Roughness Prediction Models Based on Variable Distress Parameters using Neural Network and MLRA for PMGSY roads

Abstract

Roughness indicates the performance and quality of flexible pavement. Roughness is measured
based on International Roughness Index (IRI). Roughness includes smoothness and frictional properties.
IRI directly relates to safety and Vehicle Operating Cost (VOC). To evaluate IRI, more models are
developed till today. All the models are not applicable for all types of pavements. Many models are
implemented only for standard road with high traffic volumes. The standard models are dependent of
local condition variable parameter of pavement. The variable parameters are as soil condition, pavement
material composition and traffic. In this paper, Takagi-Sugeno-Kang (TSK) and Multiple Linear
Regression Analysis Model (MLRA) is proposed to evaluate performance prediction of PMGSY roads.
Model includes variable and distress parameter such as cracking & potholes for prediction of roughness.
In India, the Thiruvallur District of Tamil Nadu is taken for the study. The proposed model utilizes 120
road data for prediction and the remaining 53 road data for validation of the model. The TSK and MLRA
models achieved 95% of accuracy in the prediction of performance.