2
$\begingroup$

I am solving the one-dimensional wave equation over regions where the bulk modulus (and thus the wave speed) vary continuously over a region. The current version seems to assume that the material properties are constant over an FEM element. Some literature suggests that isoparametric elements can help model regions where the material properties vary continuously. Can you suggest how to model regions where the material properties vary in a continuous fashion?

In the wave equation code I am using, kappa and rho can vary across an element. The code is as follows:

eqn = 1/κ[x] D[u[t, x], {t, 2}] + 
Div[-ρ[x]*Grad[u[t, x], {x}], {x}] == 
1/κ[x]*10*Exp[-50 (x^2)]*Sin[2 π f t] + 
NeumannValue[0, x == 0] + NeumannValue[-Derivative[1, 0][u][t, x], x == 
xMax]; 
ic = {u[0,x] == 0, Derivative[1, 0][u][0, x] == 0};
$\endgroup$
3
  • $\begingroup$ I don't understand your question. Have you tried to define example function kappa and rho and to submit the system to NDSolve? What is the issue? $\endgroup$ Aug 5, 2019 at 1:45
  • 2
    $\begingroup$ "The current version seems to assume that the material properties are constant over an FEM element." Might be true for element order 1. For order 2, probably a nonexact quadrature rule is applied, but I doubt that the coefficients are assumed to be constant. Anyways, with sufficiently small elements and suffciently smooth material coefficients, that should not matter at all. $\endgroup$ Aug 5, 2019 at 1:49
  • $\begingroup$ Thanks, I use order 2 elements. Some literature suggests that accuracy is improved by using isoparametric elements.This shows up in functionally graded materials (FGM). My current results seem OK, using a fine grid in the region where the coefficients are rapidly varying. $\endgroup$ Aug 6, 2019 at 13:34

1 Answer 1

4
$\begingroup$

Here is a wave equation with a $x+1$ as a factor in the wave equation.

\[CapitalOmega] = 
  RegionDifference[
   RegionDifference[Rectangle[{0, 0}, {2, 1}], 
    Rectangle[{9/10, 0}, {11/10, 4/10}]], 
   Rectangle[{9/10, 6/10}, {11/10, 1}]];
sol = NDSolveValue[{D[u[t, x, y], {t, 2}] - 
      Inactive[Div][(x + 1) Inactive[Grad][u[t, x, y], {x, y}], {x, 
        y}] == 0, DirichletCondition[u[t, x, y] == 0, True], 
    u[0, x, y] == 2*Exp[-125 ((x - 0.25)^2 + (y - 0.5)^2)], 
    Derivative[1, 0, 0][u][0, x, y] == 0}, 
   u, {t, 0, 2}, {x, y} \[Element] \[CapitalOmega]];

ListAnimate[
 Table[Rasterize[
   Plot3D[sol[t, x, y], {x, y} \[Element] \[CapitalOmega], 
    PlotRange -> {-0.75, 2}, AspectRatio -> Automatic]], {t, 0, 2, 
   1/25}], SaveDefinitions -> True]

I would be interested to know what lead you to the conclusion that "...current version seems to assume that the material properties are constant over an FEM element." This is not correct and if this appears or is suggested in the documentation it needs to be fixed. But for that I'd need to know what lead you to that conclusion.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.