CSMIP Strong Motion Data

marblefreedomAI and Robotics

Nov 14, 2013 (3 years and 9 months ago)

67 views

May 25, 2004

CSMIP Processing, Shakal et al

1

CSMIP Strong Motion Data
Processing


Anthony Shakal, Moh Huang and Vladimir
Graizer


California Strong Motion Instrumentation Program

California Geological Survey (was CDMG)

Sacramento, California

May 25, 2004

CSMIP Processing, Shakal et al

2

CSMIP Processing Development



CSMIP began joint processing project with USGS
in late ‘70s (film scanning by Towill Co.
-

software
devel. and processing at Lawrence Berkeley Lab)


In early ‘80s standalone processing at CSMIP


Scanning system installed patterned after that
developed at Univ. Southern Calif. by Trifunac & Lee


Processing software of Caltech Bluebook project
(Hudson et al) as modified by Trifunac & Lee


Software upgraded for production, with noise
level improvement at CSMIP

May 25, 2004

CSMIP Processing, Shakal et al

3

Uniform Processing


Guiding filter period
selection based on Signal and Noise Spectrum
(Trifunac, 1977)

Digitized accelerogram as sum of desired
acceleration and background noise

May 25, 2004

CSMIP Processing, Shakal et al

4


Signal spectrum moves up and to
the right with increasing
magnitude



Noise spectrum controlled by


digitization (by film scanner,
or by A
-
to
-
D converter)


sensor properties



Initial filter corner estimate
-

above & left of junction,
an SNR 2 or 3



Final corner guided by time domain output of suite of
runs having filter near this period

May 25, 2004

CSMIP Processing, Shakal et al

5

Current CSMIP procedure is to use one
filter corner for all components



Pro: Multi
-
dimensional aspects can be
studied by end user


Particle motion


Torsional response in structures


Con: Period controlled by the lowest
signal/highest noise channel (often vertical,
or lowest n building)

May 25, 2004

CSMIP Processing, Shakal et al

6

Steps in processing (analog)

1.
Baseline correction


minimal (remove mean;
perhaps remove slope)

2.
Instrument correction

3.
High
-
frequency filtering (25 Hz Ormsby
classically)

4.
Initial integration & long period filtering

5.
Maximum
-
bandwidth response spectra

6.
Time
-
history suite for long
-
period filter
selection

7.
Final product preparation

May 25, 2004

CSMIP Processing, Shakal et al

7

Example


Whittier analog record


Digitize at 200 points/cm


Digitize two fixed traces for reference
-
trace subtraction to
remove film shift (earthquake) and film drift (canister) problems


Use 2 pulse/sec time trace to correct film
-
speed change errors


Digitize fiducial marks placed on film to control multiple
-
panel
concatenation

May 25, 2004

CSMIP Processing, Shakal et al

8

Example


Digitized (Vol. 1)


Digitized, time
-
corrected, 200 pts/sec, scaled by
sensitivity


Match
-
test to film image, check for offsets, drifts,
panel
-
junction effects

May 25, 2004

CSMIP Processing, Shakal et al

9

Response Spectrum


Response spectrum
(Nigam
-
Jennings) of
Vol.1 data


‘Wide
-
open’ bandwidth


Compare the long period
decay of signal spectrum
with long
-
period noise


Initial corner estimate

May 25, 2004

CSMIP Processing, Shakal et al

10

Suite of time
-
histories


Range of corners from 12 second to 2.5 sec period


Period chosen was 3.5 second

May 25, 2004

CSMIP Processing, Shakal et al

11

Final Accel, Veloc, Displ


May 25, 2004

CSMIP Processing, Shakal et al

12

Final Spectrum


Filter corners
given on plot


Plotted only out
to corner selected

May 25, 2004

CSMIP Processing, Shakal et al

13

Usable Data Bandwidth


3 dB (half
-
power) points (whether Ormsby,
Butterworth or other filter) define UDB for user


User assumed to be knowledgeable, but not
necessarily in data processing

May 25, 2004

CSMIP Processing, Shakal et al

14

Digital Records


Frequency domain processing


Noise level controlled by A
-
to
-
D
converter’s effective number of bits (last
bits often noise)


In general, more dynamic range (72, 96 dB,
or more vs ~50
-
60 dB)


Sensor

noise/drift more critical


the next
focus in getting the most from recorded data

May 25, 2004

CSMIP Processing, Shakal et al

15

Automatically Processed Record

Record processed automatically at the time of the earthquake
(2 am Sunday morning, May 9, M4.4 off Santa Barbara).

May 25, 2004

CSMIP Processing, Shakal et al

16

CISN Internet Quick Report


tied to automatically
generated ShakeMap

May 25, 2004

CSMIP Processing, Shakal et al

17

Summary (1)


CSMIP processing evolved from the Caltech
Bluebook project as extended at USC


Nearly 1000 records digitized/processed (1000s of
traces)


General approach goal is to release as much signal
as possible, with as little noise accompanying the
signal as practical


Using one filter corner per record means that results
may be:


more conservative than another policy would yield; but


less difficult to use, for most users

May 25, 2004

CSMIP Processing, Shakal et al

18

Summary (2)


No acausal filters routinely used; processing of special
problem or offset records done by hand, on case
-
by
-
case
basis, or restricted band pass provided


Automatic processing done by straightforward processing,
with diagnostic checks/set
-
asides if potential problems (DC
shifts, electronic noise events, etc)


Automated processing and Internet Quick Report providing
rapid release for response and post
-
earthquake engineering
evaluations.