Inconsistent dimensions for damping constants in AnalyticalLinearDamper

[sy-cui]

The linear viscous damping forces and torques within the linear and angular momentum equations are given as follows (Gazzola 2018):
$$\mathbf{f}_v =-\gamma_t \mathbf{v}, ~~~~~~~~ \mathbf{c}_L^v = -\gamma_r \mathbf{\omega}_L $$
where the paper claims that the damping coefficients $\gamma_t$ and $\gamma_r$ can be set to equal. This statement is not strictly correct as they are dimensionally inconsistent:
$$[\gamma_t] = FL^{-1}T, ~~~~~~~ [\gamma_r] = FLT$$

The analytical linear damper assumes an eigenvalue problem with respect to the damping force, where the damped translational and angular velocities are respectively
$$v_{\text{damped}} = v_0 \exp \left(-\frac{\gamma_t}{m}t\right), ~~~~~~~ \omega_{\text{damped}} = \omega_0 \exp \left(-\frac{\gamma_r e}{J}t\right)$$
Current implementation assumes $\gamma = \gamma_t = \gamma_r$, and the input damping_constant corresponds to $\gamma / m$ where $m$ is the nodal / elemental mass. This includes a dimensional inconsistency and incurs problems in non-dimensional analysis.

Two possible remedies are proposed:

1. The analytically rigorous approach is to include both $\gamma_t$ and $\gamma_r$ from user input, and compute the exponential coefficients based on these inputs and rod properties
2. Assume that $k = \frac{\gamma_t}{m} = \frac{\gamma_r e}{J}$, with $k$ having inverse time dimension ($[k]= T^{-1}$). The damper class collects only one user input $k$ instead of two, and no access to kinematic/material properties is required.

The original implementation should be retained for the moment and possibly removed in a future update. If removal is expected, usage of the original approach should emit a deprecation warning.