Pull-Out Behaviour of Steel Reinforced Cement Composites

This thesis is concerned with developing an effective and comprehensive method to simulate pull-out response of fibres in fibre reinforced cement composites. In particular, it addresses cases involving randomly oriented fibre, large relative sliding and interfacial separation between fibre and matrix, and cyclic loading. This work is also extended to study the pull-out behaviour of curved bars in reinforced concrete, because this has many similar features and mechanisms to fibre reinforced concrete. The main strategy and contributions are outlined in the following. A novel pull-out modelling method is proposed. This consists of a contact algorithm with friction for the interface and damage models for the matrix. It is the first known attempt to merge various pull-out mechanisms of a fibre reinforced cement composite into one two dimensional finite element model. These mechanisms include bonding, debonding, fibre sliding, friction, fibre bending, snubbing, matrix spalling and substantial separation of the fibre at its exit from the matrix. The use of a contact algorithm with friction simplifies the simulation of the interface and only requires an experimental pull-out load-slip relationship from a single perpendicular fibre, having no requirement for additional strength or fracture criteria. Interfacial separations are dealt with by means of different normal constraints on the interface. Matrix spalling is automatically simulated using a concrete damage model. Accompanying the development of the pull-out model, four increasingly complex concrete damage models and corresponding computational algorithms are developed. These models are a pure damage model (model I), an inelastic- damage model (model II), and reverse cyclic and biaxial loading damage models (model III and model IV). Although the first two models are only suitable for monodirectional cyclic loading, they can describe concrete responses under uniaxial monotonic loading (for both models) and monodirectional cyclic loading (for model II) very well. Models III and IV are based on different damage mechanisms. In the former, the positive and negative parts of the principal strains control the tensile and compressive damage respectively. Since the model is described in strain space, the complexity due to a stress space description in existing models is avoided. In model IV, the introduction of a weighted average damage parameter overcomes the shortcoming of separating stress/strain into positive and negative parts, and greatly simplifies implementation in the finite element method. Additionally, the design of a damage multiplier distinguishes the different contributions of hydrostatic and deviatoric components of the stress/strain tensor to damage and produces the modelling under biaxial loading. The implementation of model IV under biaxial tension and biaxial compression reproduces completely the biaxial experimental results of Kupfer et al. The validation of the developed models are proved by comparing against experiments. Application are made to fibre reinforced cement composite with single and multiple inclined fibres, and to curved bar reinforced concrete. These studies provide some useful conclusions and point to several recommendations for further researches