Experimental investigations on the ‘shallow crack effect’, on the 10 MnMoNi 5 5 steel, and computational analysis in the upper shelf by means of the global and local approaches Original Research Article

Shallow cracks have been of great interest for some years now, since it has been observed that standard fracture toughness parameters, determined on deep crack specimens, exhibit a geometrical dependence and that most of the structures containing cracks (in particular pressure vessels) exhibit crack depth to width ratios a/W much lower than fracture mechanics specimens tested in the laboratory. This study performed at MPA Stuttgart in collaboration with EDF, aimed first of all at investigating the ‘shallow crack effect’ on the high ductile, high strength, low strain-hardening shape welded 10 MnMoNi 5 5 steel over the whole temperature range of the transition region of toughness. Deep crack (CT25 and TPB25 with a/W=0.55) and shallow crack (TPB25 with a/W=0.066) specimens were tested, providing KIC, Ji and JIC results. At the same time, computations were performed, in the upper shelf with, on the one hand, the usual elastic–plastic model (Von Mises model) and on the other hand, the local approach model of Rousselier. Experiments did not exhibit any shallow crack effect in the lower shelf (no increase in KIC) and show a slight effect in the transition region and in the upper shelf (slight increase in Ji, larger increase in JIC). Computations confirmed the ability of the ‘modified’ Rousselier model to describe the macroscopical behaviour (F-CMOD curves) of the three types of specimens as well as to simulate their crack propagation behaviour (CMOD-Δa curves). The combination of the Rousselier model with the line contour integral method enables us to develop the draft of an experimental method to determine the J-integral for any material and any initial crack length (a/W remaining within 0.05 and 0.7), taking crack growth into account. This method, based on the measurement of the energy under the F-CMOD curve and an ηc factor indexed on the current a/W ratio, provided an agreement within 10% and most of the time 5% between the experimental and computational J-integral results. Consequently, computations with the Rousselier model also correctly predicted the experimentally observed crack resistance curves, especially their geometrical dependence. The observation during crack growth of the triaxiality of the stress state in the first millimeters of the ligament enabled one to understand this geometrical dependence, high triaxiality accelerating crack growth leading to lower J–R curves. For shallow crack specimens, the crack tip being located close to the top surface, a stress relaxation occurs at this top surface leading to a loss of triaxiality at the crack tip and in the ligament, delaying crack growth and leading to higher crack resistance curves. Finally, the Rousselier model was found to confirm the experimentally observed Ji values on the various tested specimens. Therefore, assuming that the Rousselier model can correctly describe initiation, computations with different stress–strain curves (different yield stress values and strain-hardening behaviours) were performed: it seems that the ‘shallow crack effect’ (Ji(shallow crack)/Ji(deep crack)) is controlled by strain-hardening independently of the yield stress, the absence of strain-hardening leading to the most important effect.