水下薄壳结构低频声辐射快速预报方法研究  

Research on rapid prediction for low-frequency acoustic radiation of thin-shell structures in fluids

作　　者：孙滨 李鹏[1] 陈浩然[1] 张德春 SUN Bin;LI Peng;CHEN Haoran;ZHANG Dechun(School of Mechanics and Aerospace Engineering,Southwest Jiaotong University,Chengdu 611756,China)

机构地区：[1]西南交通大学力学与航空航天学院,四川成都611756

出　　处：《舰船科学技术》2024年第24期35-39,共5页Ship Science and Technology

基　　金：国家自然科学基金资助项目(12072298)。

摘　　要：针对含内部结构的水下薄壳振动声辐射问题提出一种快速预报方法。该方法采用贝叶斯超参数优化方法构建神经网络模型,建立薄壳结构表面振动响应与声压的映射关系。通过与理论模拟和实验结果对比验证了该方法的可靠性。与实验结果对比表明,该方法在1~100 Hz范围内,峰值处的平均绝对误差为6.31 dB;与有限元法计算结果相比,总声压级的拟合优度为97.79%,平均绝对误差为0.85 dB。该方法具有较高的计算效率,可为水下结构振动声辐射快速预报提供新思路。A fast prediction method for acoustic radiation of underwater thin shell with internal structure is proposed.The neural network model is constructed with Bayesian hyperparameter optimization method,and the mapping relationship between surface vibration response and sound pressure is established.The reliability of the proposed method is verified by comparison with theoretical simulation and experimental results.The comparison with experimental results shows that the mean absolute error at the peak value is 6.31 dB in the range of 1~100 Hz.Compared with the finite element method,the goodness of fit of the total sound pressure level is 97.79%and the average absolute error is 0.85 dB.This method has high computational efficiency and provides a new idea for rapid prediction of acoustic radiation of underwater structure vibration.

关 键 词：水下薄壳结构 神经网络 贝叶斯优化 快速预报 

分 类 号：U663.1[交通运输工程—船舶及航道工程]