A comprehensive fatigue test database was developed in PRCI SIA-1-1 and IM-3-2 projects, where there are 185 sets of fatigue crack growth rate parameters (i.e., C and m fitting coefficients in the Paris law) calculated from compact tension (CT) fatigue test data. Four sets of recommended values of C and m were presented for two flaw locations (base metal (BM) and Electric Resistance Weld seam (ERW)) and two R ratios (R < 0.5 and R ≥ 0.5). The past works did not further assess the fatigue crack growth rate parameter relation (i.e., C and m), although intuitively, two parameters should be correlated for a material type (i.e., line pipe). To this end, a systematic study was carried out to reassess the fatigue crack growth rate database. It is found that parameters C and m are strongly correlated in the form of m = A * log(C) + B, where A and B are fitting coefficients. In addition, C and m (mresidual) values follow lognormal and normal distributions, respectively, where mresidual refers to the residuals associated with fitting the predicted values. The values of A and B for BM or ERW at R = 0.1 or 0.6 (four scenarios) are reported. Note that the distributions of C and m for ERW cases have relatively heavier tails and shorter peaks.
Based upon the four newly developed sets of A and B values, a series of probability studies were carried to calculate the most-likely fatigue crack growth life in the integrity analysis. Note that two ERW pipe segments reported in PRCI IM-3-2 projects were considered in the full-scale test, where fifty-one (51) initial notches were prepared to evaluate the fatigue crack growth rate. The initial notches were cycled to develop different crack lengths, depths, materials, and flaw locations (e.g., BM and ERW seam). The fatigue crack growth lives calculated from the PRCI approach were used as the baseline for comparisons. In total, 30,000 pairs of C* and m* data points were generated from random sampling for each scenario, and there are over 1.5 million random simulations carried out. The probability densities for each full-scale test case were simulated using a Monte-Carlo approach. The fatigue crack growth life distributions for BM are relatively narrow because of the fatigue crack growth rate parameters’ lower standard deviation. The most-likely (median) fatigue life from the probabilistic study is almost identical to the fatigue life from deterministic calculations for the BM cracks. The probabilistic simulation results of ERW seam crackfatigue lives are significantly different from those for BM cracks.