recorded for m=6 and this drops below 5% for R-K steps greater than 18. These results indicate that use of lower order schemes at the same small number of steps could result in unacceptably high errors. The convergence characteristics may also be investigated in a more comprehensive manner by looking at the fatigue lives calculated via Eqn. 4 in a probabilistic way. Here, the bi-linear crack growth law proposed in BS 7910 [1] is used. 0 2 4 6 8 10 12 0 5 10 15 20 25 30 35 40 Number of R-K steps % Error m = 6 m = 5 m = 3 m = 2 Mean Mean + s.d. Figure 1: Mean and mean + s.d. curves of the errors in (a/c)f as functions of the R-K steps for a cracked plate The crack growth model comprises two lines, each being described by its own Paris parameters. The crack growth law is described by the following equation ( ) ( ) ∆ ≤∆ ∆ ∆ <∆ ≤∆ ∆ ∆ ≤∆ = K K C K C K K K K K K dN da tr m tr thr m thr 2 1 2 1 0 (9) where ∆Kthr is the threshold stress intensity range and ∆Ktr is the stress intensity range corresponding to the intersection of the two crack growth lines. The deterministic and probabilistic variables and their characteristics used here are the same as in Reference [4] with the added random variable of (a/c)in, which was assumed to be lognormally distributed with mean 0.01 and coefficient of variation 0.5 . The Paris parameters m1 and m2 were taken to be deterministic and equal to 5.10 and 2.88, respectively, in accordance with BS 7910 [1]. Errors here were calculated with respect to the estimate arising from the application of the highest number of R-K steps (nI, nII = 10) in both crack growth regions. Results are shown in Figure 3. Additional information is given by the inset of this figure. The figure shows that in this case, the influence of the number of R-K steps in the near threshold crack growth region (I) is far more important than in the Paris region.
RkJQdWJsaXNoZXIy MjM0NDE=