Scholarlink

<p>Artificial Neural Networks (ANNs) are increasingly applied in marine design to optimize ship performance, navigation, hydrodynamics, structural integrity, and energy efficiency. This study presents a Systematic Literature Review (SLR) of 36 empirical studies (2019–2024) sourced from ScienceDirect and Scopus, identified through VOSviewer co-occurrence analysis and screened using the PRISMA method. The review examines (1) dominant application segments, (2) complementary methods used with ANN, and (3) targeted design parameters and performance metrics. Results show energy efficiency; notably fuel consumption and speed prediction, and navigation; including route planning and maneuvering being the most common domains. Hybrid approaches (e.g., ANN + Genetic Algorithms, Fuzzy Inference Systems, Finite Element Methods) consistently improve prediction accuracy and model robustness, enabling faster conceptual design and more reliable operational decision-making. ANN models have achieved sub-industry-standard error margins when trained on high-quality datasets, but face persistent issues with data dependency, interpretability, and generalization across varied conditions. Future directions include Physics-Informed Neural Networks (PINNs), lightweight onboard architectures for real-time use, and explainable AI frameworks to enhance transparency and regulatory acceptance. The findings provide actionable insights for advancing sustainable, efficient, and adaptive ANN-based solutions in marine engineering.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Harnessing artificial neural networks for optimizing marine design: a comprehensive review of applications and innovations

  • 期刊论文

  • 发表日期：

    2025年9月10日

查看全文
  • Afandi Abdul Ghani,
  • Md Redzuan Zoolfakar,
  • Noor Apandi Osnin

摘要

Artificial Neural Networks (ANNs) are increasingly applied in marine design to optimize ship performance, navigation, hydrodynamics, structural integrity, and energy efficiency. This study presents a Systematic Literature Review (SLR) of 36 empirical studies (2019–2024) sourced from ScienceDirect and Scopus, identified through VOSviewer co-occurrence analysis and screened using the PRISMA method. The review examines (1) dominant application segments, (2) complementary methods used with ANN, and (3) targeted design parameters and performance metrics. Results show energy efficiency; notably fuel consumption and speed prediction, and navigation; including route planning and maneuvering being the most common domains. Hybrid approaches (e.g., ANN + Genetic Algorithms, Fuzzy Inference Systems, Finite Element Methods) consistently improve prediction accuracy and model robustness, enabling faster conceptual design and more reliable operational decision-making. ANN models have achieved sub-industry-standard error margins when trained on high-quality datasets, but face persistent issues with data dependency, interpretability, and generalization across varied conditions. Future directions include Physics-Informed Neural Networks (PINNs), lightweight onboard architectures for real-time use, and explainable AI frameworks to enhance transparency and regulatory acceptance. The findings provide actionable insights for advancing sustainable, efficient, and adaptive ANN-based solutions in marine engineering.