ABSTRACT
The pavement management system is of utmost importance for state agencies. Pavement performance prediction is an important component of the pavement management system. This paper aims to develop pavement performance prediction models with the least number of explanatory (independent) variables to predict the pavement performance for future planning of maintenance and rehabilitation (M&R). The artificial Neural Networks (ANNs) approach is utilized to develop the Jointed Reinforced Concrete Pavement (JRCP) performance prediction models. The study was carried out with the database collected from Long Term Pavement Performance (LTPP) program. The International Roughness Index (IRI), in meter per kilometer, was mainly used for dependent variables while initial IRI, pavement age, concrete pavement thickness, base/subbase thickness, average contraction spacing of pavement, cumulative equivalent single axle load (CESAL), base/subbase type (granular, stabilized), climatic region (wet-freeze, wet non-freeze), and construction number (CN) were used in the search of best models. There were several models with constructive statistical measures with a good fit of observed and predicted data. However, the best performing ANN model resulted in promising statistical measures (i.e., R2 = 0.93) which can be used to successfully estimate the IRI values considering the M&R history of the pavements.
