Closed jerbaroo closed 4 years ago
The r2 plot seems to be in reasonable agreement between OpenSees and Diana. I have verified that the OpenSees results are coming from OpenSees and not another source by changing a value in the Strain calculation.
It is strange then that the maximum and minimums are so different between Diana and OpenSees. See the case below with a load at the center of the bridge.
Given that the displacement is in good agreement (#74) and the r2 of strain isn't so bad, then I wonder why these maximum and minimums are so different. After the meshing/debugging session this weekend I will re-run this analysis.
I measure minimum strain as -5.07, while Diana is 3.06. The maximum recorded for OpenSees is 17.6 while for Diana it's -11. I note how the signs are flipped and OpenSees values are ~1.5 x the Diana values in absolute terms.
Similar situation for point A, points C and D are more widely off, but these are the "edge" case points, close to piers and abutments where I expect better meshing will help
I started the more thorough debugging/validation of the material properties, and found a discrepancy in thickness, I will fix it and validate the remaining properties further on the weekend. Focus on temperature for now.
It is strange then that the maximum and minimums are so different between Diana and OpenSees.
Maybe it is because of this modelling decision in Diana:
The variation of the vertical position of the centre of gravity of the deck is accounted for in the FE model by giving as input the eccentricity of the centroid of the deck with respect to the bottom surface.
From the bridge 705 report I shared with you earlier (Probabilistic system and load identification of bridge 705 using on-site measurements and finite element analysis)
Note that this eccentricity is supposed to be closer to reality, so it is weird that your results are better than those of Diana. I expect the eccentricity to have a small to negligible effect on displacements and potentially a marked effect on the strains.
I declare the difference unimportant at this point (low priority). The model is close enough to reality, let's focus on other tasks, e.g. temperature and one-class classification.
convergencestrain-point-load-a.pdf min-max-strain.pdf
Plan for strain:
Nice plots! If you added horizontal grid lines/guides to the plots in the first pdf that would guide the eye to assess convergence.
Once #83 is done