Jennie Watson-Lamprey COSMOS Annual Meeting Technical Session November 9, 2007 1 PEER GMSM Program:...

Jennie Watson-LampreyCOSMOS Annual Meeting Technical Session

November 9, 20071

PEER GMSM Program: Recommendations for Selection and

Scaling of Ground Motions and Evaluation of the ATC-58 Procedures

Jennie Watson-Lamprey

COSMOS Annual Meeting Technical Session

November 9, 2007


November 9, 20072

Collaboration Among Many Researchers

• Program Members: • Norm Abrahamson• Jack Baker• Yousef Bozorgnia• Allin Cornell• Christine Goulet• Curt Haselton

• Erol Kalkan • Nico Luco • Tom Shantz• Jonathan Stewart• Polsak Tothong• Jennie Watson-

Lamprey • Farzin Zareian


November 9, 20073

Collaboration Among Many Researchers

• Program Collaborators: • Arzhang Alimoradi• Paolo Bazzurro• Greg Deierlein• Jamshid Ghaboussi• Mahmoud Hachem• Charlie Kircher• Praveen Malhotra• Frank McKenna• Coleen McQuoid

• Jack Moehle• Farzad Naeim• Maury Powers• Ellen Rathje• Nilesh Shome• Brian Skyers• Gang Wang• Andrew Whittaker• Bob Youngs• Tony Yang


November 9, 20074

GMSM Mission

• To provide guidance and tools to the engineering community on appropriate GMSM methods for nonlinear dynamic analyses.


November 9, 20075

Ground Motion Selection and Modification

• Selection based only on seismological principles leads to large variability of results.


November 9, 20076

Large Variability in Results

• Building Code Mini-Study led by Jack Baker

• ASCE 7-05• 6.5 < M < 7.6• 0 < R < 30 km• Strike slip events only• Target spectrum exceeded as per code

requirements


November 9, 20077

Building Code Method


November 9, 20078



November 9, 20079


• Median estimates are approximately one sigma above the expected median response. – Bias in results– Implied increase in return period of ground motion

• From this small sample there is a factor of 1.5 difference in median estimates.– Large variability of results


November 9, 200710

Ground Motion Selection and Modification

• Selection based only on seismological principles leads to large variability of results.

• There are two solutions to this problem: – Perform a high end analysis that uses

more records– Be smarter about picking records


November 9, 200711

Solution 1. High End Analysis

• Large number of ground motions

• Regression on the results of the nonlinear dynamic analysis

• Coupling the regression with ground motion prediction equations gives a best estimate of structural response


November 9, 200712

Example building results from over 300 ground motions

0.001

0.01

0.007

0.005

0.003

0.002

0.1

0.07

0.05

0.03

0.02

EDP

.1 .07 .04 .02

1 .7 .5 .3 .2 2 3 4

Spectral Acceleration


November 9, 200713

Example building regression

€

ln EDP( ) =

a+ b ln(SaT1) − c ln(SaT1)2

+d ln(Sa2T1) + e ln(Sa2T1)2

+ f ln(SaT2) + g ln(SaT2)2

⎧

⎨ ⎪

⎩ ⎪

⎫

⎬ ⎪

⎭ ⎪

• Identify key record properties that are important to building response - this will be building specific.

• Perform a regression to determine distribution of EDP as a function of the key properties.

• For example: Use spectral values at different periods to model the EDP.


November 9, 200714

Example building best estimate of response

• Scenario Event - magnitude & distance

• Design Event - magnitude, distance & ground motion

• Risk Based - full hazard curve for an engineering demand parameter


November 9, 200715

Example - Scenario Event


November 9, 200716

Example - Design Event


November 9, 200717

Example - Risk Based


November 9, 200718

Example - Risk Based


November 9, 200719

High End Analysis

• More time consuming

• Highly accurate estimate of both median and variability of structural response


November 9, 200720

Solution 2. Smarter Selection

• A large number (100s) of records is impractical in many cases

• We want a method of ground motion selection and modification that has a higher degree of accuracy using only a small number of records– Median– Variability - not discussed today

• In order to do this we have to know something about the building so we know what’s important for response


November 9, 200721

Solution 2. Smarter Selection

• Objection: The ground doesn’t know the building is there. The motions should be the same no matter what the building is.


November 9, 200722

Solution 2. Smarter Selection• There is a distribution of record properties

(Dur, PGV, RMS) for a given design event. • These record properties affect the nonlinear

response of structures.


November 9, 200723

Solution 2. Smarter Selection• IF we could sample the entire distribution of all record

properties with the ground motions then the set wouldn’t need to depend on the building.

• BUT when using a small number of ground motions it is impossible to capture so many record properties at once.

• THEREFORE we have to know which ones are important and select based on those important properties.

• In order to do this we have to know something about the structure.


November 9, 200724

COSMOS 2006 Results

• There are many simplified methods for selecting a small set of records.

• Depending on the method used there can be a factor of 3 difference in your results.


November 9, 200725

GMSM Mission

• To provide guidance and tools to the engineering community on appropriate GMSM methods for nonlinear dynamic analyses.

• To do this we need to evaluate the simplified methods and determine which ones produce the desired response.


November 9, 200726

GMSM Evaluation Procedure

• Identify an application for which the ground motions will be employed.– 4 structural models, 4 - 20 stories– Curt Haselton will discuss

• Identify the candidate GMSM methods.– 46 methods are now being looked at– Draft summaries of the methods are

available on the GMSM website


November 9, 200727


• Identify the goal of the application.– The building code made me do it– Learn something about the building– PBEE - defined distribution of EDP

• We start with median EDP | M, R, Sa - this is the building block for other more complicated analyses


November 9, 200728


• Calculate a best estimate of response using a high end analysis to serve as the point of comparison (POC)

• Compare the results from each GMSM method with the POC


November 9, 200729

GMSM Results 2007

• Overview of Results - Curt Haselton– Bias in results can be reduced by taking

into account spectral shape.– In order to reduce the variability between

suites of motions the selection must be optimized based on record properties that are important for nonlinear response.

– This means we have to know something about the structure.


November 9, 200730

GMSM Results 2007

• Evaluation of ATC-58 35% - Farzin Zareian– Time-based selection method

overestimates median response for above average ground motion because it ignores spectral shape.

– Time-based selection has variability in median estimates of response because the motions are selected from an M, R bin.


November 9, 200731

GMSM Results 2007

• Evaluation of ATC-58 35% - Farzin Zareian– Scenario-based selection method

overestimates the variability of response because it double counts the ground motion variability at periods other than T1.


November 9, 200732

GMSM Results 2007

• Wrap up - Nico Luco


November 9, 200733

Thank You

Jennie Watson-Lamprey COSMOS Annual Meeting Technical Session November 9, 2007 1 PEER GMSM Program:...

Documents

Transcript of Jennie Watson-Lamprey COSMOS Annual Meeting Technical Session November 9, 2007 1 PEER GMSM Program:...