Internal Damping of ChElementCableANCF (possible bug?)

[David Marten]

Hi all!

recently I have started looking at the damping of cable elements ( ChElementCableANCF ).

I have noticed that the rayleigh damping coefficient specified in  ChBeamSectionCable is not having an effect (and is effectively not accessed by  ChElementCableANCF ).

However, using ChElementCableANCF::SetAlphaDamp(double a), it is possible to assign a "internal damping coefficient" to the cable element. My question is: how can this damping factor be interpreted? Is it similar to the rayleigh beta (K-proprtional) damping?

Furthermore, I think I have found a bug in: ChElementCableANCF::ComputeInternalJacobians(double Kfactor, double Rfactor). When setting a damping factor using ChElementCableANCF::SetAlphaDamp() my test system could not be solved. Looking at the source code of  ChElementCableANCF::ComputeInternalJacobians, I found the following section:

// Add part of the Jacobian stemming from elastic forces

for (int inode = 0; inode < 2; ++inode) {

   pos[inode].x() += diff;

   ComputeInternalForces_Impl(pos[0], D[0], pos[1], D[1], pos_dt[0], D_dt[0], pos_dt[1], D_dt[1], F1);

   m_JacobianMatrix.col(0 + inode * 6) = (F0 - F1) * (1.0 / diff) * Kfactor;

   pos[inode].x() -= diff;
   ...

   ...

   ...

 }

// Add part of the Jacobian stemming from internal damping forces, if selected by user.

if (m_use_damping) {

   for (int inode = 0; inode < 2; ++inode) {

      pos_dt[inode].x() += diff;

      ComputeInternalForces_Impl(pos[0], D[0], pos[1], D[1], pos_dt[0], D_dt[0], pos_dt[1], D_dt[1], F1);

      m_JacobianMatrix.col(0 + inode * 6) = (F0 - F1) * (1.0 / diff) * Rfactor;

      pos_dt[inode].x() -= diff;

      ...

      ...

      ...

    }

}

I think that in the 2nd loop (m_use_damping), the damping contribution should be added to the jacobian matrix, by changing:

      m_JacobianMatrix.col(0 + inode * 6) = (F0 - F1) * (1.0 / diff) * Rfactor;

to

      m_JacobianMatrix.col(0 + inode * 6) += (F0 - F1) * (1.0 / diff) * Rfactor;

After changing this the damping of the cable element seems to work.

Thanks a lot for your help!

Best regards,

David

[chao peng]

Hi Dr. Marten,

hank you for reporting this bug and investigating the solution.

I have just pushed your proposed changes to GitHub.

See commit: c911f6c on 19th July.

Fixed bug in ChElementCableANCF: when the damping is used, the stiffness terms are overwritten by the damping, leading to incorrect results. Now the stiffness and damping terms are accumulated.

Best regards,

Chao PENG