LBM的CFL条件

onesupeng

原帖由 lwd1981 于 2012-9-7 14:06 发表

                               
登录/注册后可看大图

"记得我问你在哪儿看到王竹溪先生的“统计热力学”了吗？你回答了我的问题，就多知道一点关于王先生的东西。所以回头去审视我的问题，对所有人都会有帮助"

       样子，你认为我并没有看过王竹溪先生的热力学和统 ...

不要说气话嘛~

通流

吵得挺火。吵架不是坏事，只是大家不要把这些当成对个人的攻击就行。也就是，即使有些话说的比较偏激，过火，也尽量当成只是对事情的态度。也许这才是和谐的意思。和谐不是掏浆糊，更应该是对事不对人的争论，或者说吵架。
其实，就是吵架，也是一种交流的方式。鲁迅的很多杂文就是吵架骂人。

onesupeng

原帖由 通流 于 2012-9-7 14:21 发表

                               
登录/注册后可看大图

吵得挺火。吵架不是坏事，只是大家不要把这些当成对个人的攻击就行。也就是，即使有些话说的比较偏激，过火，也尽量当成只是对事情的态度。也许这才是和谐的意思。和谐不是掏浆糊，更应该是对事不对人的争论，或者说 ...

哈哈，通流也过来凑热闹了

其实我也就多事，多问了几个问题而已。目的也是为了让大家搞清楚事情的原委而已。lwd1981比较认真，有问必答，所以问题多了，书翻多了有点急躁，慢慢来

onesupeng

原帖由 通流 于 2012-9-7 14:21 发表

                               
登录/注册后可看大图

吵得挺火。吵架不是坏事，只是大家不要把这些当成对个人的攻击就行。也就是，即使有些话说的比较偏激，过火，也尽量当成只是对事情的态度。也许这才是和谐的意思。和谐不是掏浆糊，更应该是对事不对人的争论，或者说 ...

你也就我提出的问题看看有没有可取的地方吧，判断下会不会是我胡闹呢？

zqb138

luo@odu.edu

Dear Mr./Ms. WDLXMZD,

I cannot disagree with you more about your comments concerning the CFL number in the context of LBE. With a fixed resolution and a given algorithm, we all wish to use the largest CFL number possible to attain best possible computational efficiency. And that's the essence of the CFL number -- computational efficiency. Your statement that "the CFL number is no use in standard lb" simply means that computational efficiency is of no concern to the "standard LB", which could not be further from the truth. I am sometimes quite puzzled by this kind of statements of yours -- their falsehood is in proportion to their definitiveness and firmness when they are asserted. I only hope that you did not mean it. Perhaps you could clarify it in Chinese.

Concerning your question regarding the MRT , clearly my message was not going through. In the recent work your were referring to, we demonstrated that the stability is affected by ALL those relaxation rates, some more severely than others. However, we also stressed that one ought not to look at the stability in isolation, and one must consider ACCURACY, STABILITY, and COMPUTATIONAL EFFICIENCY together, and in that order. In this way, we concluded that the two-relaxation-time (TRT) model is perhaps the best. There is a theory behind this, which has been published by d'Humieres and Ginzburg.

wdlxmzd

谢谢罗老师的回复，前些天在医院住院，不能用中文，不好意思。
我的观点其实很简单，LBM的稳定性等问题远比传统方法复杂，有太多的因素在影响，所以CFL数就不能在LBM中具有其在传统算法中的地位。传统的算法里面，CFL基本上与选择时间步长DT的话题有关。

标准LBM里面的CFL数，有的定义为1，有的定义为最大马赫数。无论是哪个，这里面都没有选择时间步长DT的事。

按照郭老师的理解，LBM为ian explicit time-marching method for compressible fluid flows。如果从计算效率出发，我对于以密度等变量为求解量的可压缩算法的最简单理解就是，其对声速和流速在一个级别上的流动是很有效的。对于低马赫数流动，流速远小于声速，可压缩算法的效率就很成问题。

针对稳态流动的一些LBM预处理技术，可以看作是通过减少有效声速来提高计算效率，减少了声速和流速的差距，但不知此时马赫数是否变化。是否应该采用有效马赫数来计量。

非常感谢您对MRT的松弛时间的释惑。LBM里面仍然有很多东西暂时无法解释，或者还没有发现相应理论的东西，期待各位大师指引。当然如果能发现这些松弛因子，为什么要这样选取，也许对MRT是很大进步。现在对某些松弛因子可以按照规则去选择，但其他的似乎很多时候就是摸索选取，对某一模拟问题，1.5更好，对另一问题，1.1又更好, no theory。

