Robotic Assistive Smart Touch Inspection of Offshore Oil and Gas Pipelines

Offshore oil and gas operations rely heavily on bolted flange connections, where bolt loosening poses ongoing safety and reliability risks. This project developed and demonstrated a fully integrated robotic smart‑touch inspection system capable of autonomously locating, grasping, and evaluating subsea flanges using a Remotely Operated Vehicle (ROV). The system combines long‑range sonar perception, vision‑based manipulation, advanced autonomous control, and vibration‑based bolt‑tightness classification. A redesigned gripper incorporating paired piezoceramic (PZT) transducers significantly improved stress‑wave signal quality, enabling reliable active sensing once the flange was grasped. Long‑range detection was achieved using a Ping360 scanning sonar with RANSAC‑based pipeline extraction and IMU-assisted localization, followed by a transition to close‑range machine‑vision using YOLOv8 and Hough‑Transform methods for flange alignment. AprilTag-based station‑keeping ensured stable sonar mapping in drift‑prone underwater environments. Machine-learning models—including MFCC+SVM, physics‑informed neural networks, and domain‑adaptation techniques such as FRF-based signal transformation and DANN+CADA—achieved high bolt‑tightness classification accuracy across flange sizes from 4.25 to 21 inches, with performance exceeding 90% for most configurations. The system successfully demonstrated end‑to‑end autonomous operation, reducing the need for diver intervention and establishing a practical pathway toward reliable, repeatable, and scalable robotic subsea flange inspection.