bow-simulation
Improve strain interpolation to avoid nick in stress curves
In GitLab by spfeifer on Sep 3, 2017, 02:23
The beam elements have a constant longitudinal strain. When calculating stresses at intermediate limb nodes, the strains of two adjacent elements are averaged together. This can't be done for the first and last nodes, which creates little nicks as shown in the attachment. This also shows in the resulting stress curves (See #69).
A possible solution would be to include some sort of extrapolation in the calculation of the strains of the first and last nodes.
In GitLab by ozra on Jan 8, 2018, 04:30
Make the subdivision more fine grained for ends? Reducing the level of the kink. Or even more generally: based on expected (heuristically?) deformation level at each point in limb. (The base section probably deforms less and can have longer subdivision segments, while a recurve tip likely deforms more heavily from resting position, requring tighter subdivisions) - improving practical resolution at the same number of subdivisions of limb (say 30)?
In GitLab by spfeifer on Jan 8, 2018, 17:32
Using smaller elements where needed would be nice. But unfortunately that doesn't play well with the dynamic solver, because the time step for a stable solution is limited by the inverse of the maximum eigenfrequency of the system. Introducing small elements also introduces large frequencies which makes the solution less efficient.
In GitLab by ozra on Jan 9, 2018, 11:31
Alright, that was beyond my current knowledge!