Offshore steel risers are often subjected to harsh environment and variable repetitive loading during service life. The associated variations in stresses and strains in the pipe wall may give fatigue cracks, particularly close to the girth welds that are joining the pipe segments. The potential cracks are sketched up in Figure 6.4 and they may occur both from the external toe of the weld cap or from the weld root as shown. It must be ensured at the design stage that these potential cracks do not initiate and grow to sizes that represent a threat to the integrity of the steel riser during the target service life. Such fatigue damage may give production loss, unacceptable repair costs and even endanger human safety. On this background a Joint Industry Project (JIP) led by The Welding Institute (TWI) and Det Norske Veritas (DNV) has been initiated. The goal is to provide more scientific insight into the actual fatigue damage mechanism such that more accurate and reliable life prediction models can be established. Furthermore, the guidelines and the recommendations that are given in various building codes should be harmonized. Marinres has been asked to join the project to contribute to this important work. The JIP is funded by oil & gas companies.

Marinres shall use a Random Fatigue Limit Model (RFLM) to solve the problems discussed above. The RFLM represents an alternative to the common statistical analysis used as basis for the conventional S-N cures. An important characteristic of the RFLM is that both the finite life N and the fatigue limit S0 are treated as random variables simultaneously. This will result in a non-linear S-N curve for a log-log scale. If the scatter at long live increases, this will be revealed by the model. Furthermore, the existence of a fatigue limit may be accepted or rejected based on the obtained model. As the RFLM is using a maximum likelihood approach, the run-out data can easily be included. The RFLM will give more reliable selection of the transition point between upper and lower S-N curve. The slope and the scatter pertaining to the lower curve can also be more accurately determined. Hence, the RFLM can be used to assist the construction of the conventional S-N curves. Furthermore, based on the RFLM, a Probabilistic Two-Phase Model (PTPM) can be established. This model that may account for the existence of a crack initiation phase, is a stand-alone model for life predictions and may replace the existing conventional S-N curves. Preliminary PTPM results compared with data from VAL testing of girth welded pipes are shown in Figure 6.5. As can be seen, the predictions made by the PTPM are closer to the test data and they are substantially longer than the predictions made by the conventional S-N curves. These results are promising but need to be corroborated during the present JIP. More VAL test data for welded pipes should be provided.