Scholarlink

<p>Due to the increasing complexity of turbulent flows, researchers have increasingly turned to artificial intelligence (AI) to enhance turbulence modeling, prediction, and control. This review presents a structured and critical synthesis of over 200 recent studies on the integration of AI techniques into turbulence research. We analyze a broad range of AI architectures—including multilayer perceptrons, convolutional and recurrent neural networks, generative adversarial networks, physics-informed neural networks, graph neural networks, and transformer-based models—and assess their suitability for different turbulence-related tasks such as modeling, flow prediction, control, and physical insight discovery. Key challenges are identified, including limited interpretability, uncertainty quantification, generalization issues, and computational cost. In response, we summarize emerging solutions such as observable-augmented manifold learning, physically constrained neural networks, and active learning frameworks. The review also explores future directions, including foundation models, cross-modal data fusion, and the integration of simulations and experiments under an AI-guided paradigm. This work aims to support researchers and practitioners in navigating the evolving landscape of AI-enabled turbulence research and in identifying key gaps and opportunities for interdisciplinary advancement.</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

Breakthroughs and Perspectives of Artificial Intelligence in Turbulence Research: From Data Parsing to Physical Insights

  • 期刊论文

  • 发表日期：

    2025年9月6日

查看全文
  • Junjie Fang,
  • Xujiang Qin,
  • Yanqiu Zuo,
  • Hongqiang Wang

摘要

Due to the increasing complexity of turbulent flows, researchers have increasingly turned to artificial intelligence (AI) to enhance turbulence modeling, prediction, and control. This review presents a structured and critical synthesis of over 200 recent studies on the integration of AI techniques into turbulence research. We analyze a broad range of AI architectures—including multilayer perceptrons, convolutional and recurrent neural networks, generative adversarial networks, physics-informed neural networks, graph neural networks, and transformer-based models—and assess their suitability for different turbulence-related tasks such as modeling, flow prediction, control, and physical insight discovery. Key challenges are identified, including limited interpretability, uncertainty quantification, generalization issues, and computational cost. In response, we summarize emerging solutions such as observable-augmented manifold learning, physically constrained neural networks, and active learning frameworks. The review also explores future directions, including foundation models, cross-modal data fusion, and the integration of simulations and experiments under an AI-guided paradigm. This work aims to support researchers and practitioners in navigating the evolving landscape of AI-enabled turbulence research and in identifying key gaps and opportunities for interdisciplinary advancement.