A Bayesian model updating procedure for dynamic health monitoring

Structures under dynamic excitation can undergo phenomena of crack growth and fracture. For safety reasons, it is of key importance to be able to detect and classify these cracks before the unwarned structural failure. Additionally, the cracks will also change the dynamic behaviour of the structures, impacting their performance. Here, a Bayesian model updating procedure has been implemented for the crack detection location and length estimation on a numerical model of a spring suspension arm. A highfidelity finite element model has been used to simulate experimental data, by inserting cracks of different extent at different locations and obtaining reference frequency response functions. In the following, a low fidelity parametric model has been used in the Bayesian framework to infer the crack location and length by comparing the dynamic responses. It is shown that the proposed methodology can be successfully adopted as a structural health monitoring tool.