Developing the quality assurance thresholds for screening automated pavement condition data using prediction interval methods

Abstract

ABSTRACT Data quality assurance (QA) procedures have been developed and implemented to evaluate the quality of automated pavement distress data. Most State Departments of Transportation (DOTs) use fixed acceptance thresholds to evaluate data quality at the network level and project level. However, errors of automated measurements can be affected by many factors such as pavement types and pavement conditions. It is necessary to take those factors into account while developing the QA acceptance thresholds. The objective of this research is to propose thresholding methods that could effectively utilise these factors to improve QA of the automated data. In this study, the potential problems associated with the use of fixed acceptance thresholds for QA were first discussed. Then, one conventional prediction interval method and one bootstrapping prediction interval method were proposed for QA procedures. The proposed prediction interval based methods can be used to determine variable thresholds that are sensitive to measurement errors. Lastly, QA results from the proposed methods and a fixed thresholding method were compared and analysed. The experimental results showed that prediction interval methods were superior to the fixed threshold methods in identifying pavement sections with data quality issues and evaluating the overall data quality of the network.