Assessment and prediction of high speed railway bridge long-term deformation based on track geometry inspection big data

Abstract

This paper proposes a low-cost, data-driven approach to assess and predict bridge deformation using track inspection big data, which is primarily used for assessing track conditions. Firstly, a Bridge Deformation Assessment model with a sophisticated signal processing process is introduced to manipulate track geometry inspection data for extracting bridge-related components. Secondly, a Bridge Dynamical Deformation index (BDD index) is defined to quantify bridge deformation based on track geometry inspection data. Thirdly, the Temperature-Time-Deformation model (TTD model) is established to describe bridge deformation with respect to ambient temperature and length of service time of the bridge. Three types of TTD equations are proposed, including exponential-, hyperbolic- and linear-TTD equations. Fourthly, a track geometry inspection dataset over 2.6 years involving 563 bridge spans is applied as a case study. It is found that the BDD index changes with ambient temperature by 0.02 mm/°C on average, and increases with time by 0.2 mm/year during the 2.6-year period. Furthermore, a prediction on the amount of increase of the BDD index over the following 3 years is given with a 95% confidence level. It is expected that BDD index will increase by 0.5 mm in 2 years and 0.7 mm in 3 years according to the TTD model. Finally, the model uncertainty is discussed from data aspect and model aspect. The methods in this paper are of reference value for research topics on bridge condition evolution, rail geometry degradation and prediction-based infrastructure maintenance.