.

[ 本帖最后由 luo@odu.edu 于 2011-4-17 08:34 编辑 ]

luo@odu.edu

I have been hesitating to respond to this kind of question, for I don't know where to start. First, I am not a "big shot" (I suppose that what "大牛" means). However, I sort of understand the questions. I say "sort of" because the question is not a clean one.

In stead of giving a straightforward response. Allow me to ask some questions. First and foremost, why the boundary conditions (BCs) have to be Reynolds number dependent? Isn't that implied in the condition of "Re < 1000"?

Suppose you use Fluent, what are the BCs used in it? Are they "physical" or mathematically consistent? If the answer is yes to the above, then it is easy to convert whatever you can do with Fluent to the LBE equivalence. The BCs in velocity or pressure can be easily specified by the equilibrium distributions.

It is less easy to specify the nonequilibrium part. However, this can be dealt with.

[ 本帖最后由 luo@odu.edu 于 2011-4-16 10:02 编辑 ]

血色蔷薇

I agree with you Prof.Luo and I have noticed the significance of some basic knowledge, for example, the Statistical mechanics.

However, in china, or at least in my college,  there still are many graduates who just can "use" commercial softs, like fluent and NUMECA. I mean that they ignored basic knowledge of fluid mechanics more or less. And I do not want to like that !

luo@odu.edu

Please reply in Chinese if that makes easier for you -- I still can read Chinese. It's just that my system does not allow me to type in Chinese easily.

[ 本帖最后由 luo@odu.edu 于 2011-4-17 22:07 编辑 ]

血色蔷薇

好的