Highway Research Record

Performance Modeling of Rural Road Pavements using Artificial Neural Networks

Date of Start:January 2005
College of Engineering, Thiruvananthapuram (R, I)

Scope and Objective
The rural roads are deteriorating at an alarming rate. Detailed PMS databases are required for the upkeep and maintenance of these roads. The pavement performance study conducted was on the National Highways and State Highways. The construction practice and the traffic characteristics for the rural roads are very much different from those for the NH and SH. Hence PPS models need modifications before they can be applied for the rural roads. The study aims to identify the parameters that really affect the performance of rural roads and to develop performance models for the rural roads from the performance data collected.
(i) To identify the causes of distress of rural roads from detailed distress survey and by destructive and non-destructive tests. (ii) To establish relationship between pavement deterioration and parameters identified for failure of pavements.
(iii) To develop models for the performance of rural road pavements using Artificial Neural Network.
(iv) To establish a relationship between Clegg impact value, CBR value and field dry density values for rural roads using Artificial      Neural Networks.
In this study, deterioration study will be carried out on different rural roads with varying construction quality and site conditions. Eight rural roads were selected for the present study.

Data regarding subgrade properties, drainage conditions, pavement component material properties, traffic volume and load, age and thickness of pavement was collected. Major distress modes are identified and detailed distress surveys were done periodically. One set of pavement performance data was collected. Deterioration models have been developed for rural roads relating different types of distresses, age of pavement, traffic and subgrade and pavement component material properties. The modeling has been carried out using Artificial Neural Networks. The CBR value, field dry density and Clegg impact value is to be collected for the pavement sub grade. A correlation is to be established between CBR value, field dry density and Clegg impact value for rural road pavements using Artificial Neural Network.

Findings and Conclusions
The factors that mainly affect the performance of rural roads are identified as construction quality and drainage. Neural network models were developed for construction quality and drainage. The main distresses identified on these roads were raveling, pothole, and edge failure. These distresses were modeled in neural network for the selected rural roads.
• Models were developed for Construction quality and Drainage. Drainage = f (FDD shoulder, Ponding, Camber of the shoulder,    thickness above subgrade)
   CQ = f (FDDsubgrade, Drainage, FDDshoulder)
• The deterioration models were developed for raveling initiation and progression, pothole progression, roughness progression    and edge failure.

Age of the pavement at the time of Ravelling initiation = f (Traffic, CQ, Drainage)
Raveling progression = f (traffic, initial ravelling, Drainage)
Pothole Progression = f (Drainage, traffic, CQ, THBM, MSN, Rvi, Phi)
Roughness = f (traffic, MSN, Age, Drainage, CQ, Ravelling area, Pothole area)
Edge failure progression = f (Drainage, CQ, Traffic)
Where CQ is the construction quality of the road, FDD is field dry density, THBM is the thickness of the bituminous layer, MSN is the modified structural number, Rvi is the initial raveling and Phi is the initial pothole area.

Reshmy D.S (2005), “Performance Modeling of Rural Road Pavements using Artificial Neural Networks”, M.Tech Thesis, Traffic and Transportation Engineering Division, Department of Civil Engineering, College of Engineering, Trivandrum