Bayesian Damage Estimation with Regularized Data-Driven Stochastic Time Series Model

Abstract

A probabilistic vibration-based global SHM technique is proposed. In the process, experimental data from a modal test on a wing structure is used to identify a unified model with i) a Vector-dependent Functionally Pooled (VFP) component, ii) and an Auto-Regressive eXogenous (ARX) component. LASSO regularization is incorporated as a model structure selection method while introducing model sparsity. A probabilistic damage identification/quantification method within a Bayesian architecture is applied to solve the inverse problem, which provides a decision confidence interval for damage estimation.