Earthquake Engineering GE CEE 479679 Topic 16. Predition of Ground Motions, and Other Topics in Stro

1
Earthquake EngineeringGE / CEE - 479/679Topic
16. Predition of Ground Motions, and Other
Topics in Strong Motion Seismology

• John G. Anderson
• Professor of Geophysics

2
Ground Motion Prediction Equations

• AKA Regressions
• Regression analysis is the mathematical process
  used to determine the coefficients in the
  equations.
• AKA Attenuation Relations
• Attenuation more properly refers to the distance
  dependence of ground motion for a single
  earthquake, or perhaps a single magnitude.

3
Some practical aspects of strong motion seismology

• Access to data
• Exceptional data
• Data processing

4
Data Access

• Most important sources of strong motion data
• COSMOS (http//www.cosmos-eq.org/)
• Japan K-Net and KiK-Net (http//www.k-net.bosai.go
  .jp/ http//www.kik.bosai.go.jp/)

5
(No Transcript)
6
(No Transcript)
7

• Exceedance rate per instrument-year
• PGA 1 g, 2x10-4, or once in 5000 inst-years
• Distribution decreases faster than a power law

8

• Exceedance rate per instrument-year
• PGV 100 cm/s, 2.5x10-4, or once in 4000
  inst-years
• Distribution decreases faster than a power law

9
Nature of data

• Background noise
• Need for baseline correction
• Access to data

10
Izmit, Turkey M7.6, Aug 17, 1999 1 g full scale
Timing marks, 2/sec
L
V
T
Trigger time, initial P-wave is missing
Trailer from the previous trigger.
Three components of ground motion, L parallel to
long axis of the instrument, V vertical, T
parallel to the short axis of the
instrument. Zero is not known exactly.
11
Izmit accelerogram
Continued
12
Digitization This is a first attempt to digitize
this accelerogram.
Discontinuity where the record was shifted
13
Integral of records from the previous page to
velocity.
Note low frequency behavior on the accelerogram
that almost certainly has nothing to do with the
behavior of the ground motion. Conclusion
digitization and baseline correction is not a
trivial matter.
14
Filter

• A numerical process that is applied to a time
  series (in this case an accelerogram). The
  process removes the contributions of certain
  frequencies to the time series.
• High pass filter removes low frequencies (i.e.
  frequencies below the filter frequency ff ), but
  does not affect the high frequencies.
• Low pass filter removes high frequencies (above
  ff ), but does not affect the low frequencies.
• Filter response is generally not sharp. In
  other words, there is a range of frequencies that
  are partly removed.

15
Baseline correction for accelerograms

• Baseline corrections generally filter the
  accelerograms, so that those frequencies where
  the raw signal is dominated by noise are removed
  from the time history.
• The effect of filtering is small on the
  acceleration, as seen on the next three slides.

16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
Filter effects

• High-pass filters generally have small effects on
  accelerations.
• The effect is much greater on velocity and
  displacement.
• Note next three slides all have the same scale.

20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
Comments on baseline correction

• Filter frequencies are often selected on the
  basis of noise models.
• When a record is filtered, signal is removed as
  well as noise.
• If a particular frequency is important to a
  structure, then the accelerogram you use to test
  it should not have that frequency filtered out.
• Modern digital accelerographs require much less
  filtering than older analog accelerographs.

24
Ground-Motion Prediction Equations

• Handout Abrahamson and Silva (1997)
• Next Generation Attenuation Relations (NGA)
• Decade-long project
• Papers now posted on the web
• http//peer.berkeley.edu/products/rep_nga_models.h
  tml
• Database (flatfile) is also on that web site

25

• At this point, I switched to the presentation by
  Ken Campbell to an NGA working group, to give a
  sense of the issues NGA is working on.
• NGA is the state-of-the-art in ground motion
  prediction equations.
• I think papers on the topic will come out in the
  February 2008 issue of Earthquake Spectra.