This work explores the application of Random Coefficient (RC) Linear Parameter Varying (LPV) AR models for output-only vibration-based health monitoring of structures with time-dependent dynamics and subject to uncertain operational and environmental conditions. A Bayesian framework is then employed for the model identification and damage inference. The proposed methodology is assessed on the SHM of an operating wind turbine at healthy and five different damage scenarios, via Monte Carlo vibration response simulations computed by the FAST simulation package. The obtained results demonstrate the method’s effectiveness, which allows for robust SHM and accurate modeling of the non-stationary dynamics of the vibration response.
doi: 10.12783/SHM2015/101