BeamDyn: Updated integration scheme for nodal load outputs

[bjonkman]

@andrew-platt updated the integration scheme he used for the nodal load outputs in BeamDyn. The updated integration scheme better matches the analytical result:

[ashesh2512]

There are 3 minor merge conflicts resulting from the recently merged pull request #155. I have gone through the conflicts and once this pull request is approved, I should easily be able to resolve those conflicts before merge.

[ashesh2512]

@bjonkman overall the code looks good to me. I have a few comments -

1. Following pull request BeamDyn finite differencing capability and variational consistent output in summary file #155 the error calls at lines 3657 and 3663 can very well use RETURN instead of CYCLE because the calling function (BD_Static) now checks for errors and aborts immediately upon calling BD_StaticSolution. Would you agree ?

2. In the subroutine BD_diffmtc, the data structure p is used only to access nodes_per_elem. If so, I suggest passing in nodes_per_elem instead of p. This would make that routine modular and hence unit-test-able.

3. Do you know if any of the regression test results are affected as a result of the update integration scheme ?

4. Do you (or @andrew-platt) have any documentation regarding the computation of BldInternalForceFE ? Adding this information to the BeamDyn manual might be very helpful for new developers of OpenFAST/BeamDyn.

[bjonkman]

@ashesh2512

I don't have time to address about all of your comments today, but I did want to mention something regarding the error handling in your first comment: Your static algorithm is a little different than ours, but I do not think that #155 should have added a return after the call to BD_StaticSolution. Ignoring the error is intentional there (i.e., the lines 3512-3516 that you added to BeamDyn.f90 were intentionally left out) otherwise the else branch in the statement that follows is irrelevant. You could perhaps change the if (piter .le. p%niter) then to if (ErrStat2 < AbortErrLev) instead, but the error checks after BD_StatisSolution shouldn't occur.

Also, the lines you added at the end of that routine (lines 3562-3566 in BeamDyn.f90) are redundant given the two lines that follow them.

I think there are some other potential error handling issues in your BD_StaticSolution subroutine in #155 (again, related to returning on error with the loop counter less than the max specified in that loop).

[ashesh2512]

@bjonkman

The else condition still works because I removed line 3746 in BD_StaticSolution in #163 . I believe that line wasn't significant anymore considering the fact when (piter .le. p%niter) we reduce the load and move on till the maximum number of load steps is reached. So we never really exit the code for a static simulation based on whether the maximum Newton steps have been reached.

I agree with the redundancy of lines 3562-3565. I will submit a pull request for that right away. Thank you for pointing it out.

[ashesh2512]

This pull request has been modified in #265 due codes diverging significantly.