Modeling of Pavement Roughness Utilizing Artificial Neural Network Approach for Laos National Road Network

Abstract

The International Roughness Index (IRI) has become the reference scale for assessing pavement roughness in many highway agencies worldwide. This research aims to develop two Artificial Neural Network (ANN) models for Double Bituminous Surface Treatment (DBST) and Asphalt Concrete (AC) pavement sections using Laos Pavement Management System (PMS) database for National Road Network (NRN). The final database consisted of 269 and 122 observations covering 1850 km of DBST NRN and 718 km of AC NRN, respectively. The proposed models predict IRI as a function of pavement age and Cumulative Equivalent Single-Axle Load (CESAL). The obtained data were randomly divided into training (70%), validation (15%), and testing (15%) datasets. The statistical evaluation results of the training dataset reveal that both ANN models (DBST and AC) have good prediction ability with high values of coefficient of determination (R2 = 0.96 and 0.94) and low values of Mean Absolute Error (MAE = 0.23 and 0.19) and Mean Squared Percentage Error (RMSPE = 7.03 and 9.98). Eventually, the goodness of fit of the proposed ANN models was compared with the Multiple Linear Regression (MLR) models previously developed under the same conditions. The results show that ANN models yielded higher prediction accuracy than MLR models.

Machine Learning for Prediction of the International Roughness Index on Flexible Pavements: A Review, Challenges, and Future Directions

“…Studies have demonstrated the ML algorithms' effectiveness in predicting pavement performances [9,30,[52][53][54][55][56][57][58][59][60][61][62][63][64][65][66][67]. Likewise, commonly used algorithms include ANN, SVM and support vector regression (SVR) [40], adaptive boosting (AdaBoost) [68], random forest (RF) [69], gradient boosting decision trees (GBDT) [70], categorical boost (CatBoost) [71], and ensemble models [50].…”

“…Likewise, Gharieb et al [58] developed two ANN models for double bituminous surface treatment (DBST) and asphalt concrete (AC) pavement sections within the National Road Network (NRN), using the Laos PMS database to predict IRI by analyzing only pavement age and traffic load, surpassing traditional MLR methods. Furthermore, Abdulaziz et al [30] developed ANN models that accurately predict the IRI by analyzing the effects of pavement distress across two climate regions in North America.…”

A Systematic Literature Network Analysis: Research Mapping of International Roughness Index

“…In 2022 there will be two IRI research publications; the first is a study in Laos to predict IRI models using ANN for double bituminous surface treatment and asphalt concrete pavements. This model uses Age, CESAL and YESAL parameters as input, with the result that the ANN model is more accurate than the multiple linear regression model [98]. The second study used IRI prediction models using several soft computing techniques, including gradient boosting method and random forest for flexible pavement to determine the effect of maintenance, traffic, and climate conditions.…”

Pavement Quality Evaluation Using Connected Vehicle Data

“…Current means of evaluation of pavement conditions often include estimation of the International Roughness Index (IRI) with an inertial profiler. IRI is widely employed due to its stability overtime and has been applied around the world [ 32 , 33 , 34 ]. IRI is less subjective than other pavement performance indicators as it is calculated from the road profile.…”