Re: DAE daspk solver problem

[Bill Greene]

I was able to solve this using ode5r from odepkg. My test code is below:

function radau5PendulumDAE

vopt = odeset ('InitialStep', 1e-3, 'AbsTol', 1e-8);

len=1; theta0=pi/2; m=1; g=9.81;

y0=[ len*sin(theta0), -len*cos(theta0), 0, 0, m*g];

odef = @(t,x) odeFunc (t, x, m, g, len)

sol=ode5r (odef, [0, 5], y0, vopt);

figure; plot(sol.x, sol.y(:,1));

end

function y = odeFunc (t, x, m, g, len)

lam = x(5);

y=[x(2); -lam*x(1); x(4); -lam*x(3)-m*g; x(1)^2 + x(3)^2 - 1];

end

On Fri, Oct 28, 2016 at 2:18 AM, Sebastian Schöps <address@hidden> wrote:

Lazy_Tom wrote
> I don't know why DAE solvers index capabilities are not reported in Octave
> manual. Matlab at least tells you that you can go no further than index 1
> (at least without doing some "magic"...)

You are right. Feel free to oben a bug report at savannah, maybe immediately
suggesting a string that can be added to the help. However, I assume that
dassl and daspk will be deprecated in some future version since we are
moving to odeXY compatible solvers.


Lazy_Tom wrote
> about the paper you cited, yup, I'm aware of it and I agree, it is quite
> nice.
> Well, at least I'll gave it a try, I think that for DAE problems I'll move
> to openModelica

I think that ode5r from odepkg should be able to solve index3 problems since
its based on Radau 5 [1]. We hope to have a new release of odepkg quickly
after Octave 4.2. I will keep in mind to include such information in the
help.

Regarding openModelica: yes, they do some clever tricks for higher index. As
far as I recall it's based on dummy derivatives [2]. Such stuff is easier
for modelica since it is in some sense a symbolic language. However, to my
best knowledge, it is not really well suited for big problems. So, as usual,
the best choice depends on what you want to do :)

Seb.

[1] http://www.unige.ch/~hairer/prog/stiff/radau5.f
[2] http://dx.doi.org/10.1137/0914043



--
View this message in context: http://octave.1599824.n4.nabble.com/DAE-daspk-solver-problem-tp4680377p4680395.html

Sent from the Octave - General mailing list archive at Nabble.com.

_______________________________________________
Help-octave mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/help-octave