Discussion Forum

Hello users !

From two years, I try to learn COMSOL (the license from my university) by using some personal models but the last result is not like I want...

In fact, I try to solve a very quick differential equation (COMSOL's notations) :

utt = exp(-(x^2+y^2+z^2))*exp(-t)
with the initial conditions : u(x,y,z,t=0)=0 and ut(x,y,z,t=0)=0

where u(x,y,z,t) is a spatio-temporal function and the second derivative (utt) is performed with respect to the time only.

*To find the solution u(x,y,z,t), I used the "PDE, coefficient form" with the coefficients "e_a = 1" and "f = exp(-(x^2+y^2+z^2))*exp(-t)" in the menu "Physics / subdomain Settings" (in this problem, my subdomain is simply a solid cylinder). All the others coefficients appearing in this menu are zero.

*Also, I took the initial values u(t_0)=0 and ut(t_0)=0 in the "init" menu.

Finally, I solved my model by using the solver "Direct UMFPACK" in the solver manager menu but the final result is not depends on the solver used.

Unfortunately, the result obtained is really wrong ! It is very different of the analytical solution...For example, the true solution gives u = 3.46 for x=y=z=0 and t=4.45. The solution given by COMSOL is 1.28...

You will find my model in attachement...Can COMSOL solve such a problem ? An experimented user can help me to find the problem ?

Thank you.

Sébastien