Flood-borne debris imposes direct loads on bridges and can intensify local hydraulics, increasing scour risk at foundations. This paper outlines a proposed event-driven structural health monitoring framework that integrates pier-level sensing with artificial intelligence-based debris detection to support flood-responsive bridge decision-making. The concept combines 3D tilt/vibration sensing on the substructure with a deck reference, water-level measurements, surveillance imagery, and basic weather data. Under normal conditions, periodic baseline measurements are planned. When hydraulic or structural triggers occur, such as rapid water level rise, debris obstruction, or unusual dynamics of the structure, the system is intended to switch to continuous logging and issue alerts. The computer vision module is developed using a synthetic-data training strategy and a deterministic decision layer that fuses debris coverage, temporal persistence, and pier proximity into configurable risk bands. Pilot tests of the vision module indicate reliable debris event flagging with low nuisance alarms, supporting early-warning use.
Event-driven bridge SHM concept with AI-based flood debris detection / Buka-Vaivade, K., Nicoletti, V., Gara, F.. - (2026), pp. 1678-1685. [10.1201/9781003778677-202]
Event-driven bridge SHM concept with AI-based flood debris detection
Nicoletti V.;Gara F.
2026-01-01
Abstract
Flood-borne debris imposes direct loads on bridges and can intensify local hydraulics, increasing scour risk at foundations. This paper outlines a proposed event-driven structural health monitoring framework that integrates pier-level sensing with artificial intelligence-based debris detection to support flood-responsive bridge decision-making. The concept combines 3D tilt/vibration sensing on the substructure with a deck reference, water-level measurements, surveillance imagery, and basic weather data. Under normal conditions, periodic baseline measurements are planned. When hydraulic or structural triggers occur, such as rapid water level rise, debris obstruction, or unusual dynamics of the structure, the system is intended to switch to continuous logging and issue alerts. The computer vision module is developed using a synthetic-data training strategy and a deterministic decision layer that fuses debris coverage, temporal persistence, and pier proximity into configurable risk bands. Pilot tests of the vision module indicate reliable debris event flagging with low nuisance alarms, supporting early-warning use.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


