GAUSSIAN HYPERMODEL APPLIED TO FAILURE ANALYSIS OF LAMINATED COMPOSITES

This work deals with the use of a Gaussian Hyper model to recover the interface thermal contact conductance between two materials within the Bayesian framework. Such function is assumed to be piecewise smooth with discontinuities of unknown location and size. In this work, we propose the use of a Gaussian smoothness prior, where the scaling parameter appearing in the prior was treated as an unknown hyperparameter. This parameter was estimated as part of the inference problem in a hierarchical model, with a hyperprior density in the form of a Rayleigh distribution. The feasibility of the approach was evaluated with simulated temperature measurements.