Merge of #163 (BeamDyn: Updated integration scheme for nodal load outputs)

[ashesh2512]

THIS PULL REQUEST IS READY TO MERGE

This pull request although ready to merge with current dev, requires further review of the failed tests and subsequent update of gold files

Feature or improvement description

This pull request merges #163 (authored by @bjonkman and @andrew-platt) into the current dev after review and code modifications to align with current dev.

Impacted areas of the software

Beamdyn

Additional supporting information

Automated test results

A couple of tests fail, but converge to the same result as the baseline solution

Twisted beam results -

Curved beam results -

[ashesh2512]

@andrew-platt Could you please look through these changes? A couple of tests failed. Is this an artifact of the modified computation of y%ReactionForce%Force and y%ReactionForce%Moment or do you suppose something else could have contributed to these changes?

[andrew-platt]

Hmm... these results don't look right. I'll try to take a look in more detail later in the week.

UPDATE: I misread the plots. That is time, not distance along the x-axis. This does look correct!
There are a lot of things that could contribute to the second timestep converging slightly differently. For static solves, we found things converged nicely after about 3 steps.

[ashesh2512]

I further wanted to add Fx and Mx plots along the beam axis for the standalone 5MW dynamics test case. The plots below show a clear improvement over the previously found oscillations in load profiles near the blade tip.
Blue - current approach
Red - approach modified via this pull request

the closeup -

[andrew-platt]

This looks correct to me. Do the Fy and My terms show similar trends as well?

[ashesh2512]

@andrew-platt I plotted the Fy and My and the My profile with the current approach shows load reversals near the blade tip (see close-up of red plot) which goes away with refinement in p. Furthermore I don’t see any such behavior in Fx. Also note, for the coarser p I do not see load reversal with the old approach.

Entire blade -

Zoomed in -

p-refinement

p-refinement zoom in

[ashesh2512]

After some discussion with @andrew-platt, since the load reversal vanishes with increasing the refinement, there is likely a slight under-sampling near the tip with where the trapezoidal quadrature points are defined. So it is suspected that this might be a small artifact of the spectral method resulting in a difference 4 orders of magnitude smaller than the largest load, and not an issue with the load integration scheme.

[andrew-platt]

Since this PR supercedes the PR #163, can we close that one?