Multi-input performance prediction models for flexible pavements using LTPP database

Abstract

Pavement performance prediction is a primary concern for pavement researchers and practitioners. The impact of climatic conditions and traffic characteristics on pavement performance is indisputable. The main objective of this study is to investigate the combined effect of both climate and traffic loading on pavement performance. Multi-input performance prediction models in terms of the well-known Pavement Condition Index (PCI) are proposed. The Long-Term Pavement Performance (LTPP) database is used for the models development and validation. Data from 89 LTPP sections including 617 observations from the Specific Pavement Studies (SPS-1) with no maintenance activities are collected. These data cover the four climatic zones (wet, wet freeze, dry, and dry freeze) in the USA, different pavement structures, and different levels of traffic loading. Based on these data, PCI prediction models are developed using two modeling approaches: multiple linear regression analysis and artificial neural networks (ANNs). The proposed models predict the PCI as a function of climatic factors, namely average annual temperature, standard deviation of monthly temperature, precipitation, wind speed, freezing index, total pavement thickness, and weighted plasticity index. Additionally, traffic loading, expressed in terms of the classical equivalent single-axle loads, is considered. The regression model yielded a coefficient of determination (R²) value of 0.80, whereas the ANNs model results in a relatively higher R² value of 0.88. The proposed models are not only simple and accurate; they also have the potentials of being adopted in countries experiencing similar climatic conditions and traffic loading.