Applications of non-Gaussian models to the solution of structural engineering problems

The major sources of uncertainties in engineering problems are the input actions and the system characteristics. Environmental loads such as earthquakes, ocean waves and wind are examples of input actions. The distribution of these actions are non-Gaussian, as various statistical analysis of observed phenomena shows.; This work (i) represents these non-Gaussian actions by memoryless transformations of filtered Gaussian and Poisson processes, and (ii) finds solutions for the response of linear and nonlinear systems to the proposed input model. Response analysis is performed using the Ito differentiation formula and the Monte Carlo simulation method.; Applications to the fields of offshore structures and earthquake engineering are presented. The ocean wave forces are represented by two models: (i) a superposition of a translation process and a Poisson white noise and (ii) the sum of a filtered Poisson process and two Poisson white noise processes. These input models are calibrated to wave force records. Response analysis of an offshore platform modeled as a linear system to these input models is performed. Comparison with ocean wave records and the response of the offshore platform to these wave force records show very good agreement. For the earthquake engineering application, the seismic ground acceleration is represented by filtered Gaussian and Poisson processes. The processes are equal in second moment sense and calibrated to a recorded ground acceleration. The filtered Poisson process also matches the marginal fourth order moment of the record. Response analysis of linear and nonlinear structures to these models show that the filtered Poisson model yields higher response measures than the Gaussian one. Thus, the Gaussian model is an unconservative model because it does not represent well the actual ground acceleration record.