Development of a Smart Autonomous Bilge Management System Using Synthetic Data and Machine Learning Algorithms

Abstract

Artificial Intelligence (AI) and Machine Learning (ML) are revolutionizing industrial operations by enhancing efficiency, ensuring regulatory compliance, and promoting sustainability. This paper presents the development of a Smart Autonomous Bilge Management System (SABIMS) for ships, focusing on ML integration through its AI-Driven Decision-Making Module (AIDDM) in conjunction with the highly automated SABIMS Logic Operations Module (SLOM). A synthetic dataset, carefully designed to reflect realistic maritime operational conditions, was utilized to train and evaluate Multi-Class Logistic Regression (MCLR) and Decision Tree (DT) models. The results demonstrated that DT outperformed MCLR across key performance metrics, including precision, recall, F1 score, accuracy, and fit, while maintaining good class balance. By addressing the challenges of bilge water management through predictive and autonomous decision-making, this paper outlines a practical roadmap for achieving enhanced MARPOL compliance. The findings highlight the critical role of synthetic data and robust ML models in advancing sustainable and efficient shipboard systems.