Predicting pavement condition index using artificial neural networks approach

Pavement Condition Index (PCI) is a numerical assessment of pavement conditions based on existing distresses. The PCI values are used for pavement management and rehabilitation programs. Calculating the PCIs using conventional method relies on collecting relevant field data (such as distresses types and severity) by visual inspection method. The collected data are processed to estimate the PCI values, which is a lengthy process that requires technical experience. This research aims to model the relationship between distresses type and severity and PCIs via straightforward and adaptive model. Therefore, Artificial Neural Networks (ANN) capabilities are employed to predict the PCI values of the different sections, thus reducing the required efforts and technical experiences to estimate PCI values. Moreover, the use of ANN enables the possibility of introducing new localized variables, such as the presence of manholes in pavement sections. The total of 348 directional sections from 10 different roads located in the City of Nablus, Palestine were examined to collect the distresses-related data and to estimate the corresponding PCI values using ASTM 6433 07 method. The results revealed low correlation between distresses and PCI, where the highest absolute correlation between PCI and any distress type and severity did not exceed 0.38. The results indicated that the ANN model is capable to predicting the PCI with high level of reliability, with an R-2 value of 0.9971, 0.9964 and 0.9975 for training, validation and testing datasets, respectively. The regression slope between observed and predicted PCIs ranges between 0.9964 and 0.9974. (C) 2021 THE AUTHORS. Published by Elsevier BV on behalf of Faculty of Engineering, Ain Shams University.

Automated summary

This research models the relationship between distress types and severity and Pavement Condition Index (PCI) using Artificial Neural Networks (ANN). The results show that the ANN model can predict PCI with high reliability.