Stochastic Seismic Response of RC Building and Multi-Span Bridge using Metamodel Approach

Abstract

Variability in geometry, material qualities, and loads all contribute to the stochasticity of a structure's dynamic response. Stochastic analysis gives a decent representation of the random dynamic responses. Stochastic analysis techniques fall into two broad categories: statistical and non-statistical. Although statistical methods such as Monte Carlo simulation are generally accepted as the gold standard for stochastic analysis, simpler, non statistical alternatives that need less processing resources without sacrificing accuracy are required. High Dimensional Model Representation is a relatively new non-statistical metamodel based methodology that is being compared in this work to traditional response surface methods including Central Composite Design, Box Behnken Design, and Full Factorial Design in the context of a dynamic response analysis. To examine the stochastic reactions, the High Dimensional Model Representation technique is used to the shape of a reinforced concrete frame utilising natural frequencies and nonlinear dynamic analysis. Results for the chosen issues were found to be comparable to those obtained using traditional sampling approaches, but with significantly less computational work required thanks to this technique. The recent earthquakes have shown that bridges, which are essential to transportation networks, are also one of the most susceptible parts of these networks. Bridge damage following an earthquake can hinder rescue attempts and lead to significant financial losses for affected towns. An examination of a bridge's seismic resilience both during and after an earthquake is crucial. While advances in design principles have greatly reduced bridges' seismic vulnerability, the earthquake research community remains deeply concerned about the possibility of damage to the culvert stock. The seismic risk evaluation of bridges makes extensive use of fragility curves. In this analysis, the fragility curves for multi-span bridges are presented, taking into consideration the uncertainties in the bridge's structural and material properties. Modern highways frequently have reinforced concrete box-girder bridges, therefore their seismic safety has been the subject of extensive study. However, the nonlinearity and variability in geometry, material parameters, and loading must be taken into account for a thorough seismic analysis of box-girder bridges. This study introduces a novel nonstatistical metamodel-based approach, high-dimensional model representation, to the generation of metamodels for chosen seismic behavior parameters of a box-girder bridge while accounting for uncertain input variables. Comprehensive finite element analysis is used to assess the seismic responses at the sample locations of the high dimensional model representation. In addition, the derived metamodels are used to construct seismic fragility curves, which is proven to be significantly easier than standard fragility analysis. This method yields conclusions that are consistent with those of two well-known response surface methods (Central Composite Design and Box Behnken Design), but requires vastly less simulations to reach a conclusion. High-dimensional model representation not only greatly simplifies the computing burden of fragility assessments by providing a failure metamodel that incorporates all relevant random variables, but it also gives a far more accurate representation of the underlying system. Direct Monte Carlo simulation is the foundation of the most accurate and robust method of seismic reliability analysis. However, it is difficult to compute since many nonlinear time history studies must be performed. In such cases, a metamodeling strategy using the response surface method can be a useful tool. In order to improve seismic reliability analysis of multi-span bridge piers, this work investigates the benefit of using an adaptive response surface method based on the moving least squares method. Three-dimensional finite element models of bridges are constructed on the OpenSees platform, with accurate representations of the bridge's components like the columns, super structure, bearings, and abutments, allowing for nonlinear time history analysis. High Dimensional Model Representation is a relatively new non-statistical metamodel-based approach that is being evaluated in this study in comparison to traditional response surface methods like Central Composite Design, Box Behnken Design, and Full Factorial Design for use in a dynamic response analysis.