Bayesian probabilistic damage characterization based on a perturbation model using responses at vibration nodes

摘要

One important topic for structural health monitoring methods is to find a balance in the number of sensors, the accuracy in damage detection and the requirement of a high-fidelity finite element model. Traditional deterministic damage detection techniques with a limited number of sensors can only provide coarse point estimates, and they can be vulnerable to uncertainties such as noise and changing environments. This paper adopts the Bayesian probabilistic approach combined with a perturbation model for probabilistic damage characterization utilizing dynamic responses at a few vibration nodes. Vibrational amplitudes at nodal points, also referred as node displacement or NODIS, can be considered to be an efficient structural damage indicator because they have the potential to achieve real-time damage assessment with a relatively small number of sensors. In the Bayesian framework, an efficient perturbation-based surrogate model is adopted to replace the FE model which is computationally costly. This paper proposes a vibration-based SHM method that is suitable for real-time monitoring, requires a small number of industrial sensors, does not rely on a high-fidelity FE model and can be applied for damage assessment of location and severity. The performance of the NODIS-based Bayesian framework with the perturbation method is evaluated and compared with FE results. The proposed method is applied to a supporting structure of a sailplane under different environmental temperatures.