[petsc-users] testing for and removing a null space using JFNK

[Dave Lee]

Hi Matt,

Perhaps. I need to think more about weather this system is over-determined,
and if so how to recover the velocity components from the Newton iteration.

Thanks for your help.

Cheers, Dave.

On Fri, Apr 5, 2019 at 1:27 PM Matthew Knepley <knepley at gmail.com> wrote:

> On Thu, Apr 4, 2019 at 9:11 PM Dave Lee <davelee2804 at gmail.com> wrote:
>
>> Hey Matt,
>>
>> I'm not solving NS per se, but rather wrapping up a Navier Stokes solver
>> within a SNES to iterate over the solution of the Navier Stokes equations
>> with a certain time period in order to determine approximate periodic
>> solutions to the NS equations (with some corrections).
>>
>> My residual vector is basically a difference between the final and
>> initial velocity states of the NS solve (with corrections). However since
>> one component can be diagnosed from the others via the divergence free
>> condition (up to a constant), I suspect that maybe what I should be doing
>> is just omit one of the velocity components from the residual vector, and
>> then diagnose this from the others via incompressibility, rather than try
>> and correct for this after the vectors have already been assembled. This is
>> all outside the scope of my PETSc question, and I don't expect you to have
>> an answer, just mentioning it since you asked.
>>
>
> Interesting.  It sounds like you can impose this condition the same way we
> impose \int p = 0.
>
>   Thanks,
>
>     Matt
>
>
>> Cheers, Dave.
>>
>> On Fri, Apr 5, 2019 at 12:12 AM Jed Brown <jed at jedbrown.org> wrote:
>>
>>> Mark Adams via petsc-users <petsc-users at mcs.anl.gov> writes:
>>>
>>> > On Thu, Apr 4, 2019 at 7:35 AM Dave Lee <davelee2804 at gmail.com> wrote:
>>> >
>>> >> I already have the Navier Stokes solver. My issue is wrapping it in a
>>> JFNK
>>> >> solver to find the periodic solutions. I will keep reading up on SVD
>>> >> approaches, there may be some capability for something like this in
>>> SLEPc.
>>> >>
>>> >
>>> > Yes, SLEPc will give you parallel eigen solvers, etc.
>>>
>>> Even so, computing a null space will be *much* more expensive than
>>> solving linear systems.
>>>
>>
>
> --
> What most experimenters take for granted before they begin their
> experiments is infinitely more interesting than any results to which their
> experiments lead.
> -- Norbert Wiener
>
> https://www.cse.buffalo.edu/~knepley/
> <http://www.cse.buffalo.edu/~knepley/>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20190405/315f8d2c/attachment.html>