( ) ( ) 2 7 7 2 7 2 / / / / / − − − − = in in in c a t a F a t a t t a F c a (5) where 10 8 6 4 2 00013 00251 0 1911 0 7609 1 6105 1 6305 + + + + + = t a t a t a t a t a t a F . . . . . . (6) 1 4 4 4 / ( / ) ( / ) − − − − = in in in c a t a F a t a t t a F c a (7) where 12 10 8 6 4 2 0 0005 00099 0 0908 0 4565 1 3451 1 7716 2 2547 + + + + + + = t a t a t a t a t a t a t a F . . . . . . . (8) The robustness of the Runge-Kutta (R-K) scheme employed here was investigated by assuming different combinations of (a/t)in, (a/c)in taken from the interval (0,1), calculating (a/c)f for (a/t)f = 1 by using Eqn. 3 and comparing with their analytically derived counterparts. In total, 400000 different combinations of these variables were used. The statistical characteristics of the differences between the numerical approximation of Eqn. 3 and the analytical solution are shown in Figure 1 and are presented in the form of mean and mean + standard deviation curves of the percentage difference between the proposed numerical scheme and the exact solution. The mean error lines demonstrate that, larger errors irrespective of the number of R-K steps used are observed for lower m values. However, the scatter is significantly smaller for low m values and decreases as the number of R-K steps increases. Semi-elliptical surface crack in a butt-welded plate under tension The factor YA used for this geometry is given in the Appendix. In this case, a closed form solution cannot be obtained as was done for the previous geometry. Consequently, differences are reported here with respect to the forty-step R-K scheme. Errors in (a/c)f were calculated again for the same values of (a/t)in, (a/t)f (=1) and (a/c)in as in the previous section. The results are shown in Figure 2. The scatter, which is depicted by the solid lines, diminishes with decreasing m. The highest mean errors are here recorded for m=6, which drop below 5% for R-K steps greater than 10. The highest scatter is again
RkJQdWJsaXNoZXIy MjM0NDE=