U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-RD-03-093
Date: August 2006
Study of Long-Term Pavement Performance (LTPP): Pavement Deflections
Appendix G. FWD SN 130, August 25, 1994–September 7, 1994
This previously reported sensor position error was located in the database using an automated version of SLIC which employs a model that was specifically chosen to predict the position of sensor 7 with a close to zero overall bias and the best possible precision (see appendix B). With this method, after regressions are fit to the model 7 prediction based on the data from sensors 3 through 6, two plots for each FWD are created. One is a plot of the sequence of R2 terms, and the other is the predicted position of sensor 7. The first graph shown in this appendix (figure 30) is a plot of all of the R2 values for FWD SN #130, for the regression that best fits each test date using the data from sensors 3 through 6. The period of time was 1994 through the end of 1996, and the data are for lane 1, drop height 4 FWD tests. The second graph in this appendix (see figure 31) uses the same period of time and data to predict the position of sensor 7 for each test date.
In both figures 30 and 31, it can be seen that there is a serious anomaly starting at test sequence 50, which corresponds to August 25, 1994. The average R2 value during the period of time in question was only 0.881, whereas when the FWD sensors used in the regression are correctly placed, the regression fits so well that the values of R2 are generally 0.998 or better. Based on a detailed analysis of the data, for example as shown in the succeeding three figures in this appendix, it is clear that the sensor positions used were 0, 8, 12, 18, 36, 48, and 60 inches. All three section-specific plots depict these sensor positions, with d5 and d6 simply shifted 12 inches into the wrong sensor holders, leaving the 24-inch sensor holder inadvertently empty.
In figures 32, 33, and 34, the gray lines and data points are plots of both actual and erroneous data for d5 and d6 for the particular date. The gray lines that are parallel to the rest of the data are the correct plots, with d5 and d6 set to 61 cm (24 inches) and 91.4 cm (36 inches), respectively.
Because of this information, and the previous information supplied to FHWA, it can be concluded with certainty that d5 and d6 were not positioned correctly, at 61 cm (24 inches) and 91.4 cm (36 inches) respectively; rather, they were positioned at 91.4 cm (36 inches) and 121.9 cm (48 inches) (or very close to these positions) respectively, on FWD SN #130 between August 25, 1994, and September 7, 1994. These dates correspond to the dates when the data generated included lane 1 tests from drop height 4. This period of time may need to be extended slightly, if other tests were conducted along different lanes or at different drop heights. In any case, FWD tests conducted on or before August 22, 1994 and on or after September 8, 1994, clearly show that d5 and d6 were correctly positioned at 61 cm (24 inches) and 91.4 cm (36 inches) (as per protocol), respectively.
Figure 30. Graph. R2 model for d7 prediction, unit #130, 1994–96.
Figure 31. Graph. Predicted position of d7, unit #130, 1994–96.
Figure 32. Graph. SLIC plots for section 20–0101 including unit #130, August 1994.
Figure 33. Graph. SLIC plots for section 20–0111 including unit #130, August 1994.
Figure 34. Graph. SLIC plots for section 20–3060 including unit #061, September 1994.