Use of artificial neural networks in detection of propagation of defects in pipelines rigid

The interest in monitoring equipment in real time is increasing in nowadays, mainly aiming the greater security of its operations. The Acoustic Emission (AE) testing has been the subject of developments with the aim of application in various types of equipment, especially in the inspection of rigid and flexible pipes. This paper presents the methodologies and results of a study of applying the method of Acoustic Emission to detect propagation in defects in pressurized rigid pipes, being a pioneering work in this area of research. In this way, specimens were manufactured with defects artificially inserted. These specimens were submitted to hydrostatic testing and the defect propagation was monitor by AE. The ultrasound by Time of Flight Diffraction (TOFD) was the technique chosen to monitor the defect growth. The AE resulting signals were divided into the classes No Propagation (SP), Stable Propagation (PE) and Unstable Propagation (PI) and used as inputs set in the implementation of nonlinear classifiers by error back propagation. The correct classification results reached close to 86%, proving the efficiency of the method for the conditions tested in this job.