至于TRT，我也一直很喜欢，不过似乎稳定性也并不能得到很大提高。不知罗老师对cascade LBM怎么看，我对这个不太懂。

luo@odu.edu

1) "我的观点其实很简单，LBM的稳定性等问题远比传统方法复杂，有太多的因素在影响，所以CFL数就不能在LBM中具有其在传统算法中的地位。传统的算法里面，CFL基本上与选择时间步长DT的话题有关。"

Your comment is erroneous, again. For any CFD solvers with NO exception, CFL number determines the time step size. In LBE, you can artificially set dt = 1, but that does not mean anything -- you must re-normalize it with for example, Re, so any different methods/schemes can be compared. For example, if you use the LBE to compute a steady flow, the CFL number will decide how many iterations you need. If you use the LBE for an unsteady flow, then there is a time-stepping issue.

Other comments of yours are off the mark, too.

2) "LBM里面仍然有很多东西暂时无法解释，或者还没有发现相应理论的东西，期待各位大师指引。当然如果能发现这些松弛因子，为什么要这样选取，也许对MRT是很大进步。现在对某些松弛因子可以按照规则去选择，但其他的似乎很多时候就是摸索选取，对某一模拟问题，1.5更好，对另一问题，1.1又更好, no theory。"

There is a vast wealth of kinetic theory. Most people in the LBE business do not understand kinetic theory, nor do they care. You can educate yourself by study kinetic theory more seriously. The problem is that most, if not all, theories are based on linear analysis; and the Navier-Stokes equations are nonlinear ones, thus we cannot expect an one-size-fits-all theory. This is expected. We debated this issue previously (theory or no theory), and let's not waste more time on this.

Perhaps you could solicit those experts you recommended to this group to shut some lights on this issue -- let's see what those bigshots have to say.

3)  "不知罗老师对cascade LBM怎么看，我对这个不太懂。"

The so-called cascade LBE uses a NONLINEAR collision model, as opposed to the linear one in MRT. The cascade LBE is more stable because it is more dissipative.

[ 本帖最后由 luo@odu.edu 于 2012-9-10 09:37 编辑 ]

wdlxmzd

感谢回复，请恕我要不敬了，我说的THEORY，不是仅仅kinetic theory，能读的书大家都能读得到，谁也不是个蠢蛋。

我这个话题本来只是针对MRT里面松弛因子的选取，有些松弛因子需要摸索选取，没有相应理论解释和数学解释.

我本来说的就是这么个简单的意思。

还有，您的论文里面既然是用错了ELBM，请问您是否修改过，按照KARLIN的方式重算一遍？

关于标准LBM的CFL数问题，在算稳态流动时，提高马赫数，计算效率会提高，但是请问，这是因为提高了CFL数，还是因为作为一种可压缩算法，缩小了声速和流速差距而带来的计算效率的增加，请问这点怎么判断？

最后一点，中国学生不笨。我记得有一次，您说一个MRT的9X9的矩阵，对13亿中国人就这么难吗？这样的话极具侮辱性。
实际上，中国学生造了13X13,16X16,17X17的MRT矩阵，也许在您看来这些没有用，但是在我看来，这说明中国学生不笨，而且很聪明。

不会笨到连DT=1怎么怎么地也不知道。 说句实在话，中国学者和部分中国学生做得丝毫不比您差。

[ 本帖最后由 wdlxmzd 于 2013-7-31 10:35 编辑 ]

luo@odu.edu

1) "还有，您的论文里面既然是用错了ELBM，请问您是否修改过，按照KARLIN的方式重算一遍？"

"任何人都可能犯错误，您对ELBM的错误应用，在我看来就是个低级错误，你要比，也要按照人家原来的比，就算最后比出来它还是有什么问题。"

I have yet to see what mistakes I may have made. If there were any mistakes, it was on the part of Karlin et al. The "ELBE" which we CLEARLY and EXPLICITLY defined in our paper was indeed proposed by Karlin et al., and I explained why we call the model "ELBE" for it is supposed to possess an H-theorem and associated entropic properties. So, please be specific about what mistakes I have made. I will address ALL the issues raised by Karlin et al. in my Reply, which will be published in PRE.

2) "关于标准LBM的CFL数问题，在算稳态流动时，提高马赫数，计算效率会提高，但是请问，这是因为提高了CFL数，还是因为作为一种可压缩算法，缩小了声速和流速差距而带来的计算效率的增加，请问这点怎么判断？"

Are you saying that for a very low-speed flow, one can indeed bring down the speed of sound to improve the computational efficiency?! That would be a miracle indeed!

