Analysis Of Long-term Pavement Performance Of Flexible Pavement Based On Statistical Theory

Long-term pavement performance(LTPP)is a fundamental and vital research topic in the field of pavement engineering.LTPP research investigates pavement deterioration of in-service pavement sections and its influencing factors by monitoring and collecting various types of data related to pavement deterioration such as pavement performance data,traffic load data,and climate data.LTPP research can provide theoretical support for pavement design,construction,maintenance and management.This study employs statistical methodologies and data mining techniques to explore and analyze pavement performance data of in-service flexible pavement and weigh-inmotion(WIM)data collected by the Long-Term Pavement Performance program managed by the Federal Highway Administration(FHWA).While this study is conducted based on the LTPP data collected in North America,the research framework and methodology of this study could be applied to analyze pavement performance data of other regions.Firstly,using weigh-in-motion data collected from Traffic Pooled-fund Study,multilevel models with 2-level structure are developed to model traffic volume and axle effect index(AEI),respectively.As a state-of-the-art methodology for longitudinal data analysis,multilevel model can incorporate multiple sources of variation from different test sections and different vehicle types.AEI is proposed as a measure of total damage caused by axle loads of a specific axle group(single,tandem,triple,or quad-plus)in a given period of time with regard to axle weight.Compared to truck volume,AEI can characterize the effect of traffic loads on pavement more accurately.Secondly,using pavement performance data collected from SPS-5 experiment,a multilevel models with 3-level structure are established to investigate the effect of overlay thickness,overlay material and pre-overlay treatment on posttreatment performance while accounting for the clustered longitudinal data structure of SPS-5 experiment.At each experimental site of SPS-5 experiment,multiple test sections were built side-by-side and periodically monitored.Pavement performance of multiple test sections nested within the same experimental site may tend to be similar as these test sections are subject to the same in-situ conditions such as climate and traffic loads.Three performance indicators,i.e.,international roughness index,rutting depth,and cracking percent,are analyzed in this study.Effect of overlay thickness,overlay material,pre-overlay treatment and possible interactive effect are systematically investigated.Thirdly,using distress data of flexible pavements in the LTPP database,association rule learning is adopted to explore the relationship between the occurrences of various types of pavement distress.Very few research has examined the relationship between the occurrences of different types of distress.Using Apriori algorithm,a number of meaningful co-occurrence patterns among various distress types are identified through examining and visualizing the resulting association rules.Three indicators,i.e.support,confidence,and lift,are combined to examine association rules.These findings can contribute to a better understanding of the relationships among the occurrences of various distress types and further provide clues on the causes of different distress types based on pavement performance monitoring data.Fourthly,a data-driven framework is proposed to analytically develop rating thresholds of pavement performance indicator based on historical deterioration data.With the development and widely use of pavement management system,pavement management agencies have been gradually accumulating a large amount of pavement deterioration data.How to make full use of these data has become a topic worth studying.As International Roughness Index(IRI)is one of the most widely used pavement performance indicators,the proposed framework is demonstrated using IRI and validated using IRI data obtained from the LTPP database.The proposed framework can fully utilizing local historical deterioration data available and have the potential to take into account local characteristics.