Damage quantification in laminated composites using gaussian process regression model

• Ano de defesa: 2020
• Autor(a) principal: Paixão, Jessé Augusto dos Santos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Damage quantification; Gaussian process; Composite structures; Structural health monitoring; Quantificação de dano; Processos gaussianos; Estruturas de material compósito; Monitoramento de integridade estrutural
• Link de acesso: http://hdl.handle.net/11449/193430

After detecting initial damage in a composite structure through a data-driven approach, the user needs to decide if there is an imminent structural failure or if the system can be kept in operation under monitoring to track the damage progression and its impact on the structural safety condition. In this scenario, it is essential to quantify the extension and the damage's size to decide about these points. Therefore, this work proposes a damage quantification based on the application of the Gaussian Process Regression (GPR) model to capture the trend and uncertainties associated with the damage index progression according to damage extension. The GPR model trained in a supervised approach is used to quantify the damage by the stochastic interpolation of the damage indices. The central methodology is proposed for delamination area quantification in laminated composite plates using damage indices based on Lamb wave signals. Autoregressive models are applied to extract damage-sensitive features from Lamb waves signals, and the Mahalanobis squared distance is used to compute damage indices, although any damage features extraction technique could be used and adapted to the proposed methodology. A modified version of the central methodology is proposed to demonstrate the methodology's versatility, using damage indices based on singular spectrum analysis of vibration signals, a well-established technique in the literature. Three sets of tests are used to demonstrate the effectiveness of this approach — one in carbon-epoxy laminate with simulated damage under temperature changes to show the general steps of the procedure; a second test involving a set of carbon fiber reinforced polymer coupons with actual delamination caused by repeated fatigue loads, for which the central methodology is applied; and finally an industrial example involving a wind turbine blade with damage caused by debonding in the trailing edge and using traditional vibration-based damage indices. Various damage progression levels are measured during the tests and monitored using the sensors bonded to these structural surfaces. The GPR proved to be capable of capturing the trend and accommodating the uncertainties related to the damage indices versus the damage size in the simulated spots in the tests. The results manifest a smooth and adequate prediction of the size area of the simulated and real delamination damage in the two first application cases and the debonding size in the last application case.