Scholarlink

<p>Airfoil noise plays a significant role in the design of aerodynamic devices, directly influencing both environmental impact and passenger comfort. In this study, noise prediction for the NACA0012 airfoil was optimized using hybrid deep learning techniques, specifically customized versions of Convolutional Neural Networks (CNN), Gated Recurrent Units (GRU), and Long Short-Term Memory (LSTM) networks. Among the tested architectures, the CNN-LSTM model demonstrated the best overall performance, achieving a Kling-Gupta Efficiency (KGE) of 0.983 and a Theil’s U2 statistic of 0.0051. This represents an improvement over CNN-GRU (KGE: 0.9741, U2: 0.0051), GRU (KGE: 0.9733, U2: 0.0061), and LSTM (KGE: 0.9665, U2: 0.0063). The results confirm that the CNN-LSTM approach offers superior stability and accuracy, making it a promising candidate for integration into automated analysis tools. Such integration could enable real-time noise prediction, support dynamic rotor or wing design optimization, and enhance compliance with environmental standards, while also offering potential applications in wind turbine design.</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

A comparative evaluation of multivariate hybrid deep learning methods for airfoil noise forecasting

  • 期刊论文

  • 发表日期：

    2025年11月10日

查看全文
  • Ihsan Uluocak

摘要

Airfoil noise plays a significant role in the design of aerodynamic devices, directly influencing both environmental impact and passenger comfort. In this study, noise prediction for the NACA0012 airfoil was optimized using hybrid deep learning techniques, specifically customized versions of Convolutional Neural Networks (CNN), Gated Recurrent Units (GRU), and Long Short-Term Memory (LSTM) networks. Among the tested architectures, the CNN-LSTM model demonstrated the best overall performance, achieving a Kling-Gupta Efficiency (KGE) of 0.983 and a Theil’s U2 statistic of 0.0051. This represents an improvement over CNN-GRU (KGE: 0.9741, U2: 0.0051), GRU (KGE: 0.9733, U2: 0.0061), and LSTM (KGE: 0.9665, U2: 0.0063). The results confirm that the CNN-LSTM approach offers superior stability and accuracy, making it a promising candidate for integration into automated analysis tools. Such integration could enable real-time noise prediction, support dynamic rotor or wing design optimization, and enhance compliance with environmental standards, while also offering potential applications in wind turbine design.