13th International Conference on Fracture June 16–21, 2013, Beijing, China -2- “steady-state” shear lips is a linearly increasing function of the effective ΔK, which usually increases with R. The analyses of experimental data in the literature are essentially two-dimensional. Slanted crack growth kinetic data is analysed as if it was mode I and as if the growth rate and driving force were uniform along the front. Based on “constant ΔKI tests”, empirical relations between the steady-state shear lips width, the loading frequency and the “effective ΔKI ” were derived for aluminium alloys [3-4]. This parameter will be denoted below by “apparent ΔKI”, since it is computed in 2D for a normal crack of same length as that observed on the free surface, with an empirical correction for closure effects deduced from the R ratio. The meaning of “constant ΔKI tests” is also questionable in view of the large gradient in KI, KII and KIII along the front of a partially or completely slanted crack and of the reduction in ΔKI associated with crack twisting. In an effort toward 3D analysis, Pook [8] performed finite element computations of the stress intensity factors along the front of a fully slanted crack in specimens of different thickness. He found that KI was 0.51 to 0.71 smaller than its apparent value (computed in 2D for a normal crack of same length). He also found that KIII had the same order of magnitude as KI, while KII raised near the free surfaces with a skew-symmetric profile. Bakker [9] performed 3D computations of stress intensity factors for fully or partially slanted crack. However, as in the case studied by Pook, a straight crack front was considered, while tunnelling probably plays a role in crack twisting in fatigue. The present work re-examines the problem from a 3D perspective, in order to determine what mechanical parameters control the onset of crack deviation and the kinetics of slanted crack growth. Since environment seemed to influence shear lips development, a 3D experimental characterization of the crack paths and kinetics was performed, both in air and in salt water and the crack paths were compared. A 3D numerical analysis of the crack growth rates was done, based on linear elastic fracture mechanics, taking mode-mixity into account. Elastic-plastic computations of stress and strain fields ahead of the crack front were used to rationalize the observed crack paths. A method to predict the onset of crack twisting and the twist angles was proposed. 2. Experimental procedures Fatigue crack growth tests were performed with a frequency of 5Hz and R=0.1 on 6mm-thick, 100mm-wide, 300mm-high Center-Cracked Panels (CCP) specimens. Two material were investigated: 7075-T651 aluminium alloy (σ0,2= 376MPa, σu= 537MPa, E= 75 GPa) and S355 low-alloy steel (σ0,2= 349MPa, σu= 510MPa, E= 205 GPa). Both sides of the specimens were polished to allow crack propagation monitoring with an optical microscope, at a magnification of one hundred. The tests were performed in air or in a transparent reservoir filled with 3,5g/l NaCl solution, under different loading amplitudes indicated in Table 1. Marker block loading sequences with and increased Kmin but the same Kmax were periodically applied, so that the R ratio became temporary 0.7, until approximately 100 µm propagation was achieved, in order to mark the position of the crack front and be able to derive the mean crack growth rate between consecutive markings, for any point along the front. Ten to twelve marker blocks were applied at 20Hz. To characterize crack front tunnelling, the difference in crack length between the mid-thickness and the average length on free surfaces was measured and denoted by Δa.
RkJQdWJsaXNoZXIy MjM0NDE=