A novel integrated interval prediction method based on small and non-normality distributed datasets for pavement performance from the uncertainty perspective

Abstract

The uncertainty prediction of pavement performance can promote intelligent highway tunnel operation and maintenance, but it encounters the challenges of small and non-normality distributed datasets. This paper proposes a novel integrated interval prediction method to overcome these shortcomings so that we can decrease the uncertainty. This paper also validates the effectiveness of the proposed method using the empirical test of pavement performance data from the Dalian Road Tunnel in Shanghai, China. The evaluated results of PINAW and PICP achieve the values of 0.2262 and 89.24%, respectively, demonstrating excellent uncertainty prediction. Furthermore, this paper applied the proposed method to other datasets, which exhibit good generalization ability and thereby be beneficial for formulating scientific maintenance decisions and achieving the maximum service benefits of road surfaces. At last, our code will be open-sourced.