Validation of pier displacement

[jerbaroo]

Best bet is against Axis, Displace a pier by -0.1m and showing the values of the deck displacement. Do you think this is sufficient @rozsasarpi (do not reply on holiday)

[rozsasarpi]

If you add strains (or stresses) to the displacements than it is sufficient.

[jerbaroo]

Displacement very close under pier settlement with Axis and OpenSees. Please ignore plotting alignment with Axis, will fix quickly with final plots.

Axis

Generated

Axis

Generated

[jerbaroo]

Displacements are very close for pier settlement, strains not so much...

Axis

Generated

Axis

Generated

[jerbaroo]

Oh me oh my, those are stresses in Axis.. but converting the OpenSees strains to stresses, by multiplying my E, would make them even larger

[jerbaroo]

I will regenerate these plots, converting to microstrain first then converting to stress

[rozsasarpi]

In a plate loaded with a concentrated loads (inside a plate), if elastic plate theory is used (Kirchhoff-Love plate theory, Reissner-Mindlin plate theory) the stresses (strains, moments) under the concentrated load are infinitely large. So you should not try to compare them. You can also have singularities in other situations so if you observe non-convergence and highly concentrated stresses/strains then you can suspect a singularity in the solution.

You can cite this book as a support to this claim. https://www.cap-recifal.com/ccs_files/articles/cuveaqua1_denisio/Timoshenko_-_Theory_of_plates_and_shells.pdf

[jerbaroo]

I adjusted the colour bars in the generated strain contour plots from pier settlement so that the colorbars are the exact same as in Axis (same colours for same range). Notice the range of the colorbars in the third and fourth images, here you can see how Axis has maximum and minimum responses of greater magnitude.

[rozsasarpi]

I do not follow. The shown ranges of the colorbars of the related pairs (Axis, OS) are not the same.

[jerbaroo]

0.3 - 0.4 is the same yellow in the first two images (Axis, OS), 0.4 - 0.5 is the same orange etc..

[jerbaroo]

The range for each segment in the colorbar is taken from Axis, and the colour

[rozsasarpi]

0.3 - 0.4 is the same yellow in the first two images (Axis, OS), 0.4 - 0.5 is the same orange etc..

Ok, I was mislead by the different ranges. Still, looking at the displacement plots from 1mm pier settlement, it is not entirely true.

The range for each segment in the colorbar is taken from Axis, and the colour

You could change the range in Axis manually, double click on the colorbar.

[jerbaroo]

You could change the range in Axis manually, double click on the colour bar.

Yep, I used the "Round calculated values" option in there and then the contour plots seemed to line up a lot better. I noticed how it changes the values on the colour bar but exactly exactly what the option does is not documented.

Still, looking at the displacement plots from 1mm pier settlement, it is not entirely true.

Yes I didn't take the colour bar values for Displacement plots from Axis, because the OpenSees plots would then have out of range values, but now I think I will just stretch the range of the last colour segment to include those values.

[rozsasarpi]

On comparing Axis and OS: stop comparing the strains based on locations where you have strain concentrations (relatively small area with large strain and rapidly changing strain). The actual values are mesh dependent and very likely all singularities. Your model is good, show the convergence outside of the range of singularities and do not put sensors there (in most cases it would be also physically impossible), and you are done with the pier settlement. (same for stress)

[jerbaroo]

convergencestrain-pier-settlement-4.pdf

The pier being settled is close to the edge on the south side of the bridge. Consider the bottom two plots in that PDF. The colour of each line increases from blue to red, as the distance form the center point of the pier is increased. Thus in the bottom two plots the strains seems not to converge as the point is passed where the piers connect to the deck. This is at 1.55m.

Plan for things to run overnight:

• Regenerate this plot but instead of the pier center as the point being moved away from, use the center of the line of nodes where the pier connects to the deck (just one side), where convergence appears to be worst.
• As part of the same simulations collect min, mean, max strain for the convergence plot, but without any responses within 1m of where the pier connects to the deck.

[jerbaroo]

Once I have the data for the pier settlement convergence plot (tomorrow) I will send the convergence plots (I rewrote the whole pier convergence/plotting code last night so it should allow me to iterate on the plots relatively fast once I have the data). Combining that with the convergence information from the point load tests this should allow me to set where sensors are allowed to be placed, will look something like:

unavailable-sensors.pdf

Plan for the weekend is:

• the pier convergence plots, just need the data
• time series plot of truck 705 moving along the bridge based on generated data, straightforward task. This is for comparison against the measurement data, Arthur sent me something for that but the visualization was already done so I have another meeting planned for next week to get the underlying data
• focus on Lisa, I want something being trained on Lisa on generated data, even if the features/model don't make much sense, point is to have the workflow figured out about generated data & code to Lisa and getting responses

[jerbaroo]

bridge-705-0-convergence-nesw-pier-5.pdf bridge-705-0-convergence-nesw-pier-4.pdf