Stochastic validation of structural FE-models based on hierarchical cluster analysis and advanced Monte Carlo simulation

Model validation of uncertain structures is a challenging research focus because of uncertainties involved in modeling, manufacturing processes, and measurement systems. A stochastic method employing Monte Carlo simulation (MCS) and hierarchical cluster analysis (HCA) is presented to give an accurate validation outcome with acceptable calculation cost. Parameters exhibiting the significant effect on modal features are identified by Analysis of Variance. To reduce the calculation burden during direct MCS, Radial Basis Function is employed to generate a low-order model of the response space. Particular emphasis is placed on HCA and model assessment, which are applied to distinguish the global best solution from local best solutions in the complete parameter space. The procedure integrating parameter selection, uncertainty propagation, uncertainty quantification, parameter calibration, and model assessment is suitable for models with massive degrees-of-freedom and complex input–output relationship. FE-models of a satellite are given to illustrate the approach's application on complicated engineering structures.

► The proposed method is aimed at structural FE-model stochastic validation. ► HCA can cope with complex nonlinear input–output relationship of the structure. ► Analysis of Variance can deal with models with a large number of parameters. ► Radial basis function meta-modeling enables an acceptable calculation cost.