10 Stochastic Wavenumber Estimation: Damage Detection Through Simulated Guided Lamb Waves 113 Fig. 10.6 Test-analysis correlation of LWP model Inverse problem for damage detection and quantification M(E, , , t, ) = k Prior(t) Experiments • Experimental uncertainty due to spurious peaks in data • Incorporate engineering judgment through informed prior Likelihood(t) Posterior(t) Posterior(ct) Prior(ct) Model • Parametric uncertainty due to unknown material properties • Simultaneously calibrate material properties and quantify damage Bayesian Inference Fig. 10.7 Framework for Stochastic Wavenumber Estimation to mitigate uncertainties in AWS procedure 10.4.3.1 Parametric Uncertainty As discussed previously, the ability to apply a model for inverse calculation of a damage indicator depends on the predictive capability of the model. Without confidence in the numerical model parameters, severity of damage may not be accurately quantified. While calibrating uncertain parameters to a baseline model of a “healthy” state remedies the issue, there is a
RkJQdWJsaXNoZXIy MTMzNzEzMQ==