Your comments regarding CFL are simply erroneous. I don't see how you can defend them. The more you say, the more you reveal yourself -- I will let you figure out the problem yourself. Just a hint: even for a steady-state problem, the time in the LBE is still "physical," if you know what that means.

3) "最后一点，中国学生不笨。我记得有一次，您说一个MRT的9X9的矩阵，对13亿中国人就这么难吗？这样的话极具侮辱性。"

Whether my words were insulting or not, my audience at large can decide.

"实际上，中国学生造了13X13,16X16,17X17的MRT矩阵，也许在您看来这些没有用，但是在我看来，这说明中国学生不笨，而且很聪明。"

Really?! Which work of "Chinese students" or yours is indeed ground breaking in the field of the LBE?! Show me, just one.

"不会笨到连DT=1怎么怎么地也不知道。 说句实在话，中国学者和部分中国学生做得丝毫不比您差，千万不要以为别人都是笨蛋。"

Well, stupidity has no ethnicity, but it does have individuality. I never thought myself as a smart guy or I will not stuck here, and that's why I don't take myself too seriously.

wdlxmzd

1）：错了就是错了，要是没错，你回复个REPLY 给KARLIN的COMMENT就是了，这是期刊允许的，不用回复给我，。

2) ：我没有说这个，也不是我要去缩小声速，是预处理LBM里面的有效声速实质就是减少的，你自己去看那些论文。

3）：您在反问中国学生有没有开创性的工作的时候，应该反问您自己？您的呢，自己回去梳理下自己第一作者的论文，非第一的论文，那些钱导师的论文，HE XIAO YI的论文， YU D Z的综述论文就不要拿来说事了，从您第一作者的论文看，您是一个非理想气体和BINARY FLUIDS的LBM专家，可是这个方面您的模型根本没有得到实际的应用，可以说好多甚至连数值实施都没有，OVER？

至于现如今中国学生的工作，由于LBM发展到现在，大部分能引起轰动的工作都被前人做了，我所看到的就是中国学生在踏踏实实一点一点地做工作，那么您这几年有开创性，轰动性的成果吗？ 不要跟我说您最近那篇论文，论文是很不错，可是是开创性的工作吗？

如果您这样一个号称大师的人，都没有开创性的工作，要求什么中国学生有，一个MRT松弛时间的选取，钱导师之后10年多，您都没有推进过理论解释？绕来绕去，有啥好绕的，一会1.5，一会1.1，没有理论解释就没有理论解释， OVER？

等您有一天发现了理论解释，再回来嘲笑我们这群人吧。

[ 本帖最后由 wdlxmzd 于 2012-9-10 15:13 编辑 ]

luo@odu.edu

1）：错了就是错了，要是没错，你回复个REPLY 给KARLIN的COMMENT就是了，这是期刊允许的，不用回复给我，。

You claimed I made mistakes, so I demand you explicitly tell the public what mistakes I made.

2) ：我没有说这个，也不是我要去缩小声速，是预处理LBM里面的有效声速实质就是减少的，你自己去看那些论文。

OK, you are kidding yourself again.

3）：您在反问中国学生有没有开创性的工作的时候，应该反问您自己？您的呢，自己回去梳理下自己第一作者的论文，非第一的论文，那些钱导师的论文，HE XIAO YI的论文， YU D Z的综述论文就不要拿来说事了，从您第一作者的论文看，您是一个非理想气体和BINARY FLUIDS的LBM专家，可是这个方面您的模型根本没有得到实际的应用，可以说好多甚至连数值实施都没有，OVER？

Ah ha, I obviously do not understand what you meant by "中国学生" within the context -- I did not know I am considered as one of "中国学生", what an honor!

至于现如今中国学生的工作，由于LBM发展到现在，大部分能引起轰动的工作都被前人做了，我所看到的就是中国学生在踏踏实实一点一点地做工作，那么您这几年有开创性，轰动性的成果吗？ 不要跟我说您最近那篇论文，论文是很不错，可是是开创性的工作吗？

You think important problems in the LBE are solved. Well, I think that only the easy ones have been done. There are still quite a few tough nuts left to be cracked. It's a pity that not too many smart "中国学生" take a shot at these problems.

Have I ever made any claim that my work is ground-breaking?! To the contrary, I always thought that my work is only second-rate, at the best. So I am in total agreement with you that I haven't done anything ground-breaking.

如果您这样一个号称大师的人，都没有开创性的工作，要求什么中国学生有，一个MRT松弛时间的选取，钱导师之后10年多，您都没有推进过理论解释？绕来绕去，有啥好绕的，一会1.5，一会1.1，没有理论解释就没有理论解释， OVER？

I am not a bigshot and I don't think of myself as one. But, I have a habit to shot at bigshots. Since you talk as authoritatively  a bigshot, so I like to pop a shot once a while. I hope you don't mind.

