EXPERIMENTAL SYSTEM The experimental system (Fig. 1, left) was designed to enable a steady state regime of brittle fracture and a simple mechanical analysis of the fracture energy, even for RCP [2, 3]. The geometry is that of a strip band specimen (SBS) of typical dimensions L ≈ 200 mm, 25 mm < H < 45 mm, a0 ≈ 3 H, B ≈ 2 mm. The location of the crack tip during propagation is determined by measuring the resistance of a metallic layer at a sampling rate of 250 kHz [4]. A loading device (Fig. 1, right) ensures uniform and constant displacement of the strip band boundaries and the symmetry of the loading is checked by strain measurements on the sides of the specimen.Crack propagation is started by an impact of low energy on a razor blade placed in contact with an initial blunt notch and the crack propagates symmetrically. Only one of the twin specimens undergoes fracture. Owing to the weight of the grips and the short fracture time, typically 200 µs, we assume that the boundary conditions are fixed during crack propagation. Crack branching can be obtained by increasing the mechanical potential energy of the specimen and the crack branches are generally symmetrical for polymethylmethacrylate (PMMA) and RT-PMMA specimens. RT-PMMA is a blend of PMMA and spherical rubber particles, in these experiments of diameter 200 nm and volume fraction approximately 40 %. + - a H H y x u u a L metallic coating B Figure 1: (left) Schematic representation of the strip band geometry uniformly loaded and the conducting layer used to record the crack tip position during propagation. (right) Experimental device ensuring symmetrical loading. DYNAMIC ENERGY RELEASE RATE COMPUTATION AND VALUES OF THE FRACTURE SURFACE ENERGY During RCP tests, no significant variation of the macroscopic crack speed has been observed for a given specimen at a given temperature, whether branching occurs or not (Fig. 2). Since the crack tip position during propagation and the stress state at initiation are known, the energy release rate GID may be calculated by means of a transient dynamic finite element procedure, using the software Castem2000©. Outside the singularity, thermo elastic effects are expected to be negligible since the mean stress is about 15 MPa [5]. Owing to the high strain rate, the fracture mechanics is assumed to display linear elastic behaviour [6]. The energy release rates were computed by differentiating the elastic energy integrated on the whole structure. As the geometry ensures a quasi-steady state regime of propagation, it is assumed that a specific treatment of the singularity is not necessary since the error done concerning the energy integration at the crack-tip singularity is eliminated by the differentiation. It has been shown experimentally that the impact on the razor blade influences the crack propagation only over the first few millimetres. Nevertheless, this crack initiation is simulated by imposing an initial crack tip opening, corresponding to the action of the razor blade at the crack lips. Figure 3 shows that the dynamic correction factor is generally of the order of 0.7 to 0.8 for macroscopic crack speeds of about 0.6 cr. This specimen geometry in fact induces a low dynamic correction factor [3, 7] and the remote stress field at the crack tip is not strongly influenced by inertial effects in this range of crack speeds. To simplify the results, as GID displays relatively small 2
RkJQdWJsaXNoZXIy MjM0NDE=