文献类型：期刊文献

英文题名：A hybrid machine learning framework for predicting aircraft scaled sound pressure levels: a comparative study

作者：Li, Guo[1]; Sheng, Hongyu[2]

第一作者：Li, Guo

机构：[1] State Key Laboratory of Electronic Thin Films and Integrated Devices, University of Electronic Science and Technology of China, Chengdu, 610054, China; [2] College of Robotics, Beijing Union University, Beijing, 100101, China

第一机构：State Key Laboratory of Electronic Thin Films and Integrated Devices, University of Electronic Science and Technology of China, Chengdu, 610054, China

通讯机构：[2]College of Robotics, Beijing Union University, Beijing, 100101, China|[1141739]北京联合大学机器人学院;[11417]北京联合大学;

年份：2025

卷号：72

期号：1

外文期刊名：Journal of Engineering and Applied Science

收录：EI(收录号：20253919231702);Scopus(收录号：2-s2.0-105016803812)

语种：英文

外文关键词：Acoustic noise - Acoustic noise measurement - Acoustic wave transmission - Adaptive boosting - Aerodynamics - Aviation - Fighter aircraft - Forecasting - Learning systems - Machine learning - Mean square error - Noise abatement - Noise pollution

摘要：Aircraft noise was among the significant environmental challenges, given increased air traffic affects communities and raises health concerns. Accurate noise prediction is indispensable for designing quieter aircraft and mitigation strategies supporting sustainable aviation practices. In this regard, this study aimed to develop a scalable hybrid machine learning framework to predict aircraft scaled sound pressure levels (SPL). Four models (Extra Tree, AdaBoost, Gradient Boosting, and Histogram-Based Gradient Boosting) were evaluated. The best-performing model is an Extra Tree with an R2 of 0.9542 and the minimum mean squared error (MSE) of 3.12. Optimization algorithms were performed to improve its accuracy and robustness—JAYA, Enhanced AEO, Levy JAYA, and JADE—which decreased MSE up to 20% for ensuring stable convergence within the first 300 epochs. JAYA and Enhanced AEO had the best results, balancing accuracy and computation efficiency. Optimized models increased the runtime by 20–30% and memory usage by 15%, which makes them fit for offline applications. Under the computational trade-off conditions, the hybrid models revealed a high potential for further improving the accuracy of the noise prediction. The proposed machine learning framework has really given actionable insight into optimizing aircraft design for noise reduction with a minimum loss in their aerodynamic efficiencies. Prediction of the sound pressure level of key parameters, such as the angle of attack, Reynolds number, and surface roughness, contributes to developing effective noise mitigation strategies, including regulatory compliance with sustainability of aviation. The flexibility extends to a range of aircraft components, with quieter and efficient design seriously threatened by environmental and community noise concerns. While the models demonstrated scalability and high accuracy, further refinements are needed to enhance real-time performance and integrate subjective noise metrics, broadening their applicability to diverse aviation noise management scenarios. ? The Author(s) 2025.