Quantcast Numerical method to solve high order ODEs. Text - Physics Forums Library

PDA

View Full Version : Numerical method to solve high order ODEs.


meridian
Aug27-08, 11:48 AM
here is a simplified version of my working equtions

y''' = \frac{(y'' y+y' y) y + y'y''}{y' + y''}

and 3 related boundary conditions, is there some hints to solve such equation numerically?

ThX

Defennder
Aug27-08, 11:53 AM
This might sound crazy, but note that everything here is a function of y. You may not have to resort to numerical solutions. But then again I haven't tried it out yet. Looks a little intimidating.

meridian
Aug27-08, 12:06 PM
in the equation,
y = y(x)
The original equation are much more complex, it is not possible to get a exact analytical solution for that. What I want to learn is the general numerical method to solve such equation.

HallsofIvy
Aug27-08, 01:03 PM
Let u= y'(x), v= y"(x). Then your equation is becomes
y'= \frac{(v y+u y) y + uv}{y + v}
That together with y'= u and u'= v gives you three interconnected first order equations. Do, say, a 4th order Runge-Kutta, advancing the step in all three equations at the same time.

Topher925
Aug27-08, 02:09 PM
Do, say, a 4th order Runge-Kutta, advancing the step in all three equations at the same time.

Thats would I would suggest. Although the Adams-Moulton-Bashforth method would work as well and probably give you a little bit more accuracy. But you will need the RK4 for the first couple steps anyway.