Determination of Effective Structural Number based on IRI and Surface Distress Using Regression and Neural Network Model

Being aware of the pavement structural condition can play an effective role in implementing an effective pavement management system. An effective structural number can be considered for pavement design, maintenance and rehabilitation, detection of homogeneous sections, prediction of pavement deterioration, and identification of sections requiring additional field tests. One of the most common methods of calculating structural number is the use of non-destructive methods and equipment, such as Falling Weight Deflectometer (FWD). However, the cost of data collection, the need for traffic control and speed constraints are important things that have caused the use of the FWD to be limited at the project and especially the network level. In this paper, it is tried to determine the effective structural number by considering roughness and surface distress with fast and relatively economical methods. For this purpose, 52 km of arterial roads in Kermanshah and Ilam provinces were studied with different specifications such as traffic, age, pavement thickness and distresses. Also, for the determination of deflection, Roughness and PASER index (surface distress), assessment devices such as FWD, RSP, RD-3VV, and visual survey were used respectively. In order to determine the effective structural number, the roughness index (IRI) and PASER index, were considered as input variables in modeling. Also, power regression relation, polynomial linear and neural network models with different architectures were used. The results of the proposed model show that it is possible to determine the structural number with high accuracy and low error. In addition, the superiority and ability of neural network models to estimate the effective structural number, is significant compared to other regression models.