Smart Maintenance Management of Highway Bridges Subject to Inspections

Bridge maintenance management aims to balance life-cycle costs and bridge safety based on inspection data and by predicting the time-dependent bridge deterioration using physics-based models or Markov-based models. Through four independent projects, this dissertation develops maintenance optimization frameworks based on the two approaches, contributing to intelligent infrastructure asset management and sustainable societies.Existing research focused on physics-based models has not explored the use of non-destructive inspection data in predicting the deterioration of pre-stressed concrete highway bridges. Based on a real bridge example, we predict the loss of structural capacity resulting from chloride-induced corrosion, simplified by gamma processes. To make the model more realistic, we consider random vehicle arrivals and varying loads, ensuring failure probabilities are not underestimated. Bayesian inference updates are used to select model parameters and determine the associated temporal structural resistances. Current bridge management systems (BMS) adopt Markov chains to simulate the temporal bridge condition rating drops, which ignore the time effect on bridge deterioration rates and underestimate the failure risks of aging bridges. Current fixed-interval inspection policies may also over-inspect new bridges and under-inspect old bridges. Furthermore, existing BMS fail to consider the effects and costs associated with carbon emissions due to maintenance activities. We propose multi-dimensional Markov models to describe temporal bridge deteriorations, the states of which are translated by the Weibull-baseline proportional hazard model into a single hazard indicator for decision-making. Markov decision processes (MDP) are employed to determine optimal repair schemes with the objective to minimize the expected long-term annual maintenance-associated and detouring-related costs. As an alternative to MDP, we also explore the use of Q-learning for single-bridge maintenance optimization. The decision-making procedure implemented in this thesis resembles a quality control chart with time-variant decision thresholds. Integer programming is used to optimize budget allocation processes for bridge networks. For demonstration purposes, we apply the proposed methodologies to bridges in New York state. Comparative evaluations indicate that the proposed methods outperform existing BMS in terms of information richness and total cost savings. Sensitivity analyses reveal the effects of inspection intervals and budget levels on maintenance policies and network carbon emissions, respectively.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.