Reliability of non-destructive test techniques in the inspection of pipelines used in the oil industry

The aim of this work is to evaluate the reliability of non-destructive test (NDT) techniques for the inspection of pipeline welds employed in the petroleum industry. Radiography, manual and automatic ultrasonic techniques using pulse-echo and time of flight diffraction (TOFD) were employed. Three classes of defects were analyzed: lack of penetration (LP), lack of fusion (LF) and undercut (UC). The tests were carried out on specimen made from pipelines containing defects, which had been artificially inserted on laying the weld bead. The results showed the superiority of the automatic ultrasonic tests for defect detection compared with the manual ultrasonic and radiographic tests. Additionally, artificial neural networks (ANN) were used in the detection and automatic classification of the defects.