SMIP 2018 Seminar

The 2016 MW 7.8 Kaikoura earthquake in New Zealand may have involved co-seismic slip on more than 10 distinct faults. We attempt to assess the sequence of rupture on the segments as well as the contribution of the different segments to the recorded ground motion in New Zealand. First, we approximate the segments as points sources to determine the temporal sequence of faulting and to determine the relative contribution to the seismic moment. We reduce the overall number of segments to 10 crustal faults. We invert strong motion, geodetic and teleseismic body and surface wave data to provide a spatio-temporal map of slip and rupture time.

We investigate benefits of regionalizing basin response in ergodic ground motion models. Using southern California data, we find average responses between basin structures, even when the primary site variables used in ground motion models (VS30 and depth parameters) are controlled for. For example, the average site response in relatively modestly sized sedimentary structures (such as Simi Valley) are under-predicted at short periods by current models, whereas under-prediction occurs at long periods for larger sedimentary structures. Moreover, site-to-site within-event standard deviations vary appreciably between large basins, basin edges, smaller valleys, and non-basin (mountainous) locations. Such variations can appreciably impact aleatory variability. ​​​

This research investigated the validity of accidental torsion provisions in ASCE-7 for buildings that are regular in plan and elevation with rigid diaphragms. MDOF systems of 4-, 8-, 12- and 20-story building prototypes along with plan aspect ratio of 1:1, 1:2, 1:4 and 1:8 are modeled. The building models possess translational to rotational period ratios (Ω) ranging from 1.1 to 2.0. Uncertainty in stiffness is treated as the main source of eccentricity. Equivalent design eccentricity indicates that the 5% equivalent eccentricity rule is adequate to capture the median magnification in deformation due to accidental torsion. 11 buildings selected from CSMIP database are studied to verify the analytical results.​​

We quantify the effects of dynamic soil-structure-interaction on building structures using system-identification techniques and finite element simulations. We develop analytic expressions for distributed spring and dashpot elements at the soil-foundation interface in terms of dimensionless variables. A system-identification approach based on Extended-Kalman-Filter is employed to estimate the true soil impedance as seen from the building-foundation system. The impedances estimated are next used to span the range of applicability of the proposed soil impedance model using nonlinear curve-fitting. We find good-agreement between the proposed flexible-based-model and the full finite element-model in period lengthening, radiation damping, time-history responses and their frequency contents.​​

​​In a seismic hazard analysis (SHA), the earthquake loading level should be predicted for one or more ground motion intensity measures (IMs) that are expected to relate well with the engineering demand parameters (EDPs) of the site. In this particular study, the goal was to determine the IMs that best relate to embankment dam deformations based on non-linear deformation analysis (NDA) results of two embankment dams with a large suite of recorded ground motions. The measure utilized to determine the “best” IM was standard deviation in the engineering demand parameter (e.g., deformation) for a given IM—𝜎_{ln𝐸𝐷𝑃|ln𝐼𝑀}, also termed “efficiency.” Results of the study demonstrated that for the NDA model used, Arias intensity (AI) was found to be the most efficient predictor of embankment dam deformations. In terms of spectral acceleration (SA)-based IMs, the SA at short periods and then in the general range of the natural period of the dams were seen to be the most efficient IM, but was in almost all cases not as efficient as AI. In terms of total standard deviation (𝜎_{ln𝐷𝑌𝐹|𝑀,𝑅,𝑆}) of EDP conditioned on earthquake source parameters, the poor predictability of AI relative to other IMs resulted in a higher total standard deviation given an earthquake. Within this context, CAV was deemed the best IM.

A unique opportunity for gaining knowledge and insights is facilitated by the CSMIP Eureka Bridge and Samoa Bridge seismic records, along with those of the nearby Geotechnical ground downhole array. Of special interest is the response of a bridge pier in each bridge with records at the deck level, pile cap and within the underlying pile foundation. This valuable data set is employed to evaluate the ground, pile foundation, and overall bridge seismic response. Spatial variation of the recorded motions is examined. Linear and nonlinear response of the ground and the bridge are assessed using system identification techniques. During the strong shaking phase of the 2010 Ferndale Earthquake, a clear and significant stiffness reduction was observed in the response of the columns and foundations. After the strong shaking phase, flexural rigidity was seen to increase back to its original value (i.e., no perceptible permanent reduction).

Post-earthquake damage assessment can be significantly expedited when machine learning (ML) algorithms are used. Recent earthquakes showed that even when a structure is operational and safe for occupancy, people chose to evacuate and not reoccupy it immediately. Such a behavior can be attributed to lack of knowledge about the structural conditions immediately following the event and the fear of being trapped in the building if aftershocks hit. Currently, there is a lack of rapid quantifiable methods to determine if buildings are safe for reoccupation after an extreme event. However, advances in remote sensing, computing technologies, and data science in the past few years paved the way to develop ML methods that can assess and quantify the conditions of structures in near-real time. This paper introduces a methodology to assess the severity of earthquake-induced damage using low dimensional, cumulative absolute velocity (CAV)-based feature and ML tools. The appropriate features and the ML tool are identified by analyzing a single degree of freedom (SDOF) model. The identified features are then applied to assess the severity and location of damage of two multi-degree of freedom (MDOF) systems and real structures instrumented by the California Strong Motion Instrumentation Program (CSMIP). Results show that the damage detection capability of the features is high.​

The Wilshire Grand Tower in Los Angeles and the Salesforce Tower in San Francisco are two new super tall buildings in California. Both buildings use concrete core shear walls to resist earthquake forces and were designed using a performance-based seismic design approach. During construction, the Wilshire Grand Tower and the Salesforce Tower were extensively instrumented with 36 and 32 sensors, respectively, in a joint effort by the owners and the California Strong Motion Instrumentation Program. This paper describes the sensor locations in the buildings and the instrumentation objectives. Data recorded at the Salesforce Tower during the M4.4 Berkeley earthquake of January 4, 2018, and the ambient vibration data obtained by the instrumentation systems in both buildings are presented. Results from some preliminary analyses of the data are also discussed.​