PPEEEERR
EARTHQUAKE SCIENCE-ENGINEERING INTERFACE:
STRUCTURAL ENGINEERING RESEARCH PERSPECTIVE
Allin Cornell
Stanford University
SCEC WORKSHOP
Oakland, CA
October, 2003
PPEEEERR
Objective
• λC = mean annual rate of State C, e.g., collapse
• Two Steps: earth science and structural engineering: λC = ∫PC(X) dλ(X)
• Where X = Vector Describing Interface
PPEEEERR
“Best” Case• X = {A1(t1), A2(t2), …Ai(ti)..}
for all ti = i∆t , i = 1, 2, …n
i.e., an accelerogram
·dλ(x) = mean annual rate of observing a “specific” accelerogram, e.g.,
a(ti) < A(ti) <a(ti) + da for all
∙Then engineer finds PC(x) for all x
∙ Integrate
PPEEEERR
Current Best “Practice” (or Research for Practice)
• λC = mean annual rate of State C, e.g., collapse
• Two Steps: earth science and structural engineering: λC = ∫PC(IM) dλ(IM)
• IM = Scalar “Intensity Measure”, e.g., PGA or Sa1
• λ(IM) from PSHA
• PC(IM) found from “random sample” of accelerograms = fraction of cases leading to C
PPEEEERR
Current Best Seismology Practice*:
·Disaggregate PSHA at Sa1 at po, say, 2% in 50 years, by M and R: fM,R|Sa. Repeat for several levels, Sa11, Sa12, …
· For Each Level Select Sample of Records: from a “bin” near mean (or mode) M and R. Same faulting style, hanging/foot wall, soil type, …
· Scale the records to the UHS (in some way, e.g., to the Sa(T1)).*DOE, NRC, PEER, … e.g., see R.K. McGuire: “... Closing the Loop”( BSSA, 1996+/-); Kramer (Text book; 1996 +/-); Stewart et al. (PEER Report, 2002)
PPEEEERR
241241 241241 241241
157157
105105
105105
105105
106106
105105
105105
•Beam Column Model with Stiffness Beam Column Model with Stiffness and Strength Degradation in Shear and and Strength Degradation in Shear and Flexure Flexure using DRAIN2D-UW by J. Pincheira et using DRAIN2D-UW by J. Pincheira et al.al.
Seismic Design Assessment of RC Seismic Design Assessment of RC Structures.Structures.
( (Holiday Inn Hotel in Van NuysHoliday Inn Hotel in Van Nuys))
PPEEEERR
Multiple Stripe Analysis Multiple Stripe Analysis
• The Statistical Parameters of the The Statistical Parameters of the “Stripes”“Stripes” are are
Used to Estimate the Median and Dispersion as Used to Estimate the Median and Dispersion as
a Function of the Spectral Acceleration, Sa1.a Function of the Spectral Acceleration, Sa1.
C
PPEEEERR
Best “Research-for-Practice” (Cont’d) :• Analysis: λC = ∫PC(IM) dλ(IM) ≈ ∑PC(IMk)
∆ λ(IMk)
• Purely Structural Engineering Research Questions:– Accuracy of Numerical Models – Computational Efficiency
PPEEEERR
Best “Research-for-Practice”:• Analysis: λC = ∫PC(IM) dλ(IM)
≈ ∑PC(IMk) ∆ λ(IMk)
·Interface Questions:
What are good choices for IM? Efficient? Sufficient?
How does one obtain λ(IM) ?How does one do this transparently, easily and practically?
PPEEEERR
10-4
10-3
10-2
10-1
100
Van Nuys 7 Story
Maximum Story Drift Angle, max
Sa a
t T1=
0.8
se
con
ds
[g]
a = 0.0198 b = 1.05
ln(max
)|ln(Sa) = 0.2716
60 records with 5.7 < M < 7.3
Van Nuys Transverse Frame:
Pinchiera Degrading Strength Model; T = 0.8 sec.
60 PEER records as recorded 5.3<M<7.3.10-1
100
5.6
5.8
6
6.2
6.4
6.6
6.8
7
7.2
7.4
Sa at T
1=0.8 seconds [g]
Ma
gn
itud
e
Van Nuys 7 Story
a = 6.72 b = 0.127
M|ln(Sa) = 0.4188
PPEEEERR
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 110
-1
100
101
Residual of the regression of magnitude on ln Sa(FMF)
Re
sid
ua
l of t
he
"o
rig
ina
l" r
eg
ress
ion
on
S a (
FM
F)
Van Nuys 7 Story
a = 2.31e-015 b = 0.1443
ln residual|residual2 = 0.2648
b = 0.08303
one-sided p-value = 0.04378
Residual-residual plot: drift versus magnitude (given Sa) for Van Nuys. (Ductility range: 0.3 to 6) (60 PEER records, as recorded.)
PPEEEERR
Residual-residual plot: drift versus magnitude (given Sa) of a very short period (0.1 sec) SDOF bilinear system. (Ductility range 1 to 20.) (47 PEER records, as recorded.)
PPEEEERR
Residual-residual plot: drift versus magnitude (given Sa) for 4-second, fracturing-connection model of SAC LA20. Records scaled by 3. Ductility range: mostly 0.5 to 5
PPEEEERR
What Can Be Done That is Still Better?
• Scalar to (Compact) Vector IM• Interface Issues: What vector? How to find λ
(IM)?• Examples: {Sa1, M}, {Sa1, Sa2}, …
·PSHA:
λ(Sa1, M) = λ(Sa1) f(M| Sa1) (from “Deagg”)
λ(Sa1, Sa2) Requires Vector PSHA (SCEC project)
PPEEEERR
Future Interface Needs
• Engineers: • Need to identify “good” scalar IMs and IM
vectors. • In-house issues: what’s “wrong” with
current candidates? When? Why? How to fix?
• How to make fast and easy, i.e., professionally useful.
PPEEEERR
Future Interface Needs (con’t)• Help from Earth scientists: Guidance (e.g.,
what changes frequency content? Non-”random” phasing? )
· Earth Science problems: How likely is it? λ(X)
λ(X) = ∫P(X \ Y) dλ(Y)
X = ground motion variables (ground motion prediction: empirical, synthetic)
Y = source variables (e.g., RELM)
PPEEEERR
Future Needs (Cont’d)
• Especially λ(X) for “bad” values of X (Or IM).
·Some Special Problems: Nonlinear Soils, Strong Directivity, Aftershocks, Spatial Fields of X.
PPEEEERR
Residual-residual plot: drift versus magnitude (given Sa) for 4-second, fracturing-connection model of SAC
LA20. Ductility range: 0.2 to 1.5. Same records.
PPEEEERR
Non-Linear MDOF Conclusion:
(Given Sa(T1) level) the median (displacement) EDP is apparently independent of event parameters such as M, R, …*.
Implications: (1) the record set used need not be selected carefully selected to match these parameters to those relevant to the site and structure.
Comments: Same conclusion found for transverse components. More periods and backbones and EDPs deserve testing to test the limits of applicability of this illustration.
*Provisos: Magnitudes not too low relative to general range of usual interest; no directivity or shallow, soft soil issues.
Top Related