Multicoating Damage Detection of Vessels Using IDBO and Machine Learning-Based Electric Field

Accurate localization of coating damage is crucial for timely repairs and protection against related threats. The complexity of corrosion electric field signals increases with the presence of multiple coating damages, making detection more challenging. To address this issue, this study proposes a novel multicoating damage detection method by integrating the improved dung beetle optimizer (IDBO) with a machine learning-based approach. First, IDBO was employed to decompose the multicoating damage electric field signals into a superposition of individual single damage signals using equivalent point charge calculations. Next, refined composite multivariate multiscale fluctuation reverse dispersion entropy (RCMMFRDE) was utilized for feature extraction. These feature vectors were subsequently input into the hierarchical prototype (HP) classifier to detect the damaged regions. Numerical experiments demonstrated that the proposed method detection accuracies of 94.0% or two damages and 88.0% for three damages. Physical scale experiments further validated these findings, with accuracies of 90.1% and 84.3% for two and three damages, respectively. The results confirm that the proposed method reliably detects coating damages, contributing to enhanced vessel defense and surveillance.