小雷诺数二维圆柱绕流问题

至尊仙

拜托能不能把那篇文章找出来，翻译下，我想知道他的结果，我更想知道他得到这些结果的方法！！！！！！！！！！
我看到过的实验曲线写的d都到了10的-3mm以下，说实在的，我不太相信那个年代能把这个数量级的实验做到精确。而其他的图都只标了个Re，根本不知道他们的实验参数，这样的结果居然成了你们的供奉，我真不知道诸位高人们治学严谨的态度哪去了。难道仅仅是因为那些人是高人们的老师？

[ 本帖最后由 至尊仙 于 2013-7-25 23:29 编辑 ]

onesupeng

原帖由 至尊仙 于 2013-7-25 14:42 发表

                               
登录/注册后可看大图

拜托能不能把那篇文章找出来，翻译下，我想知道他的结果，我更想知道他得到这些结果的方法！！！！！！！！！！
我看到过的实验曲线写的d都到了10的-3mm以下，说实在的，我不太相信那个年代能把这个数量级的实验做到 ...

200人民币/千字我帮你翻译。。。

通流

我是特别同意物理的重要性。不管是搞算法，或者用别人的程序，首先要有点流体力学的知识。这个网站里讨论的最多的大概是网格。一个基本的考虑是：网格是为了计算流场的。不清楚流场，何来高质量的网格？

至尊仙

可以，上帐号，不过只要你给翻译他的实验参数和结果，其他不要你翻译。

至尊仙

上账号啊，我说打钱给你就打给你。
网上我都消除不了你这样人的疑议，当面讲解完全没必要。

onesupeng

原帖由 至尊仙 于 2013-7-27 00:54 发表

                               
登录/注册后可看大图

上账号啊，我说打钱给你就打给你。
网上我都消除不了你这样人的疑议，当面讲解完全没必要。

我又突然觉得这个项目太小，不划算接。你看谁有兴趣接来做做

至尊仙

你就是个放屁都怕被人听见的人！！！！！
不过我还是要感谢你，没有你摩擦力公式推不出来，《流体悬浮XXXX》一文也不会有今天的成就，所以你放心，不管你认不认，这篇文章我都会加上你的名字。你不告诉真名，我只好加上onesupeng了。

[ 本帖最后由 至尊仙 于 2013-7-27 15:20 编辑 ]

onesupeng

原帖由 至尊仙 于 2013-7-27 05:07 发表

                               
登录/注册后可看大图

你就是个放屁都怕被人听见的人！！！！！
不过我还是要感谢你，没有你摩擦力公式推不出来，《流体悬浮XXXX》一文也不会有今天的成就，所以你放心，不管你认不认，这篇文章我都会加上你的名字。你不告诉真名，我只好 ...

放屁最好不要让别人听见了，难道大街上撅屁股了？

别，不要在你任何东西里面加我名字~

至尊仙

是你自己不要的，以后别埋怨我。

luo@odu.edu

Dear Onesupeng,

First, thank you very much for your illuminating discussion. I do enjoy reading it. However, this discussion also shows us that we have a huge problem in even talking about a specific, small problem.

Let's focus on the word "comparison" for now. It is absolutely essential that a COMPARISON is done with the EXACTLY identical flow conditions. Else it is should not be called a COMPARISON, certainly within the context of CFD. Thus one must be extremely careful when "compare" CFD with experimental results. In any event, (rigorous) comparison is an indispensable tool in CFD, and it appears that many still do not what it means.

Next, the Stokes flow, which consists of 2 equations: the (linear) momentum equation + the incompressibility condition. It is known that the solution for, say, a 2D cylinder in free space decays like (1/r), where r is the radial distant from the cylinder center. This means that the effect due to the (far field) boundary is also behaves like (1/r), thus decays too slowly. That explains why "blockage" matters so much in the flow, and this is not entirely based on just experience -- people in applied math. or PDEs would tell us they knew it all along. With this in mind, to solve the flow past a cylinder with very small Re, we need not only a good discretization scheme, but also an effective way to treat boundary conditions. More over, to enhance computational efficiency, we also need acceleration techniques (e.g., multigrid), which is not discussed here.

Within the context, the work of Labrosse & Lallemand show that LBE is a good second-order discretization scheme, while the work by Krafczyk and his student show that grid-refinement is necessary for the problem and it is doable with LBE. (Unfortunately, Krafczyk's work has never been published, it's in a student's thesis.) I apologize that my previous responses were not carefully qualified.

It is important to realize that one should not expect to use the so-called lattice BGK model (which is very silly scheme, by the way) with a uniform mesh and explicit time-stepping to solve all possible problems in CFD. However, some claim the contrary and they believe in magics, as we have seen on the forum. It is sad to see that many students are trying just to do that. My question is whether these students are pushed to do so, and they are simply lack of proper education. I only hope that our discussions would help these students.

