Abstract of Thesis presented at COPPE/UFRJ as a partial fulfillment of the requirements for the degree of Master of Science (M.Sc.)

Monitoring of Defects of Rigid Pipes by Acoustic Emission and Neural Networks

Carlos Fernando Carlim Pinto

March/2011

Advisors:

Luiz Pereira Calôba Romeu Ricardo da Silva

Department:

Eletrical Engineering

      Among the non-destructive testing, there is the method of inspection with the acoustic emission technique, which is based on the detection of sources of acoustic signals that are emitted during the propagation of discontinuities and sharp plastic deformations. The present work aims to develop non-linear classifiers, taking as input the parameters of the signs of Acoustic Emission (AE) capable of discriminating the growth defects of fracture in rigid duct into three classes of signs: No Propagation, Stable Propagation and Unstable Propagation. Discrimination between classes was made by classifiers nonlinear patterns using artificial neural networks trained by back propagation algorithm feed forwards. The results showed classification accuracy of 86% to situation of three classes of signals, proving that there was a significant evolution in the studies with the aim of separating the time of stable propagation of the unstable. There were also relevant study parameters of AE signals.