Estimating Layers' Structural Coefficients for Flexible Pavements in Costa Rica Road's Network Using Full Bayesian Markov Chain Monte Carlo Approach

Pavement layers' structural coefficients play a crucial role in managing flexible pavements using AASHTO 1993 design methods to create deterioration models for the pavement's structural capacity. Typically, a back-calculation approach can be used to estimate pavement layers' moduli, and this can be used to estimate the structural numbers. However, this estimation produces one single value with no associated uncertainty or reliability level for the structural coefficients. For this reason, this paper proposes a statistically robust approach that estimates each coefficient's probability density directly from the Falling Weight Deflectometer (FWD) readings with associated reliability through an associated confidence interval level. The data were obtained from local observations of FWD readings and layer thickness (from boreholes) of the Costa Rica National Roads network. More than 830 segments were considered for this study. OpenBugs software, an application for the Bayesian analysis using Gibbs sampling initially intended for biostatistical analysis and later extended to other fields, was used. Based on the materials, six different categories were considered. Two models prepared for a Costa Rica road network case study show that the estimated layer coefficients landed within commonly accepted ranges without relying on the back-calculation process. Two hundred thousand Markov Chain Monte Carlo (MCMC) iterations were run to evaluate the proposed approach's structural layer coefficients. The results proved that the second method showed smaller MC_error values compared to the first model. Furthermore, the proposed method demonstrated that it is possible to estimate coefficients for the whole road network through a hierarchical approach for various combinations of layers with different materials (granular, stabilized, macadam, etc.).