Comparative analysis of signal processing techniques for impedance-based SHM applications in noisy environments

• Autor(a) principal: de Castro, Bruno Albuquerque [UNESP]
• Data de Publicação: 2019
• Outros Autores: Baptista, Fabricio Guimarães [UNESP], Ciampa, Francesco
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1016/j.ymssp.2019.02.034; http://hdl.handle.net/11449/188750
• topic: Carbon fibre; Hinkley criterion; Impedance; SHM; Signal processing; Wavelet transform

Structural health monitoring (SHM) systems have been extensively studied in recent decades to determine the health statuses of mechanical, naval and aerospace engineering components. Currently, one of the most promising non-destructive tests for the detection of structural damage is the electromechanical impedance (EMI) technique, which uses low-cost, surface-bonded piezoelectric transducers operating both as sensors and actuators. Although many studies have reported the effectiveness of the EMI method, numerous practical issues, such as signal noise effects caused by environmental conditions, may severely affect the detection and quantification of damage. Therefore, this paper presents a comparative analysis of three signal processing techniques used in the context of the EMI method, which have the potential to enhance the detection of structural damage under environmental signal noise effects. These signal analysis methods include (i) damage indices, such as the correlation coefficient deviation metric, computed directly on impedance signatures; (ii) the wavelet transform computed on transducer response signals in the time domain; and (iii) a novel approach of damage feature extraction in the EMI method based on the Hinkley criterion. Experimental tests were carried out to analyse the three signal processing techniques on a damaged aerospace composite carbon fibre structure subjected to various signal noise levels. The experimental results revealed that conventional damage indices and wavelet transform were significantly affected by noise, whereas the proposed approach based on the Hinkley criterion was more effective for detecting damage in noisy environments.