RISA + SAP2000 + Eriksson Culvert Results Not Matching?

[NotDeadYet]

I am an EIT with very little experience using modeling software trying to learn via Google and User Manuals.

I used Eriksson Culvert to model a simple structure: a 12'x12' concrete box culvert. All 4 members are 5ksi concrete (150pcf) with a 12"x12" cross-section. There is no fill over the structure.

A friend with a Mathcad sheet was able to confirm the dead load shears and moments from Eriksson.

I then made the same model in SAP2000 and one in RISA3D trying to duplicate the unfactored dead load shears and moments, but no two models give me the same results. I have checked and rechecked material properties, node coordinates, boundary conditions, and load applications - but can't for the life of me figure out why I am getting different results from each program.

Eriksson, SAP, & RISA models w/ reports are in the zip file attached if anyone wants to check them out.

I would very much appreciate it if someone could point out what I am doing wrong.

 

[JoshPlumSE]

The key in comparing any output of structural programs is making sure that the input is all the same. I will assume that you have done this. That all the material properties are defined the same way, that all the section dimensions, loads and boundary conditions are the same.

The next part is a little harder..... Looking for internal program assumptions that could cause differences.

I suspect that some of the differences related to a particular "feature" of RISA-3D: The Icracked property of a member. RISA will (by default) assume that a concrete beam's stiffness for strength level load combinations is actually 35% of the gross stiffness of the cross section. But, it will assume a stiffness of 50% of gross section for service level load combinations. This can be changed by the user in the Concrete Design tab of the members spreadsheet using the Icr factor. If you set it to 1.0 then ALL load combinations will use the moment of inertia of the gross cross section.

Why would RISA do something like this? Well, it's default is more geared towards concrete frames for buildings, rather than culverts or other specialized applications.

SAP2000, has similar factors. However, I believe it always defaults to 1.0. So, my guess is that the differences between SAP2000 and RISA would be due to differences in member stiffness.

I know almost nothing about Erickson Culvert. So, I don't know where any source of differences could be. If I were to GUESS, I'd say that it's probably reporting a different location for design because there's a different location (a distance d from the face of the member?)than what an FEM model would normally report (member ends of a centerline model).

Erickson Culvert may also have built in member end offsets to model a rigid region where the various planes of the culvert overlap.

 

[IDS]

Looking at the summary:
The shear forces across the roof are the same, and the bending moments differ by a constant value, as would be expected with equal shears.
This indicates that the moments at the top corners are different. This may be due to a different stiffness factor being applied to the walls and roof (as suggested by Josh Plum), and/or the horizontal soil loads may be calculated differently.

From a quick look, it isn't clear to me how the horizontal loads are calculated, but the constant difference in the shear forces in the walls would be consistent with different load factors in each of the three programs.


Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/

 

[NotDeadYet]

@JoshPlumSE: Thank you! Wow, I don't know if I ever would have caught that.

I changed the Icr factor on the Concrete Beam tab and Concrete Column tab of RISA's Members Table from [blank] to 1 and now the results between RISA and SAP agree. I feel like this was a significant step forward. I am looking into the member end offset/result location possibility now and will report back what I find.

@IDS: Thank you too!

After fixing the Icr factor, RISA/SAP are now both reporting 50% of the shear in the walls, 33% of the shear in the bottom slab, and 100% in the top slab vs Eriksson. As you point out, the difference is not only a constant factor, but changes at the corners. While it doesn't seem like it has to do with horizontal loading (Eriksson docs say horizontal loads are not included in the result set I am looking at and there are no horizontal loads in the RISA/SAP models), this does make me wonder if I missed something having to do with my nodal boundary conditions. I'll take another look at that after the end offset/result locations...

If anyone else has any insight, I would love to hear it!

