A Performance Prediction Model for Continuously Reinforced Concrete Pavement Using Artificial Neural Network

The pavement performance prediction models are key components of pavement management systems. Predictions models are used for determining the future condition of pavement as well as types of maintenance and rehabilitation actions needed to keep the operation of transportation systems continuous. Accordingly, maintenance and rehabilitation actions must be prioritized due to limited budget allocations. Mississippi Department of Transportation (MDOT) utilizes Hidden Markov probability models. However, varying traffic conditions and unusual events in the operation are not reflected in these models due to the nature of probabilistic modeling. A new model using a more inclusive and powerful approach to predict the future condition of the pavement is needed. In this study, the distress data from Continuously Reinforced Concrete Pavement (CRCP) sections in Mississippi was used to develop a performance prediction model using Artificial Neural Networks (ANNs) approach. Additionally, rehabilitation actions were included as part of the model inputs to study the impact of rehabilitation actions on the model. The performance of all the developed models showed a good agreement between observed and predicted condition measures. However, only one model with the best statistical accuracy was selected to be utilized as the best performing model, which can be used for the prediction of CRCP performance within the allowable time range.