UC Irvine

Transportation networks of highway bridges are considered the lifelines of a community’s infrastructure as they play a significant role in joining communities and serving as the first outlet during a calamity such as earthquakes. They are expected to sustain minor damage and maintain their functionality after major natural and human-made disasters. Observations from seismic events of the last three decades reveal that bridges designed according to seismic design codes demonstrate poor performance and occasionally undergo significant damages leading to major consequence on the affected societies. In the light of these effects, numerous studies have showcased that one of the primary reasons for the unexpected performance of the bridge structures is the improper estimation of the expected seismic demands during the design and analysis phase.

With the evolution of risk, reliability, and hazard analysis in quantifying the seismic vulnerability of structures, the seismic structural design procedures are continuously updating to develop methodologies that achieve more accurate estimations of the structural demands corresponding to the target hazard levels. The most widely used conventional procedure is to conduct Incremental Dynamic Analysis (IDA) by selecting and scaling seismic ground motion records to attain a scalar Intensity Measure (IM), such as spectral acceleration (Sa), associated with a target hazard level of the IM hazard curve. The scaled ground motions are then used to conduct the Non-Linear Time-History Analysis (NLTHA) of finite-element models of the structures, and the obtained response value, i.e., Engineering Demand Parameters (EDP), are then utilized for developing EDP hazard curves by integrating the EDP-IM data over the IM hazard curve. Numerous studies are conducted worldwide describing the limitations of this type of analysis, such as sufficiency and efficiency of single scalar IM as the target, unrealistic scaling of recorded ground motions, type of IDA, ground motions not being site-specific, etc.

The research effort presented herein firstly proposes a supplementary Generalized Ground Motion Prediction Model (GGMPM) that can be used to construct a vector-based (29 x 1, representing intensity-, duration- and frequency- the content of ground motions) IM hazard curves which can be then used to select and scale ground motions targeted to a vector of correlated IMs. The proposed GGMPM consists of Recurrent Neural Network (RNN) and inter-event and intra-event covariance functional forms developed using the Covariance Matrix Adaptation Evolution Strategy (CMA-ES). Though the proposed GGMPM framework provides a swift tool to the engineering community to tackle some of the deficiencies of current IDA methods, this research work further develops EDP hazard curves by conducting large-scale NLTHA of the four bridge structures using synthetic ground motions of Site-Based (i.e., DRD model) and Physics-Based (i.e., CyberShake 15.12) simulation models. For the five sites in the southern California region, rupture variations and their respective probabilities are obtained from Uniform California Earthquake Rupture Forecast, Version 2 (UCERF2) database, which are used to simulate Site-Based synthetic ground motions using the DRD simulation model for a time-span of 100,000 years. Similarly, Physics-Based synthetic ground motions simulated for the CyberShake 15.12 study representing a time-span of 200,000 years are selected. This leads to around ~20,000 Physics-Based ground motions and ~10,000 Site-Based ground motions for each site. These simulated ground motions are then used to conduct NLTHA of the four OSBs to obtain simulation-based- Site-Based and Physics-Based EDP hazard curves. The two types of simulation-based EDP hazard curves are compared against the conventional IDA-Based EDP hazard curves and various regression models are proposed to transform IDA-Based EDPs to simulation-based EDPs. Finally, the three types of EDP hazard curves (i.e., Site-Based, Physics-Based, and IDA-Based) for the four bridges and five sites are provided for comparison.