Optimal sensor placement for uncertainty reduction in diagnostics and prognostics of composite patch repairs

Abstract

This work is focused on the design and theoretical evaluation of a new structural health monitoring scheme for composite patch repair systems that is equipped with diagnostic and prognostic capabilities and considers uncertainty. The system processes strain readings from a limited number of strategically placed sensors and returns information on the crack and debonding detection along with the repair’s remaining useful life. The problem of optimal sensor placement is approached with genetic algorithms that aim to maximize the accuracy and robustness of a trained neural network which is used as the mapping function between strains and diagnostic parameters. The later are considered as regressors in a surrogate model with the response being the stress intensity factor (SIF) of the repaired crack. The SIF range is required in a crack growth predictive model for estimating the remaining useful life of the repaired system. The SHM scheme is demonstrated in a fabricated yet realistic case study. Finite element simulations were performed to explore the problem mechanics, for training the data-driven model and finally for showcasing the online operation of the SHM system.