SMIP 2009 Seminar

This paper describes an investigation of the ground motions recorded at the Turkey Flat test site, and of the predictions of those motions in the blind prediction symposium that took place in 2006. The two-phase prediction experiment attracted numerous participants using several approaches to ground motion modeling and site data interpretation. The results of the Phase 1 predictions showed strong consistency in the predicted motions, but significant differences between the predicted and recorded motions. The Phase 2 predictions were also consistent and were also quite accurate. The paper reviews the basic experiment, summarizes the results of the Phase 1 and 2 predictions, examines potential explanations for the differences between the predicted and observed motions, and comments on lessons learned and implications for site response practice.

Methodology for reconstructing the seismic response of buildings from measured accelerations is examined. The popular cubic spline (CS) interpolation is shown to be equivalent to fitting the measured response on a basis whose dimension is equal to the number of sensors and whose span is determined by the sensor positions. The basis fitting perspective makes clear that a necessary condition for accuracy is that the number of sensors be no less than the number of modes that contribute substantially to the estimated quantity. It is shown that low pass filtering of the measured response, using a cutoff frequency in the proximity of the frequency of mode (m-1), where m is the number of sensors, is advisable. Reconstruction by blending the measurements with a nominal model is examined using the Kalman Filter, the RTS Smoother and a new approach designated as the Minimum Norm Response Corrector (MIRC). Results obtained using 3 nonlinear building models and an ensemble of 30 bi-directional earthquake motions suggest that, for the conditions that prevail in practice, (i.e., relatively poor nominal models and possible nonlinearity in the measured response) the MIRC estimator is the most accurate. The gains in accuracy offered by MIRC over the CS are modest for inter-story drift but are significant in story shears. Specifically, the mean of predictions normalized to the true result (based on 1560 story shear histories) proved to be 1.48 for the CS and 1.00 for MIRC.

This investigation focused on developing an improved understanding of challenges associated with computation of nonlinear response of three-dimensional building to recorded ground motions, and if the inertial base shear is an accurate indicator of the true base shear. For this purpose, three-dimensional models of two buildings – one reinforced-concrete building and one steel building – are developed in OpenSees and Perform3D¬. The analysis of these models included pushover analysis for lateral force distribution proportional to the first mode in each of the two principle directions, and RHA to compute response for 30 ground motions recorded during past earthquakes. It was found that modeling assumptions as well as different software may lead to significantly different pushover curves: concentrated plasticity model leads to lower strength, early initiation of yielding, and post yielding strength loss in pushover curves compared to spread plasticity model, strength loss model for beams/columns leads to significant post yielding strength loss in the pushover curve, and differences in solution schemes and convergence criteria available in different software programs also affect the pushover curves. It was also found that there prediction of median peak response from different software can differ from 10% to 40%. Finally, the median inertial base shear exceeds the true base shear by 10% to 20% with the value exceeding by as much as 50% for individual earthquake and even a small time delay between different recording channels may lead to significant error in the inertial base shear. Therefore, inertial base shear should be used with caution as an estimate of the true base shear.

The deployment and retrieval of strong motion records over the past 80 years has provided new insight into the intensity, distribution, and character of strong motion. Gathering, processing, and interpreting the records has also become a key aspect of understanding building response and damage patterns. Specific inspections of the buildings around the 2009 L’Aquila Italy Earthquake strong motion recording stations offers yet another glimpse of how well buildings seem to be responding to strong shaking that appears to exceed the design levels. The observed damage pattern beyond the instrumented locations also offers some indication of the variation of shaking and once again demonstrates the need for significantly more instruments.

This study focuses on the use of strong motion data recorded during earthquakes and aftershocks to provide a preliminary assessment of the structural integrity and possible damage in bridges. A system identification technique is used to determine dynamical characteristics and high-fidelity first-order linear models of four bridges from low level earthquake excitations. A finite element model (FEM) was developed and updated to simulate data from a damaging earthquake for one of the bridges. The difference between data recorded or simulated by FEM and data predicted by the linear model was used to detect damage. The use of this technique can provide an almost immediate, yet reliable, assessment of the structural health after a seismic event.

Seismic hazard for tall buildings in California is often dominated by large magnitude earthquakes for which few recorded accelerograms are available for response history analysis. In several recent manuscripts, we compare motions for an Mw 7.8 event on the southern San Andreas fault (known as the ShakeOut event), two ShakeOut permutations with different hypocenter locations, and a Mw 7.15 Puente Hills blind thrust event beneath downtown Los Angeles, to median and dispersion predictions from the empirical NGA ground motion prediction equations. The dispersion is represented by an intra-event standard deviation term, which is lower than NGA values at low periods and abruptly increases at 1.0 sec due to different simulation procedures at low and high periods. The simulated motions attenuate faster with distance than is predicted by the NGA models for periods under approximately 5.0 sec. This suggests ground motions away from the fault rupture are under-predicted by the simulation. After removing distance attenuation bias, we have found average residuals of the simulated events (i.e., event terms) are generally within the scatter of empirical event terms, indicating that the ShakeOut event is not unusually energetic for its magnitude. The simulated motions have a depth-dependent basin response similar to the NGA models, but also show complex effects in which stronger basin response occurs when the fault rupture transmits energy into a basin at low angle. The motions also indicate rupture directivity effects that scale with the isochrone parameter.

The Center for Engineering Strong-Motion Data (CESMD) has been established by the U.S. Geological Survey (USGS) and the California Geological Survey (CGS) to provide a single access point for earthquake strong-motion records and station metadata from the CGS California Strong-Motion Instrumentation Program, the USGS National Engineering Strong-Motion Program, and the US Advanced National Seismic System (ANSS). This paper briefly summarizes the CESMD functions, describes the new developed features at the Center and gives an update on the data recently added to the CESMD database. Users can now download multiple records from different earthquakes and stations in zip file(s) that are separated by earthquake name and date and recording station. Registered users are notified of new significant earthquakes with strong-motion data. Highlights are added to the Internet Data Reports of significant earthquakes that summarize earthquakes details and give an overview of available strong-motion data. The CESMD Internet Data Reports for major earthquakes provide access to the references, such as papers and reports that were published on the earthquakes. For each earthquake, a summary of records parameters including station location, peak ground acceleration, velocity, displacement, and peak response spectral values are downloadable as a table that can be imported to a spreadsheet. All the functions and features of the Center are organized in a page named “About CESMD”. Also, in this paper major earthquakes with strong ground motion data in CESMD that occurred since the SMIP 2007 seminar, in September 2007 are summarized.