Pavement structural evaluation based on roughness and surface distress survey using neural network model Mansour Fakhri a,⇑ , Reza Shahni Dezfoulian a,b a Department of Civil Engineering, K.N. Toosi University of Technology, Tehran 1996715433, Iran b Department of Transportation Planning, Road, Housing & Urban Development Research Center (BHRC), Marvi St, Sheikh Fazlollah Noori Exp, Tehran 13145-1696, Iran highlights  The paper presents a practical method for assessing pavement layers condition.  Decreasing the limitation of using costly structural assessment devices (FWD).  Strong relation between structural parameters and pavement performance indices.  Project field survey was conducted, using FWD, RSP and visual survey (PASER).  The superiority of ANNs over regression models was investigated. article info Article history: Received 10 September 2018 Received in revised form 14 January 2019 Accepted 25 January 2019 Keywords: Structural evaluation Pavement surface distress Roughness Neural network PASER Falling Weight Deflectometer (FWD) abstract Evaluation of pavement condition, which determines pavement maintenance and rehabilitation necessi- ties, is inevitable using structural or non-structural methods. Since factors such as cost and time required for testing limit the use of structural assessment devices, the development of cost-effective methods should be investigated. In this paper, a practical solution has been presented for pavement structural evaluation which is considered as a useful method for assessing pavement layers condition and identify- ing rehabilitation needs. With this in mind, we developed a relationship between deflection bowl param- eters derived from Falling Weight Deflectometer (FWD) and two pavement performance indices, International Roughness Index (IRI) and Pavement Surface Evaluation and Rating index (PASER), by the use of Artificial Neural Network (ANN) and regression models. To obtain the required data, project field surveys were conducted from 318 sections of the main roads of Kermanshah and Ilam provinces in Iran. The results show that our model provides a satisfactory correlation between IRI, PASER, and structural indices which are based on deflection measurements. By comparing the results of ANN and regression models, the superiority of ANN performance over non-intelligent models is appreciable. The findings of this study indicate that using both IRI and PASER indices leads to accurate structural pavement evaluation. Ó 2019 Elsevier Ltd. All rights reserved. 1. Introduction Roads, as a communicative-economic infrastructure and a valu- able asset, provide a good basis for transporting passengers and freight in the transportation network. Certainly, preserving the communicative role of the road in safe conditions and acceptable serviceability level is not possible without road planning and man- agement. Pavement management includes components such as identifying activities, planning, evaluating pavement performance, allocating resources, and analyzing results [1]. Due to pavement deterioration and pavement maintenance and rehabilitation needs, the implementation of the Pavement Management System (PMS) as a reliable decision-making tool is necessary. The existence of maintenance management methods and tools will help decision makers to choose the most cost-effective treatment at both the network and project level [2,3]. Determining the maintenance activities, required budget, and predicting pavement condition are carried out by assessing the pavement condition in the pavement management system [2,4–6]. The pavement condition is usually evaluated by measuring the ride quality or roughness, surface distress, structural adequacy and pavement friction [1]. Pavement structural assessment, using destructive or non-destructive methods, provides valuable https://doi.org/10.1016/j.conbuildmat.2019.01.142 0950-0618/Ó 2019 Elsevier Ltd. All rights reserved. ⇑ Corresponding author. E-mail address: fakhri@kntu.ac.ir (M. Fakhri). Construction and Building Materials 204 (2019) 768–780 Contents lists available at ScienceDirect Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat