集成算法在船舶辐射噪声预报中的应用研究

强以铭; 陈奕宏; 裴雨晴; 庞业珍; 谢硕

doi:10.11993/j.issn.2096-3920.2025-0111

集成算法在船舶辐射噪声预报中的应用研究

doi: 10.11993/j.issn.2096-3920.2025-0111

强以铭^{1, 3},
陈奕宏^{1, 2,},
裴雨晴^{1, 2},
庞业珍^{1, 2},
谢硕¹

1.
中国船舶科学研究中心, 江苏无锡, 214082
2.
船舶振动噪声重点研究室, 江苏无锡, 214082
3.
上海交通大学, 船舶海洋与建筑工程学院, 上海, 200240

详细信息

作者简介:
强以铭(1994-) , 女, 硕士, 工程师, 主要研究方向为数据科学

通讯作者:
陈奕宏(1980-), 男, 硕士, 研究员, 主要研究方向为水声学.

中图分类号: TJ630;U661
计量
- 文章访问数: 11
- HTML全文浏览量: 5
- PDF下载量: 2
- 被引次数: 0
出版历程
- 收稿日期: 2025-08-22
- 修回日期: 2025-09-12
- 录用日期: 2025-09-15
- 网络出版日期: 2025-11-26

Application of ensemble learning models on ship radiation noise prediction

QIANG Yiming^{1, 3},
CHEN Yihong^{1, 2
,},
PEI Yuqing^{1, 2},
PANG Yezhen^{1, 2},
XIE Shuo¹

1.
China Ship Scientific Research Center, Wuxi 214082, China
2.
Ship Vibration and Noise Key Laboratory, Wuxi 214082, China
3.
School of Ocean and Civil Engineering, Shanghai Jiaotong University, Shanghai 200240, China

摘要

摘要: 船舶多振动源产生的辐射噪声, 严重影响民船舒适性与军舰隐身性能。准确预测其辐射噪声水平与分布, 可为船舶设计阶段的减振降噪优化提供关键支撑。针对船舶振动源数量多以及噪声辐射机制复杂的问题, 文中首先采用集成算法中的随机森林(RF)与梯度提升树(GBDT)算法, 对不同工况、不同测点的1/3倍频程噪声声压级开展快速预报, 并与贝叶斯岭回归(Bayesian Ridge Regression)模型的预报效果进行对比验证。4种测试工况的验证结果显示, 集成算法在3种工况下的预报效果优于贝叶斯岭回归, 平均绝对误差均小于5 dB; 进一步对上述模型进行优化, 通过在不同层次构建集成算法与线性算法的基础单元并组合, 形成辐射噪声联合预报方案, 其精度较单一集成算法提升1.5 dB。文中研究提出的集成算法及联合预报方案, 为船舶辐射噪声的快速精准分析提供了有效技术工具。
- 船舶辐射 /
- 噪声预报 /
- 集成算法 /
- 振动
Abstract: Ship radiation noise induced by vibration is a major concern for the comfort and stealthy of commercial ships and warships. An accurate prediction for the distribution and sound level of ship radiation noise can assist effective ship designs to lower vibration and noise. In order to avoid constructing complex functions from multiple vibration sources, this paper used Random Forest(RF) and Gradient Boosting Decision Tree(GBDT) methods to build surrogate models to quickly predict the 1/3 rd octave noise level of different working conditions and measuring points. Ensemble learning models have a better performance on 3 out of 4 conditions compared to Bayesian Regression, with MAE prediction error less than 5 dB. This paper also proposed an optimized version of above models that combines ensemble learning methods and linear regression, which increased prediction accuracy by 1.5 dB. The proposed ensemble learning methods could be an efficient tool for ship radiation noise analysis.
- ensemble learning /
- ship radiated noise /
- surrogate model /
- vibrate

HTML全文

图 1 千岛湖试验场

Figure 1. Qiandao Lake test site

下载: 全尺寸图片幻灯片

图 2 样本采集试验装置示意图

Figure 2. Schematic diagram of sample collection test device

下载: 全尺寸图片幻灯片

图 3 样本训练与验证分割方式

Figure 3. Sample training and verification segmentation

下载: 全尺寸图片幻灯片

图 4 RF与GBDT模型结构示意图

Figure 4. Structure diagram of RF and GBDT models

下载: 全尺寸图片幻灯片

图 5 不同频率下预报模型噪声等级预报结果

Figure 5. Prediction results of noise level of prediction model at different frequencies

