Machine Learning Applied in the Detection of Faults in Pipes by Acoustic Means

• Autor(a) principal: Merizio, Igor Feliciani [UNESP]
• Data de Publicação: 2021
• Outros Autores: Chavarette, Fábio Roberto [UNESP], Moro, Thiago Carreta [UNESP], Outa, Roberto, Mishra, Vishnu Narayan
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1007/s40032-021-00682-y; http://hdl.handle.net/11449/208640
• topic: Artificial immune system; Decision making; Negative selection algorithm; Preventive diagnosis; Structural Health Monitoring

The Structural Health Monitoring evaluates the situation of aeronautical, civil or mechanical structures and provides a forecast of its remaining useful life, acting in decision making, being able to intervene in critical situations. It has emerged as a viable economic alternative for monitoring structures and preventing failures. Thus, this system is defined as a prophylactic measure, reliable and effective against structural failure. This work exposes the theoretical basis and a new technique for detection of failures in pipes by acoustic means, following the International Standard ISO10534-1 (1996) in the sampling. This method of fault detection using acoustic means requires considerably less training data than is usually used in the literature, with approximately 85% less data. The results presented in this work showed how it is possible and effective to detect failure in pipes by acoustic means using an artificial immune system for structural monitoring, with a 100% precision in the detection of failure.