Statistical Model to Detect Voids for Curled or Warped Concrete Pavements

Abstract

A statistical classifier was developed to interpret falling weight deflectometer data for the detection of voids under jointed concrete pavement slabs. The classifier was trained with the Seasonal Monitoring Program sections in the Long-Term Pavement Performance (LTPP) database and data from the Minnesota Road Research Facility. A two-level cross-validation process was used to assess the performance of existing void detection methods, according to a threshold of a single variable, and the least absolute shrinkage and selection operator (LASSO) classifier, which is based on several variables. Simple void detection methods based on the normalized 9,000-lb deflection were found to perform better than void detection methods based on variable deflection analysis. The LASSO classifier outperformed any of the existing void detection techniques. The LASSO classifier was validated through two field trials in Pennsylvania and an LTPP general pavement section in which significant faulting had developed.