Autoregressive Models for Damage Prognosis in Smart Structures
Wagner Francisco Rezende Cano1; Paulo Henrique de Oliveira Lopes1; Samuel da Silva2
1 Fundação Centro de Pesquisa e Desenvolvimento em Telecomunicações - CPqD; 2 Universidade Estadual Paulista - UNESP, Câmpus de Ilha Solteira
doi:10.20906/CPS/CON-2016-0209
Resumo
Damage prognosis is the estimation of the remaining useful life of a structure leading to economic and life-safety benefits. This estimation is based on the quantification of damage present in a structure through the development of predictive models to track structural integrity. Usually, after a certain level of damage is detected, its evolution is evaluated in terms of failure mechanics and a damage law. However, the novelty in the present work is based on the parametrization of the structure's dynamic behavior under different structural conditions using autoregressive models. Coefficient extrapolation is employed to enable the creation of a predictive autoregressive model capable to forecast structural behavior for a more severe damaged condition. The advantage is to enable damage prognosis without the need of intricate physics or mathematical based models. This methodology was tested with signals measured on an aluminum plate coupled with piezoelectric patches subjected to subsequent levels of localized mass loss resulting in a successful prediction of the last damage scenario. The results have shown the ability to detect and to predict correctly the structural state of the system.
Palavras-chave: Damage Prognosis; Structural Health Monitoring; Time Series; Predictive Models