This paper provides an overview of typical procedures currently used for pavement evaluation using Non Destructive Testing (NDT) deflection data, primarily focusing on backcalculation.

Some of the more typical problems encountered in these approaches are briefly discussed.

Some of the critical issues related to fundamental theoretical assumptions of static loading as well as material continuity, homogeneity and elastic behaviour is addressed, particularly in the context of validation of backcalculation results.

Introduction
Structural evaluation of pavement deflection response using Non Destructive Test (NDT) data has been growing since the introduction of the Benkelman Beam at the WASHO Road Test in the early 1950's.

Developments in analytical techniques, coupled with improved deflection measurement capabilities, have resulted in the current so-called backcalculation techniques widely employed in pavement evaluation.

This paper provides an overview of existing techniques used for structural analysis of pavement NDT deflection data, discusses some of the issues and shortcomings of these procedures, and provides some conjecture on expected and possible future developments in the field.

Usually some statistical measure of deflections on a pavement section is compared with a "tolerable" deflection level for that section under the expected traffic.

If the measured value exceeds the tolerable deflection then an empirical procedure determines the corrective measure required, usually an overlay, to reduce the measured deflections to the tolerable level.
Examples of this approach include The Asphalt Institute's MS-17 ¹) and CalTrans' Test Method 356²).
In some states maximum deflections are monitored during spring thaw and load restrictions are placed when the thawing pavement's deflection reaches a certain level.

Empirical use of deflection basin data usually involves one of the "basin parameters" which combine some or all of the measured basin deflections into a single number.

With a trend towards mechanistic pavement analysis and design, which is based on fundamental engineering principles, the use of deflection data has become more sophisticated.

Complete deflection basins are used, in a procedure known as backcalculation, to estimate in-situ elastic moduli for each pavement layer.

Knowledge of the existing layer thicknesses are typically necessary for this procedure.
The backcalculated moduli themselves provide an indication of layer condition.

They are also used in an elastic layer or finite element program to calculate stresses and strains resulting from applied loads.

These stresses and strains are used with fatigue or distress relationships to evaluate damage accumulation under traffic and predict pavement failure.

They can also be used to evaluate corrective measures such as overlays, rehabilitation or reconstruction.
It is these mechanistic analyses of pavement deflection that this paper is intended to address.

Briefly, the backcalculation procedure involves calculation of theoretical deflections under the applied load using assumed pavement layer moduli.

These theoretical deflections are compared with measured deflections and the assumed moduli are then adjusted in an iterative procedure until theoretical and measured deflection basins match acceptably well.

The moduli derived in this way are considered representative of the pavement response to load, and can be used to calculate stresses or strains in the pavement structure for analysis purposes.

Calculation of theoretical deflections, and the subsequent stress or strain calculations, currently typically involve linear elastic theory.
Application of elastic theory may be through the use of:

1. Traditional layered elastic programs based on numerical integration procedures such as ELSYM5, CHEVRON (various versions), BISAR and WESLEA.
2. The Odemark-Boussinesq transformed section approach rather than numerical integration.
3. Finite element programs, either those that have been specifically oriented towards pavement analysis, such as ILLI-PAVE or MICHPAVE, or general structural analysis programs such as SAP (various versions), ANSYS, ABACUS, ADINA, etc.
4. Plate theory such as the Westergaard solutions for PCC pavements.
5. Neural networks trained to reproduce results that emulate one of the above applications. ³⁾⁴⁾
   Backcalculation