Experimental Validation of the Population-Form to Represent Nominally-Identical Systems

Abstract

As a framework for population-based Structural Health Monitoring (SHM), a Gaussian processes (GP) is suggested to represent a population of systems. Specifically, a Gaussian processes is used to model the population-form, trained using data collected from a simulated population of nominally-identical structures. The form is then used for damage detection through comparisons to future measurements. Significantly, the GP regression, used to represent the form, generalises to experimental data, which were not (directly) used to train the GP.