 

[dold]

Looks like in your Eriksson model it actually is applying horizontal soil pressure as well as live load surcharge resultant horizontal pressure. Look at page 9 of 13 a few lines from the top: "Applied horizontal loads (k/ft)"

There could also be differences in boundary conditions. E.g., Eriksson may be using uniform bearing pressure along the bottom slab (pointing upward as the reaction), whereas SAP and RISA you have pinned/roller boundary conditions at the end joints. Page 5 of 13 towards the bottom "foundation model: uniform loads"

Also Eriksson might be using AASHTO load factors and SAP / RISA are using ASCE load factors? Page 5 of 13 towards the bottom "Load and resistance factors". In your RISA model you only have one load combination: page 8 of 18, and it is just 1.0DL.

 

[NotDeadYet]

@dold Thanks for the response!

Horizontal Loads: The results I am comparing are the values in the 'Vdc' column of the 'Unfactored Shears due to All Loads' table on p8 of the Eriksson report. There are other results in the Eriksson report that contain horizontal loads, but the column I am looking at only contains self weight of 1.95 kip/ft for each member (12"*12"*150pcf) and bottom slab soil reaction of 0.45kip/ft (1.95kip/ft*13'*3/13'). I am pretty confident about this as the shear in the wall is a constant 0.24 kip. If I had a horizontal earth pressure load applied the shear should increase with depth.

Load Factors: The results I am looking at in Eriksson are 'unfactored', and the load factors I have applied in RISA/SAP are 1.0 so it should not matter whether the program is set up to use ASCE or AASHTO load factors right?

Loads Applied: I should have defined the RISA load combination better. The load combination 'DL' includes the BLC category 'DL' at a factor of 1.0. The BLC category 'DL' includes the self weight factor of -1 in the Y Gravity column and a distributed load of 0.45kip in the Y direction applied to the length of the bottom member for the soil reaction.

Boundary Conditions: I've tried various combinations of boundary conditions (pin & roller), I & J offsets (6"), and Analysis Offsets in RISA (+/-y), but not making any progress.
Eriksson's user manual states that they use a structural model with a pin and a roller support, but I've also tried 2 pin supports to no avail.
If I remove the pin and roller supports the model would have inadequate restraint. I would need to replace them with something like springs along the bottom member, but I feel like there has to be something else that I am missing.

I'm trying to get a hold of my friends MathCad sheet at this point in the hopes that I will be able to figure out what it is doing the same as Eriksson but different than RISA & SAP...

 

[dold]

Interesting. Yeah if you're looking at just unfactored loads then the code shouldn't matter.

I'll try to look at this later on to address some of the other questions, but here's a spreadsheet for design of box culverts (credit to the almighty Alex Tomanovitch). It outputs factored forces/results and has some other limitations regarding soil pressures so your results might be a bit difficult to match up exactly. See attached.

 

[NotDeadYet]

@dold That is a very well put together spreadsheet. Thank you, I'm looking forward to going through it!


I finally figured out the final piece to make all three results agree...

The Eriksson manual talks about applying a distributed load in an upward direction to the bottom slab when we don't have precise soil information (aka the spring constant for your particular soil). The AASHTO LRFD BDS also makes reference to this practice in C12.11.2.3 "While typical designs assume a uniform pressure distribution across the bottom slab...". My problem was with the magnitude of that distributed soil reaction.

The Eriksson manual says the soil reaction load is equal to the loading applied to the top slab. That is not very intuitive, but sometimes empirically derived stuff isn't intuitive so trust and confirm right? Modeling this in RISA/SAP produces shear results that are correct in the top slab, but 50% less than expected results in the walls and no shear in the bottom slab at all.

An example box culvert design published by MNDOT in 2013 uses the same concept, but the magnitude of the soil reaction is equal to "The top slab, wall, and all haunch weights". This is slightly more intuitive, but also does not work and produces shear results that are correct in the top slab, but 50% less than expected in the walls and 33% less in the bottom slab.

If you keep reading in the MNDOT example though, they explain that "The bottom slab weight is not applied in the model because its load is assumed to be directly resisted by the soil." So I decided to model the soil reaction with a magnitude equal to the DL of all members, and now I get the expected results for unfactored DL shears.

Following this line of thought (even though neither of my examples state this), it seems that my distributed soil reaction applied to the bottom slab should have a magnitude equal to the sum of all the loads in the direction of gravity for the entire structure.

Can anyone confirm this assumption?