Modeling asphalt pavement condition using artificial neural networks

Abstract

Pavement maintenance and rehabilitation (M&R) decision-making is an important component of pavement management systems. Timely maintenance of extensive pavement network is mandatory to ensure a safe and comfortable ride to the road users. Accurate and reliable prediction of current pavement condition is a valuable input for taking sound maintenance decisions and selecting appropriate rehabilitation or repair alternatives. Maintenance or repair decisions are usually not based on structural parameters due to certain limitations regarding the collection of structural data. Due to its absence pavement may demand substantial rehabilitation in future, which could have otherwise required only preventive maintenance.Therefore, in the present study, structural performance models are developed using the tools of computational intelligence. Artificial neural networks have been used to develop various structural condition prediction models for asphalt pavements using a variety of input data pertaining to structural, functional and environmental parameters, collected from actual field testing. The reliable correlations developed in this study are expected to popularize the implications of structural adequacy factors in pavement M&R decision-making and ease the work of transportation agencies in obtaining structural condition data. The neural network models are flexible with the addition/modification of data and work well even with the limited data.