Abstract
A framework for quantifying the uncertainty propagating through the signal energy-based acoustic source localization approach in an orthotropic plate under an uncertainty in the properties of the plate material is presented. Seven mechanical properties of an orthotropic plate material, namely, density and six elastic constants, are considered as lognormally distributed and mutually independent random variables (RVs) with a fixed coefficient of variation for all seven random variables. Their means are considered such that the “mean” plate exhibits a strong anisotropy. Using Latin hypercube sampling, several design points in lognormal spaces of these random variables are selected. For each design point, an acoustic event is simulated in the corresponding plate using finite element analyses. The signal energy-based approach is applied to localize the acoustic source for each design point. The localization error for each design point is taken as the “response,” and a regression kriging metamodel is constructed through these response values at the design points. Monte Carlo (MC) points are selected in lognormal spaces of the random variables, and the response values at these Monte Carlo points are estimated using the regression kriging metamodel. The distribution parameters of the so-obtained response values are computed. Finally, a global sensitivity analysis of the random variables is carried out by computing the Sobol’ indices.
