Combined soils and structure model

[ludvik]

A major part of structure's behavior under EQ conditions is in the foundations, yet structural models represent their foundations very crudely.

I am currently working on a model of a bridge where the first mode of vibration is essentially the structure is remaining rigid, and the entire mass is moving horizontally - ie, the foudations are deflecting a lot. My foundations are modeled with a 6x6 single node spring matrix.

In this instance, the behavior of the bridge is strongly dependent on the properties of the foundations. It is not traditional to model any mass or damping in the foundations for such a model, though in this case, do you think this might be justified?

Further, soils are non linear and as such should provide a reasonable amount of damping in a high displacement scenario. If you were doing a non-linear time history analysis, surely it would make sense to model some of this soil non-linearity.

Michael

 

[JAE]

Just went to a seminar by ASCE on seismic. A resource book given for this issue was:

A. S. Veletsos, "Dynamics of Structure-Foundation Systems," Structural and Geotechnical Mechanics, a volume honoring N. M. Newmark, W. J. Hall, Ed., Prentice Hall, Englewood Cliffs, NJ, 1977.

I would agree that with a bridge, especially in the longitudinal direction, the effects of the soil-structure interaction should be included. I believe it is quite complex, though, and not too much data out there to help you.

 

[ludvik]

Thanks for the tip about the book JAE.

I don't view the problem of including non linear foundations in the same model as the above ground structure as that fantastically difficult.

You could do one of two things:

1. You could explicitly model the soil and piles using brick elements for the soil and beam elements for the piles - this is not that hard. Other simpler soil models could even be used. You may need a little advice from a geotechnical engineer in regards to the soil conditions and water tables, but its pretty straight forward. The only problems would be increased run times and post processing the soil results.

2. You could represent foundations as a single node element which models foundation mass (including tributary soil mass), foundation stiffness, and damping. Again, this is no great leap forward of technology. You don't need different software, just a slightly different approach.

A non-linear static soil analysis is commonly completed seperately to provide foundation stiffness for the structural model. I am simply suggesting that these two analyses be merged.

My feeling is that the reason this is not done is because, until recently, processing capacity was at a premium. Engineers were looking for the simplest model possible. In the age of $2000 Pentium 4's, we have that power.

Michael

 

[Qshake]

Ludvik, I understand completely what you are stating. However, as a practicing structural/earthquake engineer with many years of experience I can tell you that earthquake engineering is still relatively young. While much of what we do as engineers has been around for many centuries, the earthquake "problem" didn't attract much interest until the 1908 Messina event and wasn't significantly upgraded until after 1933 Long Beach quake. Processing time has greatly increased in the last ten years, but for much of that time the response spectra, developed in part by George Housner, remains a very valuable tool for enveloping the response of structures to vibrations. Non-linear soil analysis with linear elastic response spectra is apples and oranges.

So computing/processing time isn't an issue. The real issue lies in having the young talent to apply what they are learning in the way of dynamics/soil/structure interaction wherein the demand for complex algorithms will be born. All ready there is demand for the displacement based analysis - you're seeing more and more programs that profess to perform non-linear push-over analysis. I say profess as some only include geometrical non-linearities and that, as we know, is only half the equation.

Most engineers that I work with all employ soil-structure interaction to a certain extent. Many use the documents noted by JAE; Lam and Marin; and Gazetas. And depending on the size and importance of the structure we will perform a time history analysis and employ appropriate non-linear analysis. Yet for simple structures (three span bridges) we are perfectly content with leaving the technology on the board and use a simple elastic response spectrum analysis. A retrofit or seismic design should be run between 5%-10% of the construction cost for the structure. And it can be time consuming and consequently costly to have a rigorous geotech report, a hybird time history, not to mention the data collection and processing. Those are just a few things that come to mind as more costly than processing time.

In conclusion, I respectfully disagree that engineers are looking for simple models to reduce processing time. Engineers were/are looking for methods that appeal to a broad group and are cost effective to the client. Those of us that work on large scale projects (buildings or bridges) do employ more sophisticated methods. Simply put, our employers are willing to pay for the technology and engineers who understand that technology. You'll do well and will be part of the next level that earthquake engineering can achieve.