Investigations on classifiers for cathodically protected reinforced concrete structures

Purpose This investigation seeks to establish an effective machine learning-based classification model to discover offshore corrosion protection methods. Sacrificial anodes (Mg, AZ91D, Al, Zn) and supplementary cementitious materials (SCM) like sugarcane bagasse ash are used for corrosion management, rendering it important to verify which method was used after installation. This study predicts protective methods using structural characteristics and classification techniques. Design/methodology/approach Thirty-three classification models were trained on structural slabs with different anode materials and SCM proportions. Decision trees, support vector machines (SVMs), neural networks and ensemble models were assessed for accuracy, cost, training time and model complexity. Performance was evaluated using confusion matrices, ROC curves and parallel plots. Bagged trees ensemble had the highest classification accuracy (91.36%), followed by medium neural network (84.13%) and fine tree (83.69%). Findings Owing to their tolerance to complex data, ensemble models, notably bagged trees, outperform alternative classifiers. Over 84% accuracy was shown by neural networks in prediction. Quadratic SVM was 82.63% accurate. Conversely, traditional K-nearest neighbor and Naïve Bayes models have much poorer classification accuracy. This study shows that machine learning can detect anode kinds and SCM utilization in offshore constructions. Practical implications The suggested classification scheme helps real-time corrosion monitoring by quickly identifying protective techniques. This helps offshore maintenance teams optimize inspection and replacement schedules, saving downtime and costs. Social implications This study makes offshore buildings more durable, minimizing corrosion-induced failures and marine infrastructure concerns. Artificial intelligence-driven corrosion management classification improves material sustainability and critical infrastructure lifespan. Originality/value This study uses machine learning to identify offshore corrosion protection techniques post-installation in an automated, noninvasive manner. This study is limited by data set size, which can be increased for better generalization. In addition, real-world testing in various environments will improve model reliability.