Serviceability assessment for long-span suspension bridge based on deflection measurements

Serviceability issues related to service load deformations is of great importance to in-service bridge maintenance decision making. Development in structural health monitoring brings opportunities to serviceability assessment for long-span bridges. In this study, a reliability assessment framework of traffic-induced vertical deflection is developed by incorporating extreme values analysis and system reliability with deflection measurements. The framework includes the following steps: (a) identifying traffic-induced deflection by eliminating temperature component and extracting local extremes from the original measurements; (b) probabilistically modeling the peaks-over-threshold (POT) data to the generalized Pareto distribution (GPD); (c) selecting proper threshold of the GPD based on the mean residual life plot and goodness-of-fit tests; (d) extending the GPD to generalized extreme value distribution-based extreme value distribution with the assumption of filtered Poisson process; and (e) assessing the serviceability reliability of vertical deflection by using the concept of series system. The multisensor deflection measurements, which were collected by a previously verified system installed on a suspension bridge, are processed and studied according to the proposed framework. The results indicate that the best GDP fitting of the POT data can be obtained due to the proposed threshold selection strategy. The quasisystem reliability approach also shows its ability in addressing the relationship between the reliabilities of the multiple sections equipped with sensors and presenting a reasonable overall reliability for the prototype bridge.