A Method of Automatic Defect Recognition for Phased Array Ultrasonic Inspection of Polythene Electro-Fusion Joints

Abstract

Defect classification is the basis of defect safety assessment because defects of different types can lead to failure in different forms. However, the identification of defect type has long been a critical issue in ultrasonic inspection. Wave acoustic was applied in this study to investigate the sound scattering of metal wires in polyethylene (PE), which provided theoretical support for ultrasonic feature extraction. A method of defect recognition for PE electro-fusion (EF) joints was proposed based on pattern recognition of ultrasonic inspection images. According to location, shape, signal intensity, and cluster conditions, typical defects of EF joints of PE pipes were distinguished and identified in phased array ultrasonic images. Furthermore, an automatic defect recognition software was designed based on the proposed approach; the software was improved and verified through defect inspection and identification experiments. Results showed that accuracy can reach 80% for joints with complex defects and 100% for those with single defects.