Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > comp.soft-sys.math.mathematica > #6714 > unrolled thread

Memory low for PDE

Started bySérgio Lira <sergiobodoh@gmail.com>
First post2011-11-25 09:57 +0000
Last post2011-11-29 12:41 +0000
Articles 3 — 2 participants

Back to article view | Back to comp.soft-sys.math.mathematica


Contents

  Memory low for PDE Sérgio Lira <sergiobodoh@gmail.com> - 2011-11-25 09:57 +0000
    Re: Memory low for PDE Grischika <Grischika@mail.ru> - 2011-11-27 09:15 +0000
      Re: Memory low for PDE Sérgio Lira <sergiobodoh@gmail.com> - 2011-11-29 12:41 +0000

#6714 — Memory low for PDE

FromSérgio Lira <sergiobodoh@gmail.com>
Date2011-11-25 09:57 +0000
SubjectMemory low for PDE
Message-ID<janopr$atu$1@smc.vnet.net>
Hello folks,

I am trying to solve some coupled partial diferential equations using
NDSolve but after 10 min of calculations the kernel shuts down and
says: "No more memory available. Mathematica kernel has shut down."

The equations are pretty huge and I have tried to change some
integration options such as AccuracyGoal, MaxStepFraction,
PrecisionGoal, MaxSteps, Method, SpatialDiscretization, but still
couldn't solve the equations.

This is the program:

<< Calculus`VectorAnalysis`

Nch[f_, x_, y_, t_] := Nch[f, x,y, t] = (Grad[f, Cartesian[x, y, z]])/
Sqrt[DotProduct[(Grad[f, Cartesian[x, y, z]]), (Grad[f, Cartesian[x,
y, z]])]];
Sch[f_, x_, y_, t_] :=Sch[f, x, y, t] = CrossProduct[Nch[f, x, y, t],
{0, 0, 1}];
K[f_, x_, y_, t_] := K[f, x, y, t] = -Div[Nch[f, x, y, t],
Cartesian[x, y, z]];
G[f_, x_, y_, t_] := G[f, x, y, t] = 2*DotProduct[Sch[f, x, y, t],
(B*Grad[K[f, x, y, t], Cartesian[x, y, z]] - {x, y, 0})];

eps = 0.1; epst = 0.2; B = 0.01;

eqn2 = NDSolve[{

epst*D[psi[x, y, t], t] == Laplacian[psi[x, y, t], Cartesian[x, y,
z]]
+ 1/(eps*Sqrt[8])*G[theta[x, y, t], x, y, t]*(1 - (theta[x, y,
t])^2),

eps^2*D[theta[x, y,t], t] == Laplacian[theta[x, y, t], Cartesian[x, y,
z]] + eps^2*DotProduct[{0, 0, 1}, CrossProduct[Grad[psi[x,y, t],
Cartesian[x, y, z]], Grad[theta[x, y, t], Cartesian[x, y, z]]]],

psi[x, y, 0] == 0, psi[-5, y, t] == 0, psi[5, y, t] == 0, psi[x, -5,
t] == 0, psi[x, 5, t] == 0,

theta[x, y, 0] == Tanh[(1 - Sqrt[x^2 + y^2])/(0.1*Sqrt[2])],
theta[x, -5, t] == 0, theta[x, 5, t] == 0, theta[-5, y, t] ==0,
theta[5, y, t] == 0},
{psi, theta}, {x, -5, 5}, {y, -5, 5}, {t, 0, 5}]


Should I try to decrease the integration step and grid? Is there any
method that could help?

Cheers,
Sergio

[toc] | [next] | [standalone]


#6775

FromGrischika <Grischika@mail.ru>
Date2011-11-27 09:15 +0000
Message-ID<jasv31$758$1@smc.vnet.net>
In reply to#6714
If you want to get result at the end of time interval you can just
reduce last NDSolve parameter for example
{t,4.9,5}

It suppresses data storage on all time interval but it stores solution
only at the end. So this should reduce memory usage.

[toc] | [prev] | [next] | [standalone]


#6818

FromSérgio Lira <sergiobodoh@gmail.com>
Date2011-11-29 12:41 +0000
Message-ID<jb2jtc$67v$1@smc.vnet.net>
In reply to#6775
Hi Grischika,

I have tried this, but it didn't work =/

Tx,
Sergio


On 27 nov, 06:15, Grischika <Grisch...@mail.ru> wrote:
> If you want to get result at the end of time interval you can just
> reduce last NDSolve parameter for example
> {t,4.9,5}
>
> It suppresses data storage on all time interval but it stores solution
> only at the end. So this should reduce memory usage.

[toc] | [prev] | [standalone]


Back to top | Article view | comp.soft-sys.math.mathematica


csiph-web