Maximum strain in equivalent-linear dynamic analysis, e.g., SHAKE or QUAKE/W

[dgillette]

Do any of you know of any publication or industry consensus on the maximum strain for which equivalent-linear dynamic analysis can be considered valid, whether for finding cyclic shear stresses or for predicting ground motion at surface, perhaps in the QUAKE/W manual? (I didn't find it in a quick search of the SHAKE2000 manual.) I'm of the opinion that the assumptions of E-L are out the window before the strain reaches 0.1 percent. I say this because, with such large strains, the shear modulus gets reduced so much for the second, third, etc. iterations that the assumption of constant G through the whole ground motion is no longer close to reality. The greater the strain, the smaller the modulus, and the smaller the modulus the greater the strain, the smaller the modulus, etc. Typical G/Gmax curves drop off to 0.5 well before they get to 0.1 percent strain. With low-Vs material (up to maybe 650'/s) below stiffer material and high bedrock PHA, the later iterations treat it like a linear analysis of a brick on Jello, which I believe gives surface PHA that's too low, and certainly gives cyclic stresses that are too low.

This has become a topic of minor controversy here. Have you run into it?

Cheers!
DRG

 

[moe333]

DRG, I have seen 1% quoted as the limit (DMOD2000 Manual, and DMG 117 by SCEC (liq.)), and acceration >0.4g for soft soil. But I think you have good arguments to use a lower threshold. I bet you could find something in Steve Kramers various reports (and maybe his book).
Cheers!

 

[dgillette]

Thanks, Moe. I figured that if anybody answered, it would be you.

One of our points of contention has centered around low calculated CSR, therefore no liquefaction, even though the strains in the softened material go up to 0.2%. Dobry's Ishihara lecture is a good reference for showing that ain't necessarily so if it ever comes up for you.

DRG