State-of-the-art machine learning applications for ship performance modeling: a comprehensive review from design and operation to maintenance and retrofit

Accurate ship performance modeling, which characterizes the relationships among ship speed, engine power, fuel consumption, and emissions, under varying operational and environmental conditions. It is essential for analyzing and optimizing ship energy efficiency, and it plays a crucial role in supporting shipping decarbonization targets and ensuring compliance with International Maritime Organization (IMO) regulations. Most existing reviews focus mainly on the operational stage, while no comprehensive study has yet covered the entire ship lifecycle. However, data availability, modeling objectives, and method selection vary significantly across different stages, including design, operation, maintenance, and retrofit. This paper provides an overview of recent studies to summarize the current status, development trends, and progress of machine learning applications in ship performance modeling across various stages of the ship lifecycle. A structured review framework is proposed, categorizing the literature according to different lifecycle stages, design, operation, maintenance, and retrofit, and highlighting representative studies and methods. The review also clarifies commonly used terminologies and model classifications, and compares their principles, data requirements, and applicability. Finally, recent advances in machine learning techniques are discussed in relation to their applications and challenges at each stage, followed by insights and recommendations for future research and development.