Methodologies for Crack Initiation in Welded Joints Applied to Inspection Planning

Over the past decades, crack propagation has been extensively studied by researchers around the word. The approach based on crack propagation models have been widely used in inspection planning. This approach has the advantage that it gives measurable fatigue damage accumulation in terms of crack propagation with time and thus crack propagation models can be updated with inspection results. However, a prerequisite for using crack propagation models in inspection planning is that parameters such as initial crack size, crack growth rate, geometry function, etc. are known.  Among those parameters, initial crack size, depending on welding quality, material and the environment, is associated with the most uncertainties because of sampling and measuring problems. Another prerequisite for using crack propagation models in inspection planning is that crack initiation period can be assumed to be negligible. Both prerequisites are challenged nowadays as manufacturing and welding techniques have been improved. Some high-quality welded joins have been proven free from detectable size of flaws and the crack initiation period can account for a large part of the whole fatigue life. This gives rise to big difficulty for inspection planning of high-quality welded joins, as there is no generally acceptable approach for modelling the whole fatigue process that includes crack initiation period. Compared to as-welded joints, reliable inspection planning is more crucial for high-quality welded joins, as they are generally designed to withstand a larger stress range. In addition, they may have shorter time for inspection as crack initiation time account for a large part of fatigue life, with a shorter crack propagation period to failure due to higher stress range. To address this problem for high-quality welded joints, a robust model accounting for the whole fatigue process needs to be developed. The core issue is how the crack initiation period can be modelled and added to the crack propagation time. To help identify this issue, this paper reviews treatment methods for crack initiation period and initial crack size in crack propagation models applied to inspection planning. Generally there are four approaches, by: 1) Neglecting the crack initiation period and fitting a probabilistic distribution for initial crack size based on statistical data, e.g. Weibull distribution or lognormal distribution; 2) Extrapolating the crack propagation stage to a very small fictitious initial crack size, so that the whole fatigue process can be modelled by crack propagation models; 3) Assuming a fixed detectable initial crack size and fitting a probabilistic distribution for crack initiation time based on specimen tests; 4) Modelling the crack initiation and propagation stage separately using small crack growth theories and Paris law or similar models. Conclusion is that in view of trade-off between accuracy and computation efforts, calibration of a small fictitious initial crack size to S-N curves is the most efficient approach.