HULL-CLIMBING ROBOT FOR CORROSION INSPECTION

Abstract

Corrosion remains a major threat to maritime structures, motivating safer and faster inspections. We present an integrated system for ship-hull corrosion assessment that combines an autonomous ferromagnetic climbing robot with ultrasonic thickness (UT) sensors and a machine-learning classifier. The robot uses magnetic adhesion to achieve stable traversal of painted steel hulls and acquires UT data without removing protective coatings, enabling fully non-destructive testing (NDT). UT signals are processed to estimate local thickness and derive features for corrosion inference. A gradient-boosted classifier model (HistGradientBoosting) trained on synthetic UT datasets achieves a ROC AUC of 0.97; field trials on operational vessels are planned to validate performance and measure corrosion level. We also analyzed relationships among thickness, time in service, and coating condition to contextualize risk. By consolidating robotic mobility, UT-based sensing, and data-driven inference into a single workflow, the system provides a practical foundation for predictive maintenance of maritime assets while reducing exposure to hazardous manual inspections.