Modelling the effect of various factors on the condition of pavement marking

Abstract

Poorly maintained pavement marking might certainly contribute to road accidents. The cost of road accidents is estimated to be 10–25 billion Canadian dollars annually. Accordingly, it is necessary for municipalities to develop a strategic cost-effective plan in order to renew and restripe pavement markings. Therefore, the objective of the present research is to model the effect of various factors on pavement marking conditions. Data on Alkyd paint pavement marking material are collected from the city of Ottawa, Ontario, Canada. Since the collected data from municipalities in Canada always include input variables and fail to provide output variable(s) (e.g. condition), an unsupervised neural network (UNN) model is first developed to generate the condition of pavement marking (output). Then, regression and neurofuzzy models are developed based upon the results of UNN model. The developed models are validated in which they show satisfactory results. A sensitivity analysis is performed to show the effect of changing the input variables on the developed models’ output(s).