[Nek5000-users] RB convection with high Rayleigh numbers

[nek5000-users at lists.mcs.anl.gov]

Let me echo Aleks' comments by pointing out that any experimentalist
does the same thing.

They don't just turn the heater on at full power at the start --
invariably they ramp it up.   In nonlinear systems, it matters
how you get to the state of interest... Otherwise they would
wait forever for the system to settle down.

Thus, this is a physics issue - not a numerics issue.

Paul



On Fri, 12 Nov 2010, Aleksandr Obabko wrote:

> Hi David,
>
> The simulation would just blow up due to runaway CFL if I did not...  And lowering timestep did not help.  Plus usually it is quicker to reach statistically steady state for lower Ra/resolution and for the subsequent solutions restarted from a previous one with bumped up Ra/resolution as compared to one straight run at the target high Ra/resolution.
>
> Best,
> Aleks
>
>
>
> ----- Original Message -----
> From: "David Goluskin" <goluskin at gmail.com>
> To: "Aleksandr Obabko" <obabko at mcs.anl.gov>
> Sent: Thursday, November 11, 2010 9:32:10 PM
> Subject: Re: [Nek5000-users] RB convection with high Rayleigh numbers
>
> Hi Aleks,
>
> Back in May you told Mani this:
>
>
>
>
> A while back I was able to simulate 3D convection between stress-free surfaces at Rayleigh number Ra=1e6 (and Taylor number 5e5) with 4600 elements (and lx1=6) using time step 1e-5.
>
> I remember that to get to Ra=1e6 I had to get lower Ra solution first and then use it as initial condition.
>
>
> I was just wondering why you had to do that. What would happen if you didn`t?
>
> Best,
>
> David
>