FEA I-girder bridge model

[rabi24]

Hello,

I have used Strand7 to model a simple span I-girder bridge using plate elements for deck and beam elements for the girders with centroids of both coinciding. I am getting good results for analysis of bridge under dead load however when I add a truck load, I get very different results than what I would expect and have gotten by hand. Anyone able to maybe point out a reason for this? Thank you!

 

[ukbridge]

Are you comparisons based on stresses, moments or otherwise?

If your different results are based upon bending moment comparisons, you need to set the centroid of the slab at the correct position with respect to the neutral axis of the girder. The bending moments are calculated by using shell-slicing utilities or similar which numerically integrates the stress profile over the depth of the slice.

Also check that both the girder and slab are deflecting as a rigid body, if they are not you can apply an equivalence which will help with this. If you're using patch loads for the truck, make sure that the load is applied to what you want which can be done using a search area (I use LUSAS so not sure if Strand terminology is different).

Consider checking your hand model too if you're unsure, I'd probably use a simply transverse influence model to calculate the proportion of the truck spread to the girder considered, then use a simple line-beam model to determine BM's, SF's etc.

Goes without saying to check things like deflected shape and so on, but I think the above points should help out. Good luck!

 

[rabi24]

Hey, thanks for your reply.

Yeah the bending moment results are half of what I am expecting. What is the correct position of the centroid of the slab w.r.t centroid of the girder? Shouldn't they coincide?

Thank you!

 

[IDS]

First, I'd suggest reading a good book on bridge deck analysis to get a better understanding of the process, what simplifications are involved, and how they affect the results. I can recommend Bridge Deck Behaviour by Hambly, which is quite old but it covers the basics well.

Second, I'd recommend sticking with your simplified model until you find out what is causing the discrepancy with the results. If you offset the centroids of the girders and the slabs you will have to include the moment effect of the axial loads in the longitudinal members, which makes things much more complicated. You should be able to get a reasonable agreement between a plane analysis and a simple analysis using beam theory with a load distribution factor.

How did you calculate the stiffness of the longitudinal girders? You should use the composite section, including the top slab.

Have you checked the total bending moment across a transverse section? That should agree closely with the moment in a line beam under the same total load.

I use Strand7; if you would like to upload your model I would be happy to have a look at it.

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/

 

[ukbridge]

It might be worth looking at doing a grillage analysis alternatively, as the post-processing is much simpler and convenient than idealising the slab as a plate. They're much simpler, but require a bit more data entry, but IMO this disadvantage is massively offset when you consider how simple life is made extracting results from numerous load cases. Like IDS check out Hambly.

I agree with IDS it will complicate things due to the eccentricity, but IMO it should to be modelled if you're extracting bending moments using an in-software shell slicing utility in which the bending moment is calculated from both the stresses in the plate and beam. If I'm mistaken (and I may well be) the lever arm needs to be modelled, the exception I suppose if you have a very shallow span to depth ratio.

 

[jhardy1]

With Strand7, you can mesh the slab plates to the same nodes as the beams, but you should be offsetting the plates with respect to the beam axes, so that they are modelled in their correct spatial position - bottom of slab on top of the top flange of the I-Beam. Use the "Attributes | Plate | Offset" menu item to pull up the data entry form.

To get the effective total bending moment, you need to allow for the combined moments in the beam and the slab - and that means not just adding the moments of each component, but also considering the couple for the membrane stress in the slab (compressive stress at midspan) with the axial force in the beam, which arises because of the composite behaviour of two offset elements.

If you have modelled the slab as being at the same nodes as the beams without plate offsets, then the slab won't be acting as a top-flange for the composite beam, as is generally the assumption for composite slab-on-beam design; that is, it will not tend to pick up any in-plane compressive bending stresses at the mid-span, as it lies on the composite beam neutral axis, instead of being on the top flange. (The effective section resembles a cruciform instead of a Tee.)

http://julianh72.blogspot.com

 

[IDS]

jhardy1 - Whilst modelling the slab at the correct offset from the girder centroid will give a more accurate model it does make extracting the results considerably more difficult, and it doesn't make a huge difference to the results. Given the magnitude of the difference between the moment results and the expected values I think it would be better to keep the model simple, at least until the cause of the difference is determined. I'd suggest that the main point to check is that the girder I value should be for the composite section, not just the precast (assuming the centroids are all in the same plane). I agree with ukbridge that checking with a grillage model would be a good idea.

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/

 

[jhardy1]

@IDS
I think it can and often DOES make a big difference, whether you model the plate / beam offsets or not.

Consider the following example:
10-metre span 460UB67 with 2-metre wide x 200 mm thick composite topping slab. Analyse for self-weight due to gravity, for two cases (see attached):

a) Beams and plate elements modelled concentrically: Mid-span deflection = 12.5 mm
b) Beams and slabs offset to their correct spatial positions: Mid-span deflection = 4.5 mm

Yes, you can model beam-grillages, using equivalent composite section properties, and get reasonable results for simple cases, but I have experienced problems getting decent results for situations with significant point loads and / or concentrated supports (rather than distributed loads and continuous edge supports). Where there are concentrated loads and / or supports, the slab twisting moments have to be carried as beam torsions in a grillage model, and I find it very difficult to come up with a grillage model that captures such behaviour well.

I personally prefer to model beam-slab composites using beam and plate elements. Yes, interpretation of the results to get the total effective moment is a bit trickier than a simple beam model, but I am much more confident in the quality of my analysis.

http://julianh72.blogspot.com

 

[IDS]

There's no question that if you have the plates and beams in the same plane, then the Beams need to have composite section properties. Quite likely the problem reported in the OP is due to not doing that.

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/