|
Titel |
Advance in prediction of soil slope instabilities |
VerfasserIn |
C. Sigarán-Loría, R. Hack, J. D. Nieuwenhuis |
Konferenz |
EGU General Assembly 2012
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 14 (2012) |
Datensatznummer |
250071163
|
|
|
|
Zusammenfassung |
Six generic soils (clays and sands) were systematically modeled with plane-strain finite
elements (FE) at varying heights and inclinations. A dataset was generated in order to
develop predictive relations of soil slope instabilities, in terms of co-seismic displacements
(u), under strong motions with a linear multiple regression. For simplicity, the seismic loads
are monochromatic artificial sinusoidal functions at four frequencies: 1, 2, 4, and 6 Hz, and
the slope failure criterion used corresponds to near 10% Cartesian shear strains along a
continuous region comparable to a slip surface.
The generated dataset comprises variables from the slope geometry and site
conditions: height, H, inclination, i, shear wave velocity from the upper 30 m,
vs30, site period, Ts; as well as the input strong motion: yield acceleration, ay
(equal to peak ground acceleration, PGA in this research), frequency, f; and in some
cases moment magnitude, M, and Arias intensity, Ia, assumed from empirical
correlations. Different datasets or scenarios were created: “Magnitude-independent”,
“Magnitude-dependent”, and “Soil-dependent”, and the data was statistically explored and
analyzed with varying mathematical forms. Qualitative relations show that the permanent
deformations are highly related to the soil class for the clay slopes, but not for the
sand slopes. Furthermore, the slope height does not constrain the variability in
the co-seismic displacements. The input frequency decreases the variability of the
co-seismic displacements for the “Magnitude-dependent” and “Soil-dependent”
datasets.
The empirical models were developed with two and three predictors. For the sands it
was not possible because they could not satisfy the constrains from the statistical
method. For the clays, the best models with the smallest errors coincided with the
simple general form of multiple regression with three predictors (e.g. near 0.16
and 0.21 standard error, S.E. and 0.75 and 0.55 R2 for the “M-independent” and
“M-dependent” datasets correspondingly). From the models with two predictors, a 2nd-order
polynom gave the best performance but with a not-significant parameter. The best
models with both predictors significant have slightly larger error and smaller R2, e.g.
0.15 S.E., 44% R2 with ay and i. The predictive models obtained with the three
scenarios from the clay slopes provide well-constrained predictions but low R2,
suggesting the predictors are “not complete”, most likely in relation to the simplicity
used in the strong motion characterization. Nevertheless, the findings from this
work demonstrate the potential from analytical methods in developing more precise
predictions as well as the importance on treating different different ground types. |
|
|
|
|
|