13th International Conference on Fracture June 16–21, 2013, Beijing, China -5- are presented in Fig. 3a. As the k increases, KI will decrease. Since crack growth rate is an exponential function of K and the Paris law exponent is usually 2-4 for aluminium alloys, crack growth rate will be significantly reduced as the biaxial load ratio increases. This is reported in test measured crack growth rates [11-12]. Fig. 3b shows the non-dimensional K, i.e. the factor by eq. (1), which is independent of the y-axis applied stress y. Fig. 3 FE calculated stress intensity factors under various biaxial load ratios: a) KI; b) factor. 4.3 Prediction of crack growth life and trajectory Fig 4 shows comparison of predicted and test measured crack growth lives. There are three possible reasons for underestimating the crack growth life. 1) Residual stress effect was accounted for by using the measured da/dN data of an M(T) specimen fabricated by the same welding process with crack growing parallel in weld (perpendicular to the load direction) [20]. The magnitude of residual stresses are not exactly the same due to different geometry and size between the M(T) and cruciform specimens. However, since the welds were parallel to the crack growth path in both specimens, the transverse residual stress, which is perpendicular to the crack plane, is low in magnitude compared to the longitudinal residual stress parallel to the crack plane. Therefore the effect of transverse residual stress on crack growth is considered to be small. 2) Influence of weld metal microstructure change on crack growth rates is also contained in the measured da/dN data from the M(T). However, crack in the M(T) had was in the weld nugget centre, but the cruciform specimens had weld in the weld TMAZ zone about 5 mm from the weld nugget centre. Microstructures in these two zones are different, and this difference in microstructure is not modelled in this work. 3) FE calculated y-axis strains are found to be higher than the test measurement, leading to higher SIF values and, consequently, shorter predicted crack growth life. Fig. 4b shows the measured and predicted crack growth trajectories. Calculated crack trajectory is modelled by crack extension under static loading rather than cyclic loading. Crack turning is predicted by the maximum tangential stress criterion in the Abaqus code. Calculated maximum crack deviation for “Configure 1” is about 3 mm, which is larger than that of “Configuration 2 (test specimen geometry)”. The difference is caused by the width of the pad-up, which is 30 mm and 56 mm, respectively. Narrower pad-up (30 mm) encourages crack deviation into the thinner skin. Although crack turning is predicted by the FE modelling, calculated crack deviation magnitude is smaller than the test measured Fig. 4b. Predicted deviation also started at longer crack length (120
RkJQdWJsaXNoZXIy MjM0NDE=