Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
NCTM or The Math Forum.


Marc
Posts:
108
Registered:
7/2/10


Re: ODE stiffness problems
Posted:
Apr 22, 2013 9:49 PM


"Torsten" wrote in message <khunje$be2$1@newscl01ah.mathworks.com>... > "Matthew" wrote in message <khtlea$ba3$1@newscl01ah.mathworks.com>... > > I removed the abs and signs by putting a constant low flow rate just to see if it would work. However it doesn't make any change it still stops at the same point. > > > I solved your problem with another software and it worked without problems. > > The problem with ODE15s seems to be the abrupt change of the variables > in the time span from t=0.95 to t=1. This is caused by the parameter Av. > > A remedy (I also used within my ODE integrator) is to call ODE15s seperately for > the time span from t=0 to t=0.05, then from t=0.05 to t=0.95 and then from t=0.95 to t=1. > Do it in a loop and set the inital conditions of the variables for a new time span to the final values of the variables from the preceeding time span. > > Best wishes > Torsten.
Torsten,
Is this "other" integrator you speak of Limex? If so, I found setting ODE15s MaxOrder to 2 sometimes help get similar/almost identical results. For my system, ODE23t also got me similar results.
I also used a similar "trick" of breaking up the time span manually.
Another set of integrators I have had nice luck with is the sundials ode integrators, code and ida (for DAEs) which are free and have a nice mex setup in Matlab.
Not 100% sure why setting the MaxOrder to 2 worked but it did.
My 2 cents....



