Ductile Tearing of Welds in Pipe Submitted to Cyclic Loading

Flowlines and risers can be submitted to plastic deformation, sometimes cyclically, due to the installation technique, or sometimes due to exceptional events. In this case, a specific evaluation of defect acceptance in the girth weld is necessary. The present study investigates the possibility to predict ductile tearing during installation when the performed fracture mechanics tests are only high triaxiality specimens and that the effective application requires cyclic loading. A classical analysis is performed using DNV RP F108 to determine the acceptable defect size of for the case in which a pipe is submitted to cyclic loading. In the present investigation, tearing resistance was characterized with SENB specimens. An engineering critical assessment (ECA) was performed considering the size of the expected defects and the amount of plastic deformation to which the pipeline would be submitted. A validation of the ECA was performed by segment tests. While the application of ECA based on the fracture tests would predict ductile tearing with the considered defect, the results of segment tests only revealed blunting for the considered plastic deformation. It confirms the effect that in lower constraint conditions (like in segment tests), SENB test results are overly conservative. The tearing phenomenon was then simulated by the finite element method using two different damage models (Gurson-Tvergard-Needlemann and the Bai-Wierzbicki model) and compared to the experimental results. As the deformation at the crack tip is typically very large, one needs to have knowledge about the hardening behavior in the post-necking region. As this behavior cannot be directly deduced from standard measurements, an automatic identification procedure was developed to determine the post-necking flow behavior of the weld metal and the base material transverse to the weld. As reported in the literature, simplified models like Rambord-Osgood are then inadequate and model including two hardening zones is necessary: one for small deformation and one for large deformation. The calibration of the damage models was only performed on the tearing curve obtained from the SENB experiments, and the segment tests were then “blindly” simulated.