Bayesian Structural State Tracking from Sparse Sensor Data Using Sequential Gaussian-Process Updating in Experimental Bridge Load Tests

Sparse sensing remains a major limitation in bridge structural health monitoring, particularly when load-test data are interpreted without explicit uncertainty quantification. This paper proposes a sequential Gaussian-process Bayesian updating framework for tracking the evolving structural state of a prestressed concrete bridge from sparse measurements collected during a full-scale experimental load test. The measured response includes force, strain, temperature, local displacement, and laser-based deflections recorded over progressive displacement-controlled loading stages ranging from 5 mm to 60 mm. For each stage, a probabilistic deflection field is reconstructed from the available sensor data, while posterior information from previous stages is propagated to subsequent stages to represent the continuity of structural response under increasing demand. From the inferred posterior response fields, uncertainty-aware structural health monitoring indicators are derived, including compliance evolution, curvature concentration, and response-drift measures between loading stages. A sensor sensitivity analysis is further conducted to quantify the contribution of individual measurement locations to prediction accuracy and posterior uncertainty. Model performance is evaluated through reconstruction accuracy, posterior consistency, and credible-interval calibration. The results show the potential of sequential Bayesian inference to convert sparse experimental load-test data into an interpretable probabilistic description of structural behavior, supporting uncertainty-aware assessment, structural model updating, and future digital-twin integration.