UCLA

Strong motion building response data has been shown to be useful for supporting decision-making in the post-earthquake environment. However, most of the prior research in this area has focused on individual buildings. Moreover the few studies that have addressed regional impact assessment using strong motion data have only focused on structural response reconstruction. In other words, stakeholder-driven performance metrics such as repair costs from earthquake-induced damage have not been addressed. Another major challenge with prior research stems from the mostly low-to-moderate response demands that are present in the measurements from prior earthquakes. This research seeks to advance regional seismic impact assessment using strong motion data by addressing the aforementioned limitations and other gaps in the research. The first major issue that is addressed is the lack of a comprehensive and easily accessing database of building strong motion response recordings. To this end, a relational database of strong motion responses from buildings across California subjected to historical earthquakes is addressed. In addition to the response measurements, various characteristics of the buildings and seismic events are included in the database to support predictive modeling. As a compliment to the database, a python-based tool that enables ease of access and streamlined queries is also developed. Because all of the buildings in the constructed database are only partially instrumented, within a given structure, only a subset of floors have response measurements. Moreover, within a given inventory, only a subset of the buildings will be instrumented. To deal this challenge, a structural response reconstruction (SRR) model is developed for use in a regional impact assessment context. This model takes in response measurements in partially instrumented buildings as well as structural and seismic event parameters and estimates the full set of responses for all the building in the inventory. A dual modeling approach is utilized, which initiates with using kriging to estimate the peak ground acceleration at the location of the uninstrumented buildings. Then, the extreme gradient boosting algorithm is used to reconstruct the EDPs across all buildings in the inventory. This study raised questions about which ground motion intensity measures (IMs) are most suitable for use as inputs in SRR models. To further probe this question, the strong motion data from the database was used to evaluate the effectiveness of several IMs. An important distinction from prior IM evaluation studies is the use of strong motion response data. In other words, most prior studies on assessing IM effectiveness used data from response history analysis.

Because the input data comprised mostly linear elastic (or near-elastic responses), the applicability of the SRR model was limited to use in low-to-moderate-sized earthquakes. To increase the generalizability of the SRR model, the dataset of measured EDPs was augmented with responses obtained from nonlinear response history analyses (NRHAs). Through this augmentation, the range of structural responses was increased, particularly in terms of nonlinear responses. To facilitate this data augmentation, an end-to-end framework was developed to automate the generation of simulated structural responses in an inventory that includes multiple lateral force resisting systems (LFRS). Steel moment frames, woodframe shear walls, reinforced concrete shear walls and steel braced frames are the lateral systems that were included in the simulated response development. The benefits of the augmented dataset was demonstrated by comparing the performance of SRR models developed using (i) only the measured responses and (ii) the combined simulated and measured responses. Lastly, the SRR model based on the combined dataset was used to perform a regional impact assessment of a building inventory using the earthquake-induced repair costs as the performance metric of interest.