Online Crack Detection of Highly Curved Cylindrical Coils

Abstract

Many vision-based methods for defect detection have been proposed to replace inefficient manual inspection. However, due to the complex shape and high degree of freedom required to manipulate individual coils, an automatic optical inspection (AOI) method for cylindrical coils remains unsuccessful due to the difficulty of automatically scanning the front, rear, outer, and inner faces of complex spiral-shaped cylindrical coils. This paper presents a turnkey integrated system that can be operated in real time for real coil manufacturers. First, a novel 1-degree-of-freedom rotational progressive system is designed to automatically transport and sort individual spiral coils. Second, a multisensor imaging system is designed to capture images of 360° panoramic views of a cylindrical coil. Then, a robust automatic defect detection algorithm consisting of crack detection and geometric measurement is developed to detect defects in cylindrical coils during manufacturing procedures. The results of a large-scale validation experiment indicate that the proposed AOI system is robust and is practically immune to variations in surface normal vectors, occlusion, and defocus blur, and it has a high sensitivity and specificity of 99.75% and 94.2%, respectively. This validation experiment proved that the software and hardware designed in this study can implement defect detection of spiral coils.