Measurement-informed calibration of FEM models for strain-based structural assessment using multi-sensor data

Structural Health Monitoring (SHM) systems rely on advanced measurement technologies and multi-sensor data as the primary source of information about structural response. Interpretation of these measurements is enabled by numerical models, which due to inherent simplifications and associated uncertainties may deviate from the actual structural behavior. Therefore, these models require appropriate calibration procedure. This study proposes a measurement-informed calibration framework based on a multi-sensor system, which was implemented on a network arch bridge. The measurement locations were determined through prior structural analysis, incorporating structural specificity of the bridge. Strain measurements were acquired at 100 Hz using fiber-opting Bragg grating sensors, the specifications of which are described in detail. To account for environmental influences, temperature sensors were used for temperature compensation. The measurement process was characterized by an uncertainty budget, including sensors, loads and environmental conditions, resulting in a relative expanded uncertainty of 4.7%. Within this framework, the finite element model is interpreted as a parametric measurement model, in which structural parameters are adjusted directly based on measured strain responses. The paper presents a holistic design of the system, including incorporation of measurement uncertainties, static load tests procedure, and computational implementation. It involves a genetic algorithm and a neural network surrogate model to reduce computational cost. Proposed approach enables measurement-informed calibration and supports uncertainty reduction in strain-based structural assessment.