SCALABLE DRIVE-BY BRIDGE MODAL IDENTIFICATION USING A VEHICLE-INDEPENDENT FRAMEWORK: METHODOLOGY, VALIDATION AND FLEET GENERALIZATION

Abstract

Bridges are critical components of transportation infrastructure, yet the vast majority of the global bridge inventory lacks continuous structural health monitoring due to the cost and logistical burden of permanently installed sensor networks. Drive-by bridge monitoring, in which instrumented vehicles record the bridge dynamic response as they cross, offers a scalable alternative, but the fundamental challenge of separating structural signals from vehicle-induced dynamics has limited its practical deployment. This dissertation develops and validates a scalable, vehicle-independent framework for drive-by bridge modal frequency identification through four interrelated studies. The first study presents a pilot investigation using a mobile robot and a connected vehicle to evaluate whether adjacent road vibration data can isolate bridge modal content through frequency-domain subtraction, identifying up to six modes with a mean error of 2% relative to a portable reference system. Motivated by the limitations of this approach, the second study introduces the Vehicle-Independent Coherence (VIC) filtering framework, a model-free method that constructs a frequency-dependent spectral weight from three pairwise coherence measures to suppress vehicle-road contamination without requiring any knowledge of vehicle dynamics. An uncertainty analysis across 455 run combinations provides the first statistical characterization of drive-by estimate convergence and variability. The third study extends the VIC framework to a heterogeneous four-vehicle fleet spanning electric and internal combustion engine platforms and demonstrates that a single unified parameter set supports fleet-level deployment with decent precision, while a crowd-sensing simulation of over 12,000 mixed-fleet samples achieves a relatively high detection precision. The fourth study evaluates the sensitivity of the framework to tire condition, speed, passing lane trajectory, and bridge type, establishing operational robustness and cross-bridge transferability. Collectively, these findings establish the VIC framework as a rigorous foundation for cost-effective, network-scale bridge monitoring using opportunistic traffic data.