[Nek5000-users] proj_ortho error in pipe flow

Mon Apr 10 07:42:42 CDT 2017

ok, great!

As Paul says, ideally one wants an initial condition that quickly leads 
to turbulence, and that is done with some streamwise vorticity near the 
walls. The way we typically do that is not with an initial condition, 
but rather with our trip force, because this leads to an efficient 
vertical mixing (across the shear), and thus to quick transition. We 
then turn the force off after a short time. This way, you do not need to 
care much about continuity, and you can just have a force in the 
wall-normal direction at some selected positions in the flow. This works 
well for higher Reynolds numbers, at least from our experience.

Another thought that one could experiment a bit is to use the synthetic 
eddy method as a way to generate not just boundary conditions, but also 
initial conditions.

Philipp

On 2017-04-10 14:32, nek5000-users at lists.mcs.anl.gov wrote:
> Hi Paul & Philipp,
>
> thanks for your quick replies and suggestions. I started the simulation
> again with a slightly perturbed laminar flow field as initial conditions
> and there are no proj_ortho errors anymore.
> For now, this perturbation is more or less random and probably not ideal
> for tripping the laminar flow to turbulence. Therefore, I might try to
> implement a more effective numerical tripping method in the future. I
> can let you know when (and if) I succeed.
>
> The parameters p94 and p95 are set to "9" just because I have found that
> in another simulation somewhere. Do you suggest trying different values
> here?
> The Nek5000 version was downloaded as a zip file on 21/03/2017 so I
> suppose it is the commit 95b42f08c960d522679e0f0881d284ec841ae09
>
>
> \Steffen
>
>
> On 09/04/17 19:00, nek5000-users-request at lists.mcs.anl.gov wrote:
>> Message: 1
>> Date: Sat, 8 Apr 2017 19:55:36 +0200
>> From: nek5000-users at lists.mcs.anl.gov
>> To: nek5000-users at lists.mcs.anl.gov
>> Subject: Re: [Nek5000-users] proj_ortho error in pipe flow
>> Message-ID:
>>     <mailman.4792.1491674149.2967.nek5000-users at lists.mcs.anl.gov>
>> Content-Type: text/plain; charset=windows-1252; format=flowed
>>
>> Hi Steffen,
>> which revision of the code are you using? Is the initial field you are
>> using already turbulent, or are you going through transition? Any
>> special reason why you have 9 for p94/95?
>>
>> Philipp
>>
>> On 2017-04-07 19:54, nek5000-users at lists.mcs.anl.gov wrote:
>>> Hi Steffen,
>>>
>>> It looks like your solution is hitting a steady-state condition and
>>> that the
>>> projection onto prior solutions is thus projecting in to a set of
>>> linearly-dependent
>>> vectors.
>>>
>>> In principle, the code is supposed to correct for this condition but
>>> apparently
>>> it is not doing so correctly.
>>>
>>> I see you're running at Re > 5000, so presumably you are interested
>>> in turbulence,
>>> rather than laminar steady state.   I would recommend adding a random
>>> perturbation
>>> to the initial condition or, better still, adding something with
>>> bound vorticity near the
>>> walls that will get stretched and bring about a more rapid transition.
>>>
>>> I find that something like the attached works well with channel flow
>>> (snap through within
>>> about 10 convective time units).  I was disappointed that it didn't
>>> work as well with pipe
>>> flow, where I had to go up to Re=40k to get the thing started and
>>> then bring it down.
>>> I assume this has something to do with the lack of linear transition
>>> in Hagen-Poiseuille flow.
>>> If you find a better mechanism to speed transition, I'd be interested
>>> to find out about it.
>>>
>>> Best,
>>>
>>> Paul
>>>
>>> ________________________________________
>>> From: nek5000-users-bounces at lists.mcs.anl.gov
>>> [nek5000-users-bounces at lists.mcs.anl.gov] on behalf of
>>> nek5000-users at lists.mcs.anl.gov [nek5000-users at lists.mcs.anl.gov]
>>> Sent: Friday, April 07, 2017 10:18 AM
>>> To: nek5000-users at lists.mcs.anl.gov
>>> Subject: [Nek5000-users] proj_ortho error in pipe flow
>>>
>>> Dear Nek users,
>>>
>>> I am running a simple pipe flow on a Cray XC40 (Hazel Hen). After 32
>>> steps, I get this problem (see attachment for complete logfile):
>>>
>>> Step     32, t= 3.2000000E-02, DT= 1.0000000E-03, C=  0.025 3.2660E-01
>>> 1.0526E-02
>>>                Solving for fluid
>>>            32  Project velx           6.6444E-09   2.4759E-09
>>> 2.6836E+00   8   0
>>>            32  Hmholtz velx       1   3.9238E-09   6.7918E-07 1.0000E-08
>>>            32  Project vely           6.8730E-09   2.1046E-09
>>> 3.2656E+00   8   0
>>>            32  Hmholtz vely       1   3.6166E-09   5.9974E-07 1.0000E-08
>>>            32  Project velz           5.2396E-03   2.5121E-06
>>> 2.0858E+03   8   0
>>>            32  Hmholtz velz       3   7.6028E-10   9.3795E-04 1.0000E-08
>>>          32proj_ortho:    2   8 velz   Detect rank deficiency: NaN
>>> 1.0000E+00
>>>          32proj_ortho:    1   8 velz   Detect rank deficiency: NaN
>>> 1.0000E+00
>>>            32  U-PRES gmres      29   9.9376E-09   5.2213E-04
>>> 1.0000E-08   8.7926E-03   1.7780E-02
>>>             32  DNORM, DIVEX  2.704668739973178E-005
>>> 9.937641774554489E-009
>>>         32 0.3200000E-01  1.61568E-01  6.88948E-05  7.85330E-01
>>> 7.85399E-01 volflow Z
>>>            32  Fluid done  3.2000E-02  2.4042E-02
>>>
>>> In the next step I get CFL, CTarg! and the simulation exits.
>>>
>>> This problem does neither show up on my local machine nor on another
>>> smaller cluster. Moreover, when I set either parameter 94 or 95 to 0, I
>>> do not see this error. It only appears when p94 and p95 are both unequal
>>> 0. In my case I set them to 9.
>>>
>>> Do you know what might be causing this issue and how to solve it for the
>>> Cray XC40 system?
>>>
>>>
>>> Best Regards
>>> Steffen
>>>
>>>
>>>
>>> _______________________________________________
>>> Nek5000-users mailing list
>>> Nek5000-users at lists.mcs.anl.gov
>>> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
>>>
>>>
>
> _______________________________________________
> Nek5000-users mailing list
> Nek5000-users at lists.mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users