Modeling the Joint Behavior of PCC Kansas Pavements Via Dynamic ANN Approach

Joint Deterioration of PCC pavement is a process that is controlled by a variety of quantifiable and other non-easily quantifiable parameters. In this study, artificial neural network (ANN) approach was used to correlate the time-dependent joint deterioration behavior with quantifiable experimentally-based aggregate durability parameters such as durability factor, percent expansion and freeze-thaw index. To achieve this objective, the historical Kansas pavement management system database, along with corresponding aggregate and core durability reports for a number of PCC sections, were combined to produce the needed time-dependent joint deterioration database. The resulting database was split into training, testing and validation sub-bases. Upon a number of ANN training, testing and validation processes, a time-dependent ANN-based joint deterioration model was developed. The developed dynamic model utilizes a number of aggregate durability factors to project the time-dependent joint deterioration behavior for up to 20 years after construction. Accordingly, model predictions for year n are sequentially used as inputs for predictions for year n+1. Overall, model predictions are noted to be logical and in good agreement with field observations.