I used some examples just to illustrate the effects due to the relaxation rates of higher-order moments. You therefore concluded that there is NO theory. Of course this is OVER for me. What do you expect?

等您有一天发现了理论解释，再回来嘲笑我们这群人吧。

Theory is already there -- quasi-conserved modes and separation or non-separation of scales in kinetic theory. They are textbook materials, there is no need to re-discover them, really. All you need to do is to read, and read carefully.

wdlxmzd

1）；是KARLIN COMMENT了你，你应该去回复他，KARLIN也已经告诉你，你错在哪；

2）：你睁开你的眼睛去看看Z.L. Guo, T.S. Zhao, Y. Shi, 2004, “Preconditioned lattice-Boltzmann method for steady flows”, Physical Review E, 70, 066706 (2004).
论文公式13后面的那段话，是不是写着有效声速，是不是会减少？我只是转述我看到的。

3）唉，估计您在美国时间长，国语都看不懂了，我说反问中国学生的时候，你应该反问你自己，因为你自己都没有，何来要求中国学生。我很惊讶你把这句话理解为你被囊括到中国学生里面了，这表明，你应该要好好复习下中文了；

您如果认为只有些简单的东西被做了，那好嘛，你自己列下您这8年内，从2005年来的第一作者的论文，看看自己又解决了哪些？ 8年应该够了吧，因为抗战都可以结束了。

我没有EXPECT什么东西，我就一句话，一会1.5好，一会1.1好，有没有解释，有，给个解释就行，不要绕，不要把其他的扯进来。

对了上面的1.5，1.1是泛指，不要对号入座。

[ 本帖最后由 wdlxmzd 于 2012-9-10 16:42 编辑 ]

wdlxmzd

我尊重很多学者，因为这些学者是用行动在说话，无声的言语，用行动感染后辈，尊重敬仰那些70多还在科研前线，40-50岁还自己编程序的中国学者。

[ 本帖最后由 wdlxmzd 于 2012-9-10 16:49 编辑 ]

luo@odu.edu

1）是KARLIN COMMENT了你，你应该去回复他，KARLIN也已经告诉你，你错在哪；

You are the one bring the issue into this discussion, although it has nothing to do with CFL. Now, you just want to be someone else's mouth-piece, right? What is between your ears?

2）：你睁开你的眼睛去看看Z.L. Guo, T.S. Zhao, Y. Shi, 2004, “Preconditioned lattice-Boltzmann method for steady flows”, Physical Review E, 70, 066706 (2004).
论文公式13后面的那段话，是不是写着有效声速，是不是会减少？我只是转述我看到的。

Never mind who did and/or said what. What is your own thoughts? Are you telling me that just someone published something, therefore it is THE TRUTH?

I repeat myself again:  Your comments concerning CFL number in the context of the LBE are ludicrous. It was your statements in this discussion that I commended. Defend your own statements -- "LBM的稳定性等问题远比传统方法复杂，有太多的因素在影响，所以CFL数就不能在LBM中具有其在传统算法中的地位。传统的算法里面，CFL基本上与选择时间步长DT的话题有关。"  Please explain to the public why it is so -- quoting from someone else's paper won't help you.

3）唉，估计您在美国时间长，国语都看不懂了，我说反问中国学生的时候，你应该反问你自己，因为你自己都没有，何来要求中国学生。我很惊讶你把这句话理解为你被囊括到中国学生里面了，这表明，你应该要好好复习下中文了；

True, true, all true -- I ought to go back to kindergarten.

您如果认为只有些简单的东西被做了，那好嘛，你自己列下您这8年内，从2005年来的第一作者的论文，看看自己又解决了哪些？ 8年应该够了吧，因为抗战都可以结束了。

Well, precisely because I have not done anything worth talking about, that's why I do not take myself so seriously. I have said this repeatedly and say it once again here. So, why asking me? How about those bigshots you were talking about? In any event, that does not stop me asking questions, or potting some shots at some "bigshots." Since you are such an authority on LBE, CFD, and more, why don't you just show us what those bigshots of yours have done, and let's open a discussion and have some fun here.

Amuse us, please.

我没有EXPECT什么东西，我就一句话，一会1.5好，一会1.1好，有没有解释，有，给个解释就行，不要绕，不要把其他的扯进来。

对了上面的1.5，1.1是泛指，不要对号入座。

Hmmm, 1.5 or 1.1, that's the question. It is indeed perplexing.

[ 本帖最后由 luo@odu.edu 于 2012-9-13 00:40 编辑 ]