Data Set Size Magnitude 8 75 7.5 7 6.5 6 5.5 5 4.5 Pre 996 Post 996 4 0.0 0. 0 00 000 Distance (km)
Main Changes NGA from 997 Vintage Use of Vs30 Standard deviation independent of magnitude for most models Non-linear site response Hanging wall and Footwall factors Additional predictive parameters Depth to Top of Rupture Depth of Soil Dip (HW factor only)
Generic Rock and Vs ased Generic Rock categories Models Many previous studies had incorrect classification of site Vs30 approx 550 m/s for A&S 997 generic rock Expect a reduction in the ground motion model to account for Vs30 differences If increase Vs30 from 550 to 750m/s 2% decrease for PGA 25% decrease for T= sec
Conclusions from Previous COSMOS Meetings and Importance of Today s Topic Norm Abrahamson Nov 6, 2009
Summary from the 2004 Meeting Large variability of non-linear response of structures from recordings with similar M,R and ground motion level For small number of time series (e.g. 3-7), results sensitive to the selection of the time series No well founded objective criteria for selecting time series Left to judgment Problem is getting worse as the number of recordings grows Can t develop an objective selection criteria until the intended use of the time series is specified. Need more interaction between ground motion analyst and engineer evaluating the structure
Summary from the 2004 Meeting Need to decide if we are after average response or variability of response Most participants agreed we are after the average response given the design spectrum, not the variability of the response The design spectrum already has the return period of the ground motion in it We can do better than just randomly selecting records from similar magnitude-distance bin Epsilon value (Cornell s approach) Simplified non-linear system PEER DGML Records selected to capture variability of the response
Summary from 2005 Meeting Time Histories for uilding Code Requirements Intent t Examples of time series selection and modification from different projects Scaling Spectrum compatible
Modification Approaches Scaling multiply Acc(t) by (smallest) factor to meet code requirements Same factor for two horizontal components Spectrum compatible Scale and also adjust the frequency content to Scale and also adjust the frequency content to be consistent with the design spectrum (meet code requirements)
Scaled vs Matched 0 0. Target Matched Scaled 0.0 0.0 0. 0 Period (sec)
Selecting Good Recordings Selecting only based on seismological properties leads to large variability Choosing good records suppresses variability
Summary from 2005 Meeting What do we do with a record that leads to unacceptable performance? Ignore it. Just consider average response Consider it Require structure to pass for all records This corresponds to an increase in the return period of the ground motion No common basis for comparing methods For 2006, apply multiple methods for the same structure
2006 Meeting: Satisfying Code Requirements Select time histories with similar M, R, mechanism Not sure what features of time histories control response SRSS of two horizontal components are not more than 0% below.3 DE in period range 0.2T 0 to.5t 0 Which spectral periods to check? Equal spacing on a log period scale 00 periods per decade Avoids missing a trough in the spectrum Other periods will be < 5% lower
Summary from 2006 Meeting Non-linear response is sensitive to the selection of the time histories Large variability from the recordings with similar M,R est approach for selecting and modifying time histories depends on what we want to get out of the analyses Average response Variability of response Strongly held opposing opinions on different approaches and objectives
Selection Approaches Seismological Properties Similar Mag, Dist, Mech Goal: capture key unknown characteristics of ground motion that are important to the structural response Recording Properties Wider Mag, dist, mech Identify key characteristics of ground motion that are important to the structural response E.g. spectral shape, pulses, duration, Select recordings that sample the key characteristics
Variability Reducing Variability Seismological Properties 2.8 Full range for design spectrum Seismological Properties.6 Good Records.4 Spectrum Compatible.2 Good Records 0.8 0.6 Spectrum Compatible 0.4 02 0.2 0 0.000 0.00 0.0 0. Structural Response (e.g. Drift)
Sets of 7 Time Histories What do we do with the variability of response? Require structure to perform well for all 7 Use the mean+sigma (84th) response Just use the average response, ignore variability Replace bad records Using above average recordings is equivalent to increasing the return period of the design ground motion Why put conservatism in the time histories and not in the return period
Objectives for Use of Time Histories For fragility (PEE) Need full variability Can t use spectrum compatible or subset of good records 7 is a small number to define variability For meeting code requirements Average response: reduce variability for more stable estimate of the average response Can use spectrum compatible or good records To learn something about structure What you learn is at the mercy of the selection of the records
2006 Meeting Summary No consensus in best approach Difficulty is due to the different objectives Mean response Variability of response Given an objective, we can do better than randomly selecting based on seismological properties Requires considering the non-linear behavior of the structure
2007 Meeting Summary Review of selection and modification methods used for performance-based engineering ATC-58 ATC-63 PEER GMSM Working Group results Testing of multiple methods for record selection and modification Test methods for 3 buildings
2007 Conclusions Use of CMS for uildings Realistic spectrum given a single spectral point on the UHS Avoids the enveloping characteristics of the UHS Leads to more accurate estimate of the structural response
2008 Meeting Consider record selection for geotechnical problems, not just buildings Optimal small set of recordings may be different for geotechnical problems than for building response Conditional mean spectrum focused on building period What are key features of ground motions for geotechnical problems? What about projects with both geotechnical and structural parts? Conclusion Improved estimates of the response if record properties are considered, not just seismological properties
Importance of Todays Program Move to the Maximum Rotated Component How are 2-component time histories developed using the Ma Rotated Spectrum? Immediate issue for OSHPD projects Issue for ASCE 07-0 changes
Determining Max Rotated Ground Motion models Spectrum Most ground motion models are for the average horizontal Scale factors available to adjust from average to maximum rotated Impacts on median (easy to apply) Impacts on standard deviation (more difficult to apply) Directivity Effects Should directivity effects be applied if the max rotated component is used? Impacts using old and new directivity it models
CMS from Maximum Rotated Spectra Previous COSMOS meetings Using records that match the CMS provide a better estimate of the performance than just using seismological properties Issues of Realistic Spectra UHS is an envelop of multiple earthquakes Max rotated spectrum is envelop (at multiple directions) of the envelope (UHS) The max Rotated spectrum is moving away from use of more realistic ground motions
CMS from Maximum Rotated Spectra Methods to compute CMS are from average horizontal UHS How to compute CMS from Max rotated spectrum? What should CMS represent? Max rotated only at T0, same component at other periods? Different set of correlation coefficients are needed
Time Series Modification Modification of Two Horizontal Components Scaling How are records scaled? Scale to meet the maximum rotated spectrum Scale so SRSS meets.0 times max rotated target» Computed max rotated will be lower than max rotated target Spectral matching What spectrum is used for matching the two horizontal components? Max rotated ttd and Min rotated ttd Average horizontal Max rotated/sqrt(2) What to use outside the critical period range? What requirements (same issue as scaling)
Example Ave SRSS meets.0 Max Rotated target Need to scale by.66 to get calculated max rotated to meet 0.9 Target Is this additional scaling Needed?