Rubblization is an effective rehabilitation method for deteriorated Portland Cement Concrete (PCC) pavements due to its low initial cost, minimum traffic disruption, and ability to minimize reflective cracking in asphalt overlays. However, the loss of strength in PCC slabs due to rubblization creates the demand for a subgrade that is strong enough to handle traffic after rubblization. AASHTO recommends Falling Weight Deflectometer (FWD) testing of PCC pavements before rubblization to ascertain the subgrade strength after rubblization.
However, the existing deflection analysis methods do not adequately handle the change in stress states in the subgrade before and after rubblization, and therefore result in erroneous prediction of subgrade strength.
The primary objective of this study is to develop an analysis method that allows the realistic estimation of subgrade strength after rubblization from deflection measurements on intact PCC slabs before rubblization. As a forward model, stress/strain dependent nonlinear subgrade models were incorporated into a finite element analysis. ABAQUS and NCPAVE, a finite element code developed at North Carolina State University, were used in the analysis. Multi-load FWD testing was conducted in the field to generate varying stress states in the subgrade under intact PCC pavements. Based on the synthetic database generated from the finite element analyses, a number of relationships were developed using regression and Artificial Neural Network (ANN) approaches to predict the coefficients in the nonlinear subgrade model from multi-load FWD deflections. The verification study was performed on the resulting algorithms using limited field data derived from US 29 in Guilford County and I-85 in Rowan County. The research suggests that the stress-based regression approach, which determines the nonlinear coefficients by regressing between the subgrade moduli and stresses predicted at several radial distances from the FWD load from multi-load deflections, is the most promising method of analysis. The research team strongly recommends the further verification of this procedure using additional field data before the implementation.
