"Fragility curves are useful tools for reliability evaluation of structures as well as for identifying the most vulnerable components. This study focuses on the seismic fragility analysis of highway bridges. Two main approaches are used for this purpose: component-based and system-based fragility analyses. The seismic vulnerability of two existing bridges located in Montreal are assessed as case studies.The main goal of this study is to develop reliable seismic fragility curves for highway bridge structures considering all significant uncertainties involved. Uncertainties include those associated with modelling structural behavior, seismic inputs and definition of component capacities. The procedures are implemented for the fragility assessment of two existing bridges as case studies. For this purpose, deterioration due to corrosion of reinforcing steel and its effects on structural behavior are included, as well as validation of the Finite Element Model using dynamic properties obtained from ambient noise measurements. Proposed methods for the selection of appropriate set of ground motion records, the type of model analysis and probabilistic modeling of component capacities are presented and illustrated for the two case studies.Two stochastic methods are proposed for validating the Finite Element Model of a bridge. The first method is based on classical hypothesis testing procedures while the second uses a Bayesian updating approach. The stochastic methods are also used to update the input parameters, detect probable major damage in the bridges and determine the confidence interval on model responses as a function of laboratory test data and field observations.In order to limit the uncertainties involved in seismic inputs, a state-of-the-art ground motion record selection procedure based on Conditional Mean Spectrum (CMS) is used. Incremental Dynamic Analysis (IDA) is performed to evaluate the record to record variability in seismic responses and to capture the nonlinearity in structural component behaviors.The first part of the thesis describes the application of component-based fragility analysis for the seismic vulnerability assessment of highway bridge structures. IDA is performed on the validated Finite Element model of the structure using an appropriate set of ground motion records. The results are used for estimating the relationships between ground motion intensity measures and component demands. A Joint Probabilistic Seismic Demand Model (JPSDM) is fitted to the results in order to develop component and system fragility curves of the structure.Since the component based fragility analysis of complex structures comprising a large number of components requires enormous computational efforts, in the second part of this study, a system-based approach for developing seismic system fragility curves is proposed which uses Support Vector Machines (SVM). SVM is a state-of-the-art machine learning technique which is used to discover patterns in highly dimensional and complex data sets. In this application, SVM is used to determine the relationship between ground motion intensity measures and peak structural responses. Seismic fragility curves are developed using Probabilistic SVM (PSVM). Finally, the efficiency of the proposed PSVM method for its application to vector-valued ground motion Intensity Measures (IM) as well as traditional single-valued IM are investigated." --