I have an 2D elastodynamics problem, that is a problem which is driven by the Cauchy equation: $$\rho\ddot u-\mathrm{div}\sigma=\rho f$$ where $u$ is the displacement, $\sigma$ the Cauchy stress tensor, $f$ the density of volumetric forces, and $\rho$ the mass density. The constituve law is Hook's law from linear elasticty (homogeneous and isotropic medium). To solve it numerically, I use the finite element method. After space discretization I will have the following equation: $$M\ddot u+Ku=F$$ where $M$ is the mass matrix, $K$ the stiffness matrix and $F$ the load vector. The solution of such an equation oscillate around an equilibrium position and my aim is to add a viscous damping term $C\dot u$ that will critically damp the system, by which I mean I would like to avoid the oscillatory behaviour as much as possible. The equation now reads: $$M\ddot u+C\dot u+Ku=F$$ The problem is how to choose the matrix $C$ that will achieve the desired effect?

In modal analysis of dynamic structures, the same term $C\dot u$ is added to the discretized Cauchy equation to take damped vibrations into account but for a different purpose: we search for proper modes of vibration. Here we already know the two proper modes of vibration: S-waves and P-waves. So taking into account this information, I would like to build a matrix $C$ that will reflect this physical fact and that contains only constants like $E$, $\mu$ (two classical elasticity modulus) and $\rho$, rather than assuming a Rayleigh damping term $C=\alpha M+\beta K$ and finding the relevant values of $\alpha$ and $\beta$ by hand (numerically).

Thank you very much for any idea!


I'm a bit rusty on structural dynamics, but pages 87-92 of Meirovitch's "Fundamentals of Vibration" explain that a system

$$ m\ddot{x} + c\dot{x} + kx = 0 $$

is critically damped when $c=2m\omega$, where $\omega=\sqrt{\frac{k}{m}}$ is the natural frequency of vibration.

There is a natural generalization to the system of ODEs

$$ M \ddot{x} + C \dot{x} + Kx = 0. $$

For an eigenpair of the pencil $(K,M)$, say $(\omega^2,u)$, where $u$ is a unit vector, we have that

$$ K u = M u \omega^2, $$

and thus we could construct a $C$ which critically damps this mode of vibration if we set

$$ C := 2 \omega (Mu)u^T, $$

where the matrix $(Mu)u^T$ transforms $u \mapsto Mu$, so that $(Mu)u^T$ is analogous to the $m$ term in the equation $c = 2 \omega m$, and the $2$ and $\omega$ require no generalization.

For some set of modes $\{u_j\}_j$, with corresponding eigenvalues $\{\omega^2_j\}_j$, you could critically damp those modes by setting

$$ C := 2 \sum_j \omega_j (Mu_j)u_j^T. $$

  • $\begingroup$ Thank you very much Jack, your answer gives me an idea that I will investigate soon. $\endgroup$ – Benjamin Apr 21 '12 at 6:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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