Experiment Design Approach to Nondestructive Testing of Pavements

Abstract

An analysis of the errors involved in both the nondestructive testing data collection and the modeling of a pavement structure is presented. The errors are separated into: (1) Random errors associated with the measuring devices; (2) random errors due to the intrinsic variability of the materials and pavement structures; and (3) systematic errors associated with the modeling of pavement structure, loading conditions, and material characterization. An analysis of the backcalculation process is conducted to investigate the effect of the errors on the backcalculated moduli. An experimental design approach is applied, using the first‐order approximation for the expected values and variance of the layer moduli. It is shown that (1) Mean values can be backcalculated from an average deflection bowl in which all random errors are reduced to a negligible value; and (2) the intrinsic variability of the materials can be correctly estimated only if the random errors associated with the measuring devices are eliminated. In recognition of these two findings, two testing programs are proposed, denoted as programs I and II. A case study with measured deflection data illustrates and supports the findings of the analysis and of the experimental design approach: (1) The backcalculated modulus of the upper layer depends strongly on the accuracy of the sensor at the center of the loading plate; (2) the number of tests per location required for obtaining a set of reliable moduli is about 4; and (3) testing program I is superior to testing program II.