Robust structural health monitoring under environmental and operational uncertainty with switching state-space autoregressive models

Existing methods for structural health monitoring are limited due to their sensitivity to changes in environmental and operational conditions, which can obscure the indications of damage by introducing nonlinearities and other types of noise into the structural response. In this article, we introduce a novel approach using state-space probability models to infer the conditions underlying each time step, allowing the definition of a damage metric robust to environmental and operational variation. We define algorithms for training and prediction, describe how the algorithm can be applied in both the presence and absence of measurements for external conditions, and demonstrate the method’s performance on data acquired from a laboratory structure that simulates the effects of damage and environmental and operational variation on bridges.