134
1.5 to 6.0 in. had an R2 of 0.958 with 192 number of cases analyzed.
Error analysis indicated that only 9 out of the 192 total cases had
predictive errors above 15 percent but all did not exceed 20 percent.
These occurred in pavements with value of 40 ksi and tx of 1.5 or 6.0
in.
Because of the high reliability of Equation 4.23, the database was
expanded to include other moduli and thickness combinations. Analysis
of the large data set (N = 400) resulted in Equation 4.24 with an R2 of
0.935. In 62 out of 400 cases, predictive errors were more than 20 per
cent. Three pavements had 40 percent or more prediction errors with one
of the them being 55.6 percent.
Additional evaluation of the practical limits of Equation 4.24
indicated that prediction accuracy was satisfactory even when the Ex
range was expanded to 600 ksi. Errors were generally less than 30 per
cent and only 7 out of 80 cases had predictive errors between 30 and 43
percent using Ex = 600 ksi. Overall assessment of this E3 prediction
suggests that the best accuracy is obtained from 3 to 6 in. of asphalt
concrete.
4.4.2.4 Subgrade Modulus, E,. Five equations were derived to pre
dict subgrade modulus from the 9-kip FW0 load deflections. The first
three equations (4.25 to 4.27) were essentially dependent upon individ
ual sensor deflections, Dg to D8, respectively. The other equations
(4.28 and 4.29) included two sensor deflections in each equation. All
five equations had R2 values greater than 0.997 for N = 400.
Considering the high R2 values obtained for Equations 4.25 through
4.29, their prediction accuracy is expected to be good. Also, a high
degree of accuracy is required for E^ prediction since small changes