Transportation Planning through Pavement Performance Prediction Modeling for Botswana Gravel loss Condition

Abstract

Botswana is a Southern African country with an area of about 582,000 sq. km and its small population of about 2 million people. The road transportation network has grown beyond all expectations since independence in 1966. Out of the 18,300 km Botswana Public Highway Networks, gravel road networks are significant in providing access to rural areas where the majority of the population lives. Modelling of gravel loss conditions are required in order to predict their conditions in the future and provide information on the manner in which pavements perform. Such information can be applied to transportation planning, decision making processes and identification of future maintenance interventions. The results of previous attempts to develop gravel loss condition forecasting models using multiple linear regression (MLR) approach have not been reliable. This paper intended to develop accurate and reliable performance models which best capture the effects of gravel loss condition influencing factors using Feed Forward Neural Network (FFNN) modeling technique. As extension of knowledge in unpaved road transportation network, FFNN trained with Levenberg-Marquardt (L-M) method was used to develop gravel loss performance prediction model for Botswana gravel road networks to achieve a reliable result of a higher coefficient of determinant R2 = 0.94 compared with MLR analysis of R2 = 0.74.