关于bounce-back

vims

bounce-back有两种类型，一种是墙在边界上的，一阶精度，另一种是墙在流体点和最外层点之间的，二阶精度。在succi的书中称第一种是on grid，第二种是mid grid。但是在succi书中84页有行小字是说on grid也可以达到二阶精度，请问这是为什么？谢谢
还有，平衡分布函数是二阶精度，是因为忽略三阶速度项由maxwell分布展开得到的。那么文献里所说的bounce-back具有一阶精度或者二阶精度是怎么得来的？谢谢

[ 本帖最后由 vims 于 2010-7-29 09:22 编辑 ]

ywang

原帖由 vims 于 2010-7-29 07:53 发表

                               
登录/注册后可看大图

bounce-back有两种类型，一种是墙在边界上的，一阶精度，另一种是墙在流体点和最外层点之间的，二阶精度。在succi的书中称第一种是on grid，第二种是mid grid。但是在succi书中84页有行小字是说on grid也可以达到二阶 ...

通常说on grid是一阶，，，

vims

谢谢

[ 本帖最后由 vims 于 2010-8-3 02:19 编辑 ]

xtshi

可以参看郭照立写的那本书，on grid分传统的反弹和修正反弹，修正反弹可以达到二阶精度；mid grid是二阶精度的；

luo@odu.edu

Bounce-back (BB) boundary conditions (BCs) are simplest, most popular implementation of Dirichlet boundary conditions in LBE, and yet is the most misunderstood ones, too. You can solve the channel flow analytically to find out what is the problem. This problem has been long solved!

The real culprit is the lattice BGK model coupled with the BB-BCs. The problem can be completely overcome by the MRT LBE.

vims

Thank you for your reply. I will check it by channel flow.
I have a MRT-LBE code for simulating cavity flow, it is more convenient in CUDA C parallel code than SRT-LBE, and better than SRT-LBE in high Reynolds number.
But i found the SRT-LBE will be more flexible in some questions from my opinion, e.g. the curve BCs. I am just a newer in LBE, maybe my idea is not correct.

luo@odu.edu

I just don't see how the SRT-LBE (i.e., LBGK) can be MORE convenient. For curved boundary, the case was well treated by the work of Bouzidi, Firdaouss, and Lallemand. The point is: the LBGK model produces erroneous results, regardless it's convenient or not.