Backcalculation of damage density of in‐service asphalt pavements using artificial intelligence‐based finite element model updating

Abstract

Fatigue damage is a common distress significantly affecting asphalt pavements in the performance, economy, and safety aspects. Accordingly, this paper aims to propose a methodology to backcalculate a continuum damage mechanics (CDM)‐based damage density of in‐service asphalt pavements from field nondestructive testing (NDT). An artificial intelligence (AI)‐based finite element (FE) model updating was applied with the falling weight deflectometer (FWD) test results for the calibrations of the damage density and supporting layer moduli. The development of the damage density with pavement service time was categorized and fitted with proposed analytical models. The damage density was compared with field observations on the fatigue cracking performance of asphalt pavements. It was found that the damage density had close relations with the material type, structural configuration, and fatigue cracking area of the pavement. This study shows a promising application of the damage density in the evaluation and prediction of the fatigue damage of asphalt pavements, which will lead to a timely and cost‐effective pavement maintenance and rehabilitation.

Feasibility of determining asphalt pavement condition from falling weight deflectometer test and finite element model updating

“…It can be seen as a potential warning of the difficulty in backcalculating the base modulus from sensor deflections. The results of sensitivity analysis can serve in the multi-step backcalculation scheme that associates the corresponding sensor deflections with the more influential layer properties as in previous studies [1,13].…”

“…Artificial intelligence (AI)-based FE model updating has recently gained increasing applications in determining the deterioration condition of asphalt pavements from NDT [1,10,13]. In this method, FE models were constructed in commercial packages as pavement models to simulate asphalt pavements in NDT.…”

“…As a developing methodology, the AI-based FE model updating needs to achieve high accuracy and reliability of the backcalculated results, which calls for the feasibility investigation in this study. Compared with its previous applications [1,10,13], the consideration of viscoelasticity and nonlinearity of asphalt pavement materials enlarges the number of model parameters to be backcalculated in this application. Accordingly, issues hindering accurate, stable and robust backcalculation results are more likely to occur in any of the four parts.…”

“…Deterioration of the asphalt layer can be characterized by degradation of the relaxation modulus of the asphalt mixture [5]. The master curve model can be transferred to the Prony series model at specific temperatures as equation (3.3) for viscoelastic material modelling in ABAQUS and COMSOL [1,15],…”

“…This section was investigated in the project and its basic layer material properties such as density and Poisson's ratio, structural parameters such as layer thickness and loading configuration such as the radius of the loading plate and the distances of the measuring sensors to the loading centre can be found in the final report [22] • calculate relaxation modulus with specific δ, α, β and γ in the time range 0.001-l00 s using equation ( 3 and a published paper [6]. This information was directly utilized to construct the pavement FE model as in previous applications [1,13].…”