Abstract:
Routine pavement maintenance and rehabilitation are the most crucial parts of the pavement management system (PMS) to keep the roads in serviceable condition. Depending on the severity of the surface deterioration, rehabilitation is prioritized over maintenance to increase the pavement’s lifespan. The International Roughness Index (IRI) serves as a key functional performance indicator for assessing overall surface roughness and guiding the selection of appropriate treatment types. Therefore, an efficient IRI model capturing the influence of climate and treatment types is the prerequisite to making any fruitful decision on proper treatment types. Utilizing the United States flexible pavement sections data from the Long-Term Pavement Performance (LTPP) program, the research examined the effectiveness of different rehabilitation treatments across various climate conditions through the IRI prediction models. The fuzzy c-means clustering technique was used to categorize the climate conditions into three clusters. Moreover, feature importance analysis was performed to evaluate the influential variables in IRI prediction model. This study employed multiple linear regression (MLR) and artificial neural networks (ANNs) to predict IRI, leveraging machine learning techniques to capture both linear and non-linear relationships among variables. The results revealed that both models performed well under different climatic conditions with reasonable accuracy (R2 > 80% and MSE < 0.2). The results showed that milling the existing pavement and overlay with hot mix recycled asphalt concrete (MILL_REC_AC_OL) consistently outperformed the control asphalt concrete overlay (AC_OL) in all climatic conditions, while recycled asphalt overlay (REC_AC_OL) was more effective than AC_OL in moderate and hot climate. Warm mix asphalt overlay (WMA_OL) seemed to perform better than AC_OL in cold and moderate climate conditions but less effective in hot regions.
Contributors: Khan S, Barzegaran J, Swarna S, Hossain K.