Dr Ze Zhou Wang

Strategic Themes

Smart Materials

Research Project

Title: An innovative zonation-based machine-learning methodology for studying the interactive impacts of traffic, microclimate and natural hazards on pavement deterioration

Theme: Smart Materials

Abstract: 

Pavement monitoring and health evaluation play a key role in highway operation and maintenance worldwide. As a result, extensive research has been devoted to improving pavement health evaluation methods. Determining the bulk material properties is one of the most critical steps in estimating a pavement’s remaining service life. Although various strategies exist to estimate these properties, non-destructive methods remain the most popular choice. The Traffic Speed Deflectometer (TSD) is an emerging non-destructive testing instrument for evaluating the in-situ stiffness of pavements. However, TSD does not directly measure the stiffness properties; instead, it measures the slopes of vertical deflection at a set of points along the right rear wheel of a measuring truck trailer travelling at normal traffic speed. Therefore, TSD measurement data needs to be interpreted to obtain the in-situ stiffness of pavements, and back analysis is commonly employed for this task.

A typical back analysis requires three components: (i) a calculation model to simulate pavement responses under the dynamic TSD loading, (ii) TSD measurements, and (iii) a back calculation algorithm. The calculation model generates simulated pavement vertical deflections with a given set of stiffness parameters, which are then compared against the actual TSD measurements. The back analysis algorithm then iteratively adjusts the input stiffness parameters until the generated vertical deflections reasonably match the actual measurements. At this point, the calibrated stiffness values are considered representative of the in-situ pavement properties. Many algorithms have been proposed for this task. However, existing algorithms are often based on deterministic analysis, thereby overlooking the inherent uncertainties in material properties. In this regard, this work pioneers Bayesian probabilistic back analysis of TSD measurements. Bayesian back analysis considers both the inherent uncertainties in material properties and uncertainties in TSD measurements, thereby providing more comprehensive information of in-situ stiffness properties for decision-making.

This work starts from a parametric analysis that investigates the effects of uncertainties in surface, base, and subgrade layer modulus. The results show that pavement deflection is most sensitive to uncertainty in subgrade layer modulus, followed by base layer modulus and surface layer modulus. This ranking implies that back analysis of TSD measurement is most effective in estimating subgrade layer modulus and least effective for surface layer modulus. The dominant sensitivity to subgrade layer modulus makes the simultaneous back analysis of surface, base, and subgrade layer modulus a practical challenge. Bayesian back analyses using simulated TSD measurements are then carried out to corroborate these results. In practice, it is recommended to obtain subgrade layer modulus through other means so that surface and base layer moduli can be more effectively interpreted using TSD measurements.

Project TRL: Start: TRL 3; End: TRL 6

Datasets: TSD measurement data