Sousa, Romain2020-02-192020-02-192015-11http://hdl.handle.net/10400.8/4688The last decades have witnessed significant progresses in the development of improved numerical models for structural analysis, as proven by hundreds of dissertations, theses, reports and journal papers dedicated to advanced constitutive models of materials and algorithms to model sectional, element and structural response. Nevertheless, the seismic behaviour of structures involves a number of nonlinear material and geometrical phenomena that, ultimately, are impossible to capture exhaustively in a single model. Furthermore, past studies showed that the most correct modelling options from the scientific viewpoint are sometimes challenged by experimental evidence. This thesis intends to contribute to the ongoing effort of progressively bridging the existing gap between solid theoretical principles adopted in nonlinear modelling and experimental results from shake table or other experimental techniques. Such goal is firstly pursued through the application of a sensitivity analysis to the simulation of the dynamic behaviour of three distinct structures with distributed plasticity beam-column fibre-based elements based on Euler-Bernoulli beam theory. The latter were tested in international blind prediction challenges wherein the author and/or supervisors participated with encouraging results. The goodness-of-fit for each approach is assessed through comparisons between numerical and experimental results in terms of lateral displacements as well as accelerations (when available), following two post-processing strategies: a more conventional one based on the error associated to the peak values measured during each record, and another using the frequency content characteristics of the entire response history. Sensitivity parameters included equivalent viscous damping, element discretization scheme, strain penetration effects, material constitutive models, numerical integration algorithms and analysis time-step size. The conclusions, which are interpreted in the light of state-of-the-practice recommendations and established theoretical frameworks, address fundamental modelling decisions for engineers and researchers. The referred sensitivity analysis identified the simulation of strain penetration effects as particularly relevant. They can significantly impact the seismic response of structures, contributing up to 40% of the overall lateral deformation of RC framed structures. Within this context, the last chapters of the thesis present a novel bond-slip model for RC structures that simulates the member-end deformations associated with strain penetration effects. The model, which in its final form is implemented as a zero-length element, was developed so that it is compatible with any general fibre-based frame analysis software. In a nutshell, the element response is determined from cross-sectional fibre integration, where at each rebar the anchorage mechanism is explicitly modelled through a series of virtual integration points distributed along the anchorage length. The analysis is carried out by an algorithm that enforces both equilibrium and compatibility at every integration point, making use of a state-of-the-art bond stress-slip cyclic constitutive relation applicable to a wide variety of anchorage conditions. Therefore, features such as the expected failure mode (pullout or splitting), or parameters such as the concrete strength, embedment length, cyclic degradation, amplitude of steel strains, rebar type (plain or ribbed), transverse pressure, level of confinement and bond conditions can be explicitly modelled. The element was implemented in a structural analysis software and its performance was assessed against several experimental tests, showing an encouraging accuracy while retaining appreciable computational efficiency.engmaster thesis2020-02-19cv-prod-1088372