Issue 31

H.F.S.G. Pereira et alii, Frattura ed Integrità Strutturale, 31 (2015) 54-66; DOI: 10.3221/IGF-ESIS.31.05 55 response. These complex phenomena involved in the bond behaviour have led engineers in the past to rely heavily on empirical formulas for the design of concrete structures, which were derived from numerous experiments, e.g. [7 - 15]. The properties of the adherence between rebar / matrix depends on several factors, such as friction, mechanical interaction and chemical adhesion [24]. In the past, several experimental investigations have been carried out in order to clarify and understand the behaviour of deformed bars pulled out from a concrete bulk under monotonic as well as cyclic loading conditions. Some of these experimental results are well documented in literature, e.g. [7-19, 25]. Based exclusively on the experimental results it is difficult to filter out the influences of material and geometrical parameters on the bond behaviour. In order to understand thoroughly the bond behaviour, a reliable numerical model (simulation of the transmission of forces at the interface zone, see Fig. 1a) should be employed, thus a three-dimensional finite element, analysis is needed. The numerical modelling of the bond behaviour is principally possible at two different levels: (1) detailed modelling (see Fig. 1b) in which the geometry of the bar and the concrete are modelled with three-dimensional elements and (2) phenomenological modelling (see Fig. 1c) based on a smeared or discrete formulation of the bar-concrete interface [21]. (a) Idealized bond zone. (b) Detailed modelling. (c) Phenomenological modelling. Figure 1 : Schematic simulation of the idealized bond zone [21]. The phenomenological modelling of bond between rebar / concrete can be discretised by three-dimensional finite elements. The link between the rebar and the concrete can be achieved by a discontinuous approach, where bond is defined by discrete or reduced thickness cohesive elements. Within these elements the behaviour is controlled by the local bond stress-slip relationship. This approach is able to realistically predict the pullout behaviour for different geometries and for different boundary conditions only if a realistic constitutive bond relationship is used. However, the model is not able to straightforwardly predict the bond behaviour of a given bar geometry. Consequently, the influence of these parameters must be stored in advance in the basic parameters matrix of the bond model. Thus, one has the possibility to realistically simulate the behaviour of reinforced concrete structures with relatively low effort in modelling and computing time. By the use of detailed modelling, such as both modelling of the ribs of the reinforcement and the concrete lugs (see Fig. 1b) between the ribs of the reinforcement a quite refined finite element mesh has to be generated. This leads again to a high effort in modelling, and also to really long computational time, in particular while carrying out a finite element analysis of complex reinforced concrete structures [21 - 23]. In the present work a parametric study of the numerical simulations of galvanized rebar pullout tests under the finite element framework is presented and discussed. Afterwards, the numerical simulations of galvanized rebar pullout tests are compared with the results obtained by using an analytical model [31], namely, a shear-lag model. The adopted local bond- slip laws were similar to the one proposed by the CEB-FIP Model Code 2010 [27]. Finally, an attempt is made to model the pullout tests by isolating the distinct bond mechanisms, in particular, the mechanical component of bond due to the rebar’s ribs. Therefore, to fulfil this purpose a three-dimensional finite element model considering the geometrical modelling of the rebar ribs was implemented. D ESCRIPTION OF THE NUMERICAL MODEL Geometry he experimental pullout tests were carried out on concrete cubic specimens with a rebar positioned in the middle of the specimen. The rebar’s embedded length was equal to half of the cube edge length, i.e. 100 mm, Fig. 2. Only one quarter of the experimental specimen was geometrically modelled, because the specimen has double symmetry T