Finally, I agree with your comments about benchmarks. A set of WELL-DEFINED benchmark problems are essential in CFD. Students must be taught how to conduct comparative studies. Comparison with experimental data is a much harder subject, which I shall say nothing.

[ 本帖最后由 luo@odu.edu 于 2013-7-28 22:22 编辑 ]

luo@odu.edu

Absolutely --- numerics obeys physics, NOT the other way around.

onesupeng

原帖由 luo@odu.edu 于 2013-7-28 13:51 发表

                               
登录/注册后可看大图

Dear Onesupeng,

First, thank you very much for your illuminating discussion. I do enjoy reading it. However, this discussion also shows us that we have a huge problem in even talking about a specif ...

非常感谢罗老师回复这么多，大部分观点都同意你的看法。

我这里讲一些有趣的现象，无论中国还是美国的学者都存在的。

第一个现象，还是拿圆柱绕流来说。我发现前几年的结果，尤其是用FVM和贴体网格的结果，阻力略小；而近几年用IBM、LBM等结算的，阻力略大。目前人们计算的结果，如果更大一些，那么大家会引用结果偏大的文献。久而久之，发现阻力报道的越来越大。例如原来大家算的Re=100，Cd大约是1.38，后来有人算到1.42,1.44,1.50等，表面上大家误差都不大，但回头一看最大的和最小的差异很大。。。

第二个现象，我们看到的结果都很漂亮，结果过几年作者本人说原来的方法不好，计算的生阻力振荡（oscillation）比较厉害，现在新方法抑制了这个振荡，可是我们却没有看到当年他的结果有什么振荡。具体一点，比如阿猫2008年发表一篇JCP文章，结果很漂亮，圆柱绕流、圆球绕流等报道很多算例，结果很漂亮，曲线很光滑。到2013年阿猫有发表了一篇JCP，说改进了08年的算法，曲线更光滑了，然后他给出一张图片，两条曲线，一条是08年方法算的（很不光滑），一条是现在算的（很光滑）。然后的问题是：他为什么08年的文章那么光滑，文章也没有说做了一定的光滑处理，而到13年却那么的不光滑。

还有一个现象就是，貌似大家都拿自己结果和牛人的结果比，哪怕自己算不出来，也要怎么凑一个数据。我最近发现有个牛校的牛人在流体top杂志上的一个文章，我很验证不上，然后我打电话问几个做同领域的人，他们也说验证不上，计算结果比不上。但偏偏有那么几个JCP文章“比得上”，后来我研究了一下，这个问题是个定常问题，有点类似小Re（例如Re=20）圆柱绕流，圆柱阻力随着迭代增加而减小，最后到达一个稳定值。大家结果都比文献小，然后几个JCP取了还没有到达稳定值的一个值，这不就比上了么。。。

上面讲的是很普遍的，都是中外都存在的现象，如果你也发现文献有这些现象，或者更多这种类似不端行为，欢迎讨论

onesupeng

有几个错别字不影响阅读就不回头更改了。。。哈哈

luo@odu.edu

Dear Onesupeng,

Again, appreciate very much for the discussion. I had some similar observation, but definitely not as acutely as you have done. I usually don't have any "faith" on any numerical (or experimental, for that matter) results, unless I know the author(s) and the way they work. And I do have a black list of people whose results I do not trust no matter what.

When I write a paper, I insist that the paper MUST disclose ANY possible details and must NOT leave any room of ambiguity. Thus, the results can be easily reproduced by others if they wish to.

The issue mentioned in your comment (matching or fixing results) is indeed a serious problem. And it is particularly so in "complicated" problems, which are hard to repeated (e.g., for stringent requirement of computational sources, etc.).

I assume you know the card game "Blackjack". The way these people play the game is that they want to look at their hands first before they bet -- of course they always win. My point is that if CFD simulation has any capability to predict at all, it MUST be done in such a way that you pretend the answer is not there. It's good we quote results which agree with ours in the ballpark. However, it is far more important to understand why there are differences. If we cannot afford to dig into other's results, we should at least honestly report our own results as they are -- regardless of the differences with others.

I do not know if those results you mentioned are worth to be commended. If you do so, it would be a great public service to the community. But you may risk to piss off some people and even become a public enemy (like me).

Best,

-- LSL

[ 本帖最后由 luo@odu.edu 于 2013-7-30 06:25 编辑 ]

通流

onesupeng对不少“大牛”很不感冒。按照我的理解，真正的大牛的一个必要条件是“诚实”。在我的领域，也就是叶轮机械领域，CFD的大牛之一是剑桥大学的Denton。其实Denton的CFD用的都是一些老掉牙的东西。紊流模型是普朗特的混合长模型。不过，他的意见总是比较受到尊重。原因就是他的诚实。

我认识的一位学校的教授，他的观点之一就是同时在实验和理论计算方面工作的人，容易保持诚实。我想向前更推进一步。同时在产品设计，实验，和搞理论计算的人，更容易保持诚实。