On Sun, Feb 16, 2014 at 11:50 PM, Caleb Hattingh <caleb.hatti...@gmail.com> wrote: > Given a typical DiffustionTerm(coeff=D), and typical CellVariable phi, is > there an easy way to obtain the Jacobian of such a system?
This is a good question. It would be really nice if FiPy could do this, but it can't at present. The difficulty is of course that the coefficients are functions of the variables so this would have to be automated somehow, probably with sympy. However, if you know your equation, say it is eqn = TransientTerm() == DiffusionTerm(phi**2) for example and you are solving for "phi", then the "correction equation" for a Newton step would be eqnCorr = TransientTerm() == DiffustionTerm(phi**2) + ConvectionTerm(2 * delta_phi * phi) + R where R is the residual calculated from "eqn" and you are now solving for "delta_phi", the correction to "phi". I don't have any working code to show you on this, but I know that others have used FiPy in this way. If you just want the numerical Jacobian, I would guess that Trilinos can do this. -- Daniel Wheeler _______________________________________________ fipy mailing list fipy@nist.gov http://www.ctcms.nist.gov/fipy [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]
