Scholarlink

<p>Underwater explosions (UNDEX) represent a critical threat to the structural integrity of submerged and floating systems across marine, offshore, and civil engineering sectors. Simulating their complex effects—driven by strong fluid–structure interaction (FSI), cavitation, and nonlinear material behavior—typically relies on numerical methods that, although accurate, are computationally expensive and not ideal for rapid predictions in practical scenarios. To overcome these limitations, the integration of Artificial Intelligence (AI) and Machine Learning (ML) into UNDEX research has recently emerged as a promising, albeit still developing, alternative. This paper presents a systematic review of AI and ML applications for predicting structural and fluid responses under underwater blast loading, focusing on developments over the past two decades. After introducing the governing physics of UNDEX events and summarizing the numerical techniques commonly used for dataset generation, the paper provides a detailed analysis of existing AI and ML approaches for UNDEX prediction. Despite the relative novelty of this research field, the paper concludes by highlighting the main challenges and proposing future research directions, including the adoption of advanced learning paradigms tailored to UNDEX scenarios. Practical guidelines are also offered to support future efforts in developing ML-based frameworks for accurate and efficient UNDEX response prediction.</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

Machine Learning for Underwater Explosion Prediction: A Comprehensive Review of Methods, Challenges and Future Directions

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月21日

查看全文
  • Jacopo Bardiani,
  • Claudio Sbarufatti,
  • Andrea Manes

摘要

Underwater explosions (UNDEX) represent a critical threat to the structural integrity of submerged and floating systems across marine, offshore, and civil engineering sectors. Simulating their complex effects—driven by strong fluid–structure interaction (FSI), cavitation, and nonlinear material behavior—typically relies on numerical methods that, although accurate, are computationally expensive and not ideal for rapid predictions in practical scenarios. To overcome these limitations, the integration of Artificial Intelligence (AI) and Machine Learning (ML) into UNDEX research has recently emerged as a promising, albeit still developing, alternative. This paper presents a systematic review of AI and ML applications for predicting structural and fluid responses under underwater blast loading, focusing on developments over the past two decades. After introducing the governing physics of UNDEX events and summarizing the numerical techniques commonly used for dataset generation, the paper provides a detailed analysis of existing AI and ML approaches for UNDEX prediction. Despite the relative novelty of this research field, the paper concludes by highlighting the main challenges and proposing future research directions, including the adoption of advanced learning paradigms tailored to UNDEX scenarios. Practical guidelines are also offered to support future efforts in developing ML-based frameworks for accurate and efficient UNDEX response prediction.