下载: 全尺寸图片幻灯片

图 6 不同测点下模型噪声等级预报结果

Figure 6. Prediction results of model noise level at different measuring points

下载: 全尺寸图片幻灯片

图 7 优化后模型结构示意图

Figure 7. Schematic diagram of optimized model structure

下载: 全尺寸图片幻灯片

图 8 优化后模型对于工况4预报效果

Figure 8. Model performance of the optimized model for condition 4

下载: 全尺寸图片幻灯片

表 1 集成模型超参数列表

Table 1. List of integration model superparameters

模型名称	模型超参数
RF回归	CART数量, 袋外评分, 最大特征数及最大深度等
GBDT回归	CART数量, 学习率, 阿尔法, 子样本比例及最大特征数等

下载: 导出CSV

表 2 不同频率区模型预报误差比对

Table 2. Comparison of model prediction errors in different frequency regions

频域/ Hz	RF	GBDT	Bayesian
10 ~100	2.49	2.21	2.56
100 ~ 1 000	2.83	2.53	2.76
1 000~10 000	2.12	2.31	3.33

下载: 导出CSV

表 3 不同工况下测点一模型预报误差比对

Table 3. Comparison of prediction error between measurement point and model under different working conditions

	MAE/dB				RMSE /dB
	工况1	工况2	工况3	工况4	工况1	工况2	工况3	工况4
RF	3.2	3.0	1.2	4.8	3.2	3.1	1.2	4.9
GBDT	2.9	2.9	1.1	4.3	2.9	3.0	1.1	4.4
Bayesian	4.3	3.2	1.7	3.8	4.4	3.4	1.8	3.9

下载: 导出CSV

参考文献(19)

[1]	HODGES C H, WOODHOUSE J. Theories of noise and vibration transmission in complex structures[J]. Reports on Progress in Physics, 1986, 49(2): 107-170. doi: 10.1088/0034-4885/49/2/001
[2]	庞业珍, 裴雨晴. 基于船上振动噪声监测的水下辐射噪声实时预报研究进展[J]. 隐身技术, 2023(1): 1-10. PANG Y Z, PEI Y Q. Research progress on real-time prediction of underwater radiated noise based on shipboard vibration and noise monitoring[J]. Stealth Technology, 2023(1): 1-10 (in Chinese).
[3]	王学杰, 单衍贺, 秦新华, 等. 舰船水下辐射噪声快速预报方法[J]. 噪声与振动控制, 2018, 38(3): 1-5. WANG X J, SHAN Y H, QIN X H, et al. Rapid prediction method of ship underwater radiated noise[J]. Noise and Vibration Control, 2018, 38(3): 1-5 .
[4]	庞业珍, 俞孟萨. 基于多距离声阵实测回归传播损失的浅水域水下辐射噪声源级测量方法[J]. 船舶力学, 2023, 27(4): 598-606. PANG Y Z, YU M S. Measurement method of underwater radiated noise source level in shallow water based on multi-range hydrophone array measurement and regression propagation loss[J]. Journal of Ship Mechanics, 2023, 27(4): 598-606 .
[5]	CINTOSUN E, GILROY L. Estimating ship underwater radiated noise from onboard vibrations[C]//SNAME Maritime Convention. SNAME, Italy: 2021: 1-15.
[6]	ZHANG B, XIANG Y, HE P, et al. Study on prediction methods and characteristics of ship underwater radiated noise within full frequency[J]. Ocean Engineering, 2019, 174: 61-70. doi: 10.1016/j.oceaneng.2019.01.028
[7]	GUO J, WANG M, KANG Y, et al. Prediction of ship cabin noise based on RBF neural network [J]. Mathematical Problems in Engineering, 2019: 1-12.
[8]	徐源超, 蔡志明, 孔晓鹏, 等. 船舶辐射噪声分类卷积神经网络的可视化分析和卷积核剪枝[J]. 电子与信息学报, 2023, 45(1): 1-9. XU Y C, CAI Z M, KONG X P, et al. Visualization analysis and kernel pruning of CNN for ship radiated noise classification[J]. Journal of Electronics & Information Technology, 2023, 45(1): 1-9.
[9]	SHIKI T. Estimation of prediction error by using K-fold cross-validation[J]. Statistics and Computing, 2011, 21(2): 137-146. doi: 10.1007/s11222-009-9153-8
[10]	DIETTERICH T G. Ensemble learning[J]. The Handbook of Brain Theory and Neural Networks, 2002, 2(1): 110-125.
[11]	DONG X, YU Z, CAO W, et al. A survey on ensemble learning[J]. Frontiers of Computer Science, 2020, 14(2): 241-258. doi: 10.1007/s11704-019-8208-z
[12]	WEBB G I, ZHENG Z. Multistrategy ensemble learning: Reducing error by combining ensemble learning techniques[J]. IEEE Transactions on Knowledge and Data Engineering, 2004, 16(8): 980-991. doi: 10.1109/TKDE.2004.29
[13]	BREIMAN L. Bagging predictors[J]. Machine Learning, 1996, 24(2): 123-140. doi: 10.1023/A:1018054314350
[14]	SCORNET E, BIAU G. A random forest guided tour[J]. Test, 2016, 25(2): 197-227. doi: 10.1007/s11749-016-0481-7
[15]	SCHAPIRE R E. The strength of weak learnability[J]. Machine Learning, 1990, 5(2): 197-227. doi: 10.1023/A:1022648800760
[16]	FRIEDMAN J, HASTIE T, TIBSHIRANI R. The elements