The ability to model crack-closure behaviour and aggregate interlock in finite element concrete models is extremely important. Both of these phenomena arise from the same contact mechanisms, and the advantages of modelling them in a unified manner are highlighted. An example illustrating the numerical difficulties that arise when abrupt crack closure is modelled is presented, and the benefits of smoothing this behaviour are discussed. We present a new crack-plane model that uses an effective contact surface derived directly from experimental data and which is described by a signed-distance function in relative-displacement space. The introduction of a crack-closure transition function into the formulation improves its accuracy and enhances its robustness. The characteristic behaviour of the new smoothed crack-plane model is illustrated for a series of relative-displacement paths. We describe a method for incorporating the model into continuum elements using a crack-band approach and address a previously overlooked issue associated with scaling the inelastic shear response of a crack band. A consistent algorithmic tangent and associated stress recovery procedure are derived. Finally, a series of examples are presented, demonstrating that the new model is able to represent a range of cracked concrete behaviour with good accuracy and robustness. Figure 10. Uniaxial opening and closing response.Figure 11. Cracks opened in uniaxial tension and subsequently loaded in shear.
EMBEDDED CRACK PLANES IN A THREE-DIMENSIONAL CONSTITUTIVE MODEL
Embedded smeared cracksWhen the crack-plane model is applied to continuum elements, cracks are effectively smeared over elements and are represented by directional damage. The term h is no longer employed directly to define the local strains; instead, the crack-band approach of Bazant and Oh [3] is used. This involves scaling the governing fracture softening stress-strain curve, using the fracture energy parameter .G f /, such that the energy consumed in fully opening a mode 1 crack is the same irrespective of the element size. A key assumption in this method is that (numerically) strains localise into a band of elements, one element wide, early enough in the fracture process for the fracture energy consumed outside this band to be negligible. If this assumption is valid, the inelastic fracture strain vector (i.e. the reduced form of the strain tensor) can be related to the inelastic relative-displacement vector . M u/ as follows:The characteristic length is equal to the length of the longest straight line that could be drawn within a finite element normal to the crack plane. This gives a length in between that is computed by 'Oliver's first and second methods', which are described in References [52] and [64].The element local strain vector .Q ©/ is then defined as the sum of the inelastic and elastic strain components:in which ı denotes a third-order-matrix vector contraction. The incremental fracture strains are then given by Equation (35):