A model-based approach for in-situ automatic defect detection in welds using ultrasonic phased array

The accuracy of diagnosis performed by human operators is closely related to different factors such as fatigue and subjectivity. The solution can be in developing intelligent systems that aid humans in decision-making. In that sense, the present paper proposes a model-based methodology for automatic analysis of ultrasonic data to detect, locate, and size defects in welds. For this purpose, a novel reference image construction technique is developed basing on Shannon entropy. Using this reference image, an original image metric is proposed. This metric also enables to determine the length of the detected defects. Three case studies are used to validate the proposed model where defects are first simulated numerically, and then artificial and natural. To assess the relevance of the proposed model, results are compared with those obtained by experts. The comparison is satisfactory, which offers the possibility of transposing the proposed approach of diagnosis to other techniques in which a sequence of images can be obtained.