Abstract:
Structural evaluation of road networks has always been a challenge for road agencies, especially in developing countries like Iran, from both economic and operational perspectives. Many researchers have sought innovative solutions such as developing structural indices or predictive models to address this issue. However, the proposed approaches either rely on costly equipment or have low accuracy. In this study two approaches were employed to address this issue: a weighted summation technique to develop the Pavement Structural Condition Index (PSCI), and a Random Forest Classification (RFC) model. Both methods utilized surface distresses, such as International Roughness Index (IRI), rut depth, and alligator cracking acquired by Laser Crack Measurement System (LCMS) to assess the structural condition. Additionally, a threshold value of PSCI was identified that performs as a crucial discriminator between poor and sound pavement structural status, based on the required structural overlay thickness and using the Receiver Operating Characteristics (ROC) curve analysis. Ultimately, the efficacy of PSCI and RFC models in predicting the structural adequacy of pavement sections was validated with an independent dataset. Consequently, the PSCI model demonstrated an accuracy of 0.842 and a precision of 0.964, while the RFC model achieved superior performance with an accuracy of 0.984 and a precision of 0.976. While the RFC model outperformed the PSCI in identifying structural deficiencies, the PSCI framework provides a practical, cost-effective alternative for network-level pavement condition monitoring, particularly in scenarios where implementing machine learning models is not feasible.
Authors: Barzegaran J, Karimi S M, Qorbaninik M, Swarna S T, Hossain K.
Link(s) for the Paper: Journal Website | ResearchGate