Alloy Cast Product Defect Detection Based on Object Detection

Abstract

In the metal casting process, a lot of fine pores and air holes are formed in the metal cooling process, and the number of air holes determines the quality of products. Responding to high grade development, high value cast inspection is mostly nondestructive, and full inspection is required to guarantee the quality. The X-ray detection becomes the target of industrial manufacturing and inspection. The X-ray detection technique has been extensively used in industrial production for nondestructive accurate detection. The ray image quality and staff qualification are established according to ASTM E1742 and E2973. This study captured defect images, and used X-Ray-CT equipment to create defective samples for defect images. For internal defect detection, the air holes and defects in the alloy metal cast product are located by adopting object detection technique. The defect spot contour in the image is enhanced by image pre-processing, meanwhile the noise or non-concerns are reduced, assisting the deep learning model to look for the features of defect spots. This model can assist the quality inspection personnel to determine the defect type and grade, reduce the inspection errors and shorten the detection time. Moreover, this model can solve the problems of time consuming inconsistent criteria, while making it to be time efficient and accurate judgment.