Probability of Detection Curves for Global Damage based on the Bridge Response to a 2-axle Truck Fleet

Structural Health Monitoring (SHM) under realistic traffic and noise conditions remains a critical challenge due to the inherent variability in vehicle parameters as well as measurement uncertainties. This paper assesses damage detection reliability using Probability of Detection (PoD) curves, a statistical approach that is widely adopted in non-destructive testing but rarely utilised for traffic-induced bridge responses. Here, prior work applying the PoD curve concept to bridge responses from a single test vehicle with known properties is extended to incorporate traffic fleets comprising 2-axle trucks with statistically randomised properties. A half-car model is employed to simulate the 2-axle truck, whilst the bridge is represented by a 15 m simply supported finite element beam model. The stiffness of the beam is assumed constant across the span and diminishes uniformly to simulate global damage. The traffic fleet is characterised using statistical distributions for gross vehicle weight, axle weights and spacing, speed, lateral position, and suspension properties of the half-car, as informed by literature and weigh-in-motion data. Bridge responses to multiple crossings of randomly sampled 2-axle trucks are generated via Monte Carlo simulations. These simulations embed additive white Gaussian noise and an ISO class 'A' road profile to approximate field conditions. The bridge responses investigated are midspan displacement, midspan strain, and rotation at the support—each potentially measurable by an SHM system. For each crossing, the change between the bridge’s response at baseline and reduced stiffness is calculated; when this difference exceeds a pre-established threshold, the stiffness change is regarded as detectable. This threshold is intrinsically linked to the false alarm rate, and the proportion of responses above it defines the PoD for each level of stiffness loss. Through integrating SHM data, stochastic fleet modelling, and PoD analysis, this paper presents actionable criteria for engineers, based on measured response and noise levels, to improve decision-making in bridge management.