Newmark Deformation Analysis Procedures

[moe333]

When developing time histories for a Newmark displacement analysis, it was my understanding that the seismic site response analysis is performed relative to the critical failure surface which does not incorporate a pseudo-static coefficient? Can anyone confirm or refute this? This critical failure surface should however incorporate dynamic shear strengths.

The time histories (actually HEA time histories) developed along this critical failure surface (which does not incorporate a pseudo-static coefficient) are then integrated twice to estimate deformation. It would follow that the critical failure surface for estimation of the yield acceleration may be different than the failure surface where the site response was performed.

I have seen some references that indicate that the seismic site response analysis is performed relative to the critical failure surface from which the yield acceleration is calculated? However I believe this is not correct.

Thanks

 

[dgillette]

I.M. Idriss's most recent version of QUAD4M (2D FEM equivalent-linear response analysis, a DOS program unfortunately) has a feature that allows one to convert the force time history along the failure surface into an equivalent acceleration time history to be input into the Newmark analysis. He does it that way because the slide mass is not actually rigid and, at least in theory, the centroid of the slide mass could correspond to a node for some of the more critical vibration modes. My hunch is that it's a minor issue, and you could use the acc time history from the centroid of the slide mass without any real loss of accuracy. (You are estimating deformation of a pile of dirt being shaken by ground motions that are really only an educated guess.)

If you're dealing with a dam, try to get ahold of one of Jim Swaisgood's papers on case histories of settlements. He collected a large number of case histories and plotted settlement (normalized by height) as a fn. of PHA and M. I don't really like the way he fit equations to the data, but the data are quite valuable as a reality check. In summary, a rolled fill dam on a good foundation won't settle more than a couple of feet in some pretty severe loadings. (Liquefiable sand or soft clay in the fndn would make for a different story.)

The Makdisi-Seed method (similar to the chart solution Newmark published in 1965 +/-) incorporates elastic response of the embankment only, assuming the dam is on bedrock and not a soil foundation. It's also a good quick check.

That help any?

Regards,
DRG

 

[moe333]

DRG, thanks for the reply. The use of HEA time histories I mentioned is the same feature that you describe from QUAD4M. As you say, it accounts for the failure mass not being rigid. This is a ”decoupled” analysis which assumes that the dynamic response of the failure mass is not influenced by permanent displacement that occurs on the failure surface. This decoupled analysis is appropriate for particular ratios of the period of the slide mass and the input ground motion.

After more reading, I think I may be changing my original opinion in that I think it is the failure surface from which the yield acceleration is calculated. As you say, it may be a minor issue, I'm not sure, maybe I'll take a look at the QUAD4M manual to see if I can glean anything from that.

I am trying to apply this modified Newmark method to a dam that has localized liquefiable zones. The seismic stability is adequate based on analyses with residual strengths for liquefiable zones. I realize most researchers state that this method is not appropriate for liquefiable soils, however, it does seem to be used for my specific condition probably because the alternative is a very complex, expensive, nonlinear effective stress FEA.

I'm thinking estimated deformations will be large since the calculated yield acceleration will be low due to the use of residual strengths for liquefiable zones. In addition to the deformations of the Newmark analysis, I will have to include volumetric compression by Tokimatsu and Seed.

One thing I'm considering is using a degrading yield acceleration to account for the liquefaction not being present throughout the entire time history. Another issue is the seismic site response analysis will be performed with an equivalent-linear total stress method that will not accurately model liquefiable zones. However, I think this will be conservative since the liquefiable zones should dampen the response.

As you state, I am estimating deformation of a pile of dirt being shaken by ground motions that are really only an educated guess, however the amount of research that has gone into this type of analysis is staggering and I'm trying to make the best use of it. I will be checking the results with the various simplified methods as well as Jim Swaisgood's papers.

 

[dgillette]

QUAD4M is so labor intensive that it might actually make sense to have a couple of FLAC runs done for you. You're pushing the assumptions pretty far out there with applying Newmark and QUAD4M with stuff that may liquefy.

How much freeboard are you starting with, and how big is your EQ?

I'd bet on the volumetric settlement being just peanuts compared to the dynamic deformation settlement.

 

[moe333]

Don't cringe but the plan was to use 1D site response at several sections along the failure surface. I am trying to at least use non linear eff. stress 1D (DMOD) for this but may not be able to. I agree that FLAC is the way to go. Very prelimminary, no info on freeboard or magnitude, design PGA is 0.3g.

Only a part of the Newmark settlement would be vertical as it is tangent to the failure surface.

 

[jheidt2543]

moe333:

Your topic is out of my area of work, but I did find some information at the following link that may be of some help to you:

http://pubs.usgs.gov/of/1998/ofr-98-113/ofr98-113.html

It is the following paper:

"A Method for Producing Digital Probabilistic Seismic Landslide Hazard Maps: An Example from the Los Angeles, California, Area"
By
Randall W. Jibson, Edwin L. Harp, and John A. Michael