A Modular Vision‐Based Robot for Efficient Noninvasive Inspection of External Pipeline Surfaces

This paper presents a prototype of a modular autonomous robot aiming for the rapid inspection of steel pipelines, with a focus on detecting surface cracks. The robot incorporates a novel detachable mechanism, enabling full‐coverage inspection without requiring the robot to rotate completely around the pipe. This design not only simplifies attachment and detachment but also enhances operational flexibility. A 360° rotating scanner, integrated with a vision system featuring on‐device artificial intelligence inference, enables real‐time crack detection while maintaining consistent velocity. An automated crawler system ensures smooth linear movement along the pipeline, keeping a uniform distance from the surface. Additionally, a suspension system enables the wheels to adapt to irregularities of up to 25 mm, ensuring reliable performance on uneven surfaces. Experimental results showcased the robot's capabilities, achieving a 96.67% crack detection accuracy through on‐device inferencing and an enhanced 99.3% accuracy during post‐processing. However, certain limitations were observed, attributed to constraints within the microcontroller. Motion analysis confirmed that the scanner exhibited consistent motion during testing, completing 42 circumferential rotations along a 5‐m pipeline in an average time of 72.45 s, and completed a complete inspection of the pipeline in 50.43 min. While the modular robot was specifically designed and evaluated for straight pipelines with end couplings and was not intended to navigate couplings or bent sections, the system demonstrated stable and accurate performance during laboratory trials. This design enables stable external inspection without requiring the robot to rotate along the pipeline, ensuring rapid and reliable inspection.