SMIP 2024 Seminar

A new period-dependent duration based on the cumulative-squared response of a 50%-damped single-degree-of-freedom oscillator is presented. For T>1 sec, the duration is systematically larger than the accelerogram-based duration. There is a large variability of the ratio of the long-period duration to the acceleration-based duration. To develop design time histories for structure sensitive to the long period ground motion, the long-period duration of the time histories should be considered in the selection process. A preliminary conditional ground-motion model for period-dependent duration is presented. A final updated period-dependent ground-motion model will be developed by the end of the project. 

One-dimensional ground response analyses (GRA) can introduce model error to site response estimates when wave propagation is not dominated by vertically propagating shear waves. We identify sites suitable for GRA based on microtremor horizontal-to-vertical spectral ratios (mHVSRs). We analyzed 300 microtremor recordings from 17 vertical array sites in California, comparing mHVSRs at varying spatial separations. We find that low mHVSR spatial correlation, as measured using Longest Common Subsequence, tends to occur at vertical array sites that are poorly modeled by GRA. Conversely, stronger mHVSR correlations tend to occur at sites where GRA is relatively effective. 

The horizontal intensity of earthquake ground motion intensity varies significantly with changes in orientation in what it is referred to as ground motion directionality. Unfortunately, this directionality has historically been ignored or not properly accounted for in ground motion models and in seismic provisions. This study investigates the implications of directionality in the estimation of ground motion intensity. In particular, it investigates the effect of focal mechanism on directionality by making use of a large number of records obtained in well-recorded recent earthquakes in California, Taiwan, and Türkiye. Results show that for strike-slip earthquakes, it is possible to anticipate the orientations in which current ground motion models tend to underestimate spectral ordinates and orientations in which they overestimate them.  

We have used 0-1 Hz 3D numerical wave propagation simulations and strong motion recordings for 7 Mw4.4-5.1 events around the greater Los Angeles basins to estimate the optimal thickness distribution for a generic low-velocity overlay taper (LVT) implemented in the top 1-2 km of the Statewide California Earthquake Center Community Velocity Model (CVM) version S4.26.M01. The optimized LVT model with spatially-varying thickness leads to a reduction of the Fourier Amplitude Spectrum bias by 39% from the CVM without LVT, and by 22% from the CVM including a LVT with a constant thickness of 600 m. 

This paper presents a methodology to determine the time history of the structural response using the Temporal Convolutional Network (TCN), a deep learning method. The methodology, in conjunction with sensor data from instrumented buildings, facilitates the prediction of the response in future earthquakes without a structural analysis model, providing a computationally effective complement, or even alternative, to nonlinear time history analysis. The developed TCN model is applied for predicting the responses of numerical models, a shaking table tested structure, and instrumented buildings for a broad range of response characteristics. Furthermore, ground motions are predicted from the response accelerations using the TCN model. It is observed that the method successful predicted these responses using a practical number of recorded signals for training. The limitations of the adopted methodology are identified, which can be resolved in future studies with physics-informed AI methodologies. 

Viscous damping has long served as a surrogate for the aggregate of non-damage dissipative mechanisms. When the seismic response is strongly nonlinear most of the energy is dissipated by hysteresis but a question that has lingered is whether the viscous model holds unchanged during plasticity. It is shown in this paper that data supported information on this question can be obtained by inspecting the rate of change of the base shear over time intervals of strong plasticity. The results suggest that the constants that specify the pseudo viscous model decrease when hysteretic dissipation is active. 

Pier walls used in highway bridge overpasses typically are squat shear walls with a height to width ratio lower than two. Squat walls are characterized by brittle shear failure modes. Unlike squat walls used in buildings, bridge pier walls have different reinforcement amounts, detailing, and loading demands, which affect the wall’s strength and ductility. Three large scale squat shear wall experiments have been tested to better understand the failure mechanism and behavior of bridge pier walls. All specimens had an ultimate drift exceeding 1.5% and saw sliding shear failure. CSMIP sensor data of a squat pier wall bridge is also analyzed to better understand the relationship between a component pier wall model and a global bridge system model. 

The ShakeAlert Earthquake Early Warning (EEW) system aims to provide advance notice to residents on the US West Coast seconds before ground shaking begins, if the expected intensity exceeds a set threshold. However, occupants of tall buildings may feel amplified motion due to the buildings' unique dynamic response. Therefore, efforts are underway to improve ShakeAlert by accounting for building response, to better estimate shaking intensity in high-rises. Currently, using detailed Finite Element models of buildings for this purpose is not feasible. The author recently reviewed simple methods for estimating Peak Floor Acceleration (PFA) and found them impractical for real-world application (Ghahari et al., 2022a). This paper explores an alternative by extending the Performance-Based Earthquake Engineering (PBEE) framework from the Pacific Earthquake Engineering Research Center (PEER) to EEW, acknowledging that each component of building response prediction is uncertain in the EEW context. Although others have proposed this idea (Iervolino, 2011; Cheng et al., 2014), it has two main issues: 1) the simple beam model introduces unquantified modeling uncertainty, and 2) ground motions in probabilistic demand models are unsuitable for EEW. This work addresses these issues by incorporating modeling errors into the beam model (Ghahari et al., 2022b) and using new ground motions. The approach was demonstrated using data from a 52-story building in Los Angeles (Ghahari et al., 2024), showing that accurate Peak Ground Acceleration (PGA) estimation can predict expected human comfort levels in tall buildings.