Optimised Surface Condition Classification of Flexible Road Pavement Using AutoWEKA Model

Abstract

Abstract The development of pavement management tools using intelligent algorithms requires a robust form of data mining – data classification for efficient and reliable analysis. The aim of this study is to investigate and optimally classify the surface condition of flexible road pavement along 60 km length of the Zaria – Kaduna Federal Highway in Northern Nigeria for maintenance decision. The study used data mining technique for the classification of pavement surface condition into good, satisfactory, fair, poor, very poor, serious or failed. A field survey was carried out to examine the surface area and length of various surface defects such as cracks, potholes, rutting and edge failure within Chainages measuring 200 meters apart, which was used to compute the Pavement Condition Index (PCI) values and section classification in accordance with procedures stated in ASTM D6433. The AutoWEKA model of Waikato Environment for Knowledge Analysis (WEKA) software was used to optimally classify the surface condition of the highway. Results indicated that 79.67% of the 300 total instances considered by the model were correctly classified while 20.33% of the instances were incorrectly classified. The optimum surface condition classification showed that worse pavement surface conditions of the sampled site were ‘Poor’, ‘Very Poor’ and ‘Failed’ at 77 (32.22%), 51 (21.34%) and 54 (22.59%) instances respectively of the correctly classified 239 instances out of the 300 total instances sampled. Based on its present condition, 76.15% of the road segment was bad. The rehabilitation or reconstruction of the Zaria – Kaduna Federal Highway was therefore recommended for improved condition and optimum performance.