Experimental Relative Relative values Kriging error BPNN error 1 235 238 1.18% 235 0.16% 2 237 237 -0.09% 235 -0.62% 3 252 254 0.78% 252 0.01% 4 228 232 1.77% 230 1.13% 5 258 254 -1.55% 257 -0.48% 6 246 245 -0.67% 246 -0.13% 7 242 242 -0.07% 240 -0.65% 8 254 250 -1.56% 256 0.82% 9 223 232 4.27% 222 -0.54% 10 241 241 0.08% 240 -0.24%

Average relative error 0.42% -0.06% Confluence angle ([omicron]) No.

The

average relative error comparison between Fast Marching Methods and the method combining the Fast Marching Method with the Wavefront Construction Method.

Actual value Calculated Number ([x.sub.(0)](k)) value [mathematical expression not reproducible] 1 2.793 2.380227 2 3.591 3.946891 3 4.629 4.827466 4 5.546 5.339695 5 5.986 5.934339 6 7.018 7.281263 7 8.235 8.037644 Residual Relative Number ([q.sup.(0)](k)) error (e(k)) 1 0.412773 0.14779 2 -0.355891 0.09911 3 -0.198466 0.04287 4 0.206305 0.03720 5 0.051661 0.00863 6 -0.263263 0.03751 7 0.197356 0.02397

Average relative error f(%) 5.67

Using the WMPDS-S/B algorithm, the

average relative error ([error.sub.a]), the maximum relative error ([error.sub.m]), and RMSEP were obviously decreased (Table 3).

If the

average relative error MAPE satisfies the requirements, then the training is terminated and the reliability in future S steps of reliability series at the current period t is forecasted.

Compared with the predicted results of theoretical model, the combined method of modified damage law and fuzzy neural network can predict the fatigue life with a less

average relative error.

Figure 4 shows the reconstruction errors of the MART algorithm for two tested images, where the x-axis is the relaxation factor, and the y-axis is the

average relative error and root-men-square relative error.

(1) WARE is the

average relative error of all the edge weights.

TABLE 3:

Average relative error between simulated results by EOS and experimental data.

The results show that the average absolute error of the peak and off-peak is about 4.5 km/h, mean relative error the

average relative error is about 6.5%, and the mean square deviations are 2.38 km/h and 3.39 km/h.

We adopt variance, mean square error, the

average relative error, and maximum absolute value error to measure the goodness of fit.

From Table 1, when CFLN increases, the

average relative error becomes larger; however the CPU time is reduced.