127
It was shown in Figure 4.20 that a simplified equation originally
developed from field measured Dynaflect data could be used to predict E4
values varying from 10.0 to 100.0 ksi. The simplified equation
= 5.4 (D T1-0 Eqn. 4.35
was used to predict all the data set originally used to develop Equation
4.16. The R2 value obtained from linear regression of predicted and
actual E values was 0.830. However, when the extreme E values (0.35
and 200.0 ksi) were deleted, the R2 increased to 0.982. Thus, within
the range of 10.0 to 100.0 ksi of subgrade modulus, Equation 4.35 is
considered to be simple and accurate enough to predict E4.
4.4.2 Prediction Accuracy of FWD Equations
4.4.2.1 Asphalt Concrete Modulus, E1: Three equations were devel
oped to predict Ex for different ranges of asphalt concrete thickness,
tj. Equation 4.18 is valid for t: values varying from 1.5 to 8.0 in.
The prediction accuracy of this equation is summarized in Table 4.8. In
this table, the number of cases with percent predictive errors falling
within certain limits are listed for varying tx values. The table
illustrates that higher errors occur at tx values of 1.5 and 3.0 in. As
tx increases, the percent error reduces substantially with t1 equal to
6.0 in. having most cases with less than 20 percent prediction error.
Because of the high errors obtained with tx values of 1.5 in. in
Equation 4.18, these values were deleted from the data set and the
remaining data were analyzed to obtain Equation 4.19. This equation
thus holds for tx values from 3.0 to 8.0 in., and had an R2 of 0.979 for
320 number of cases. The prediction accuracy of Equation 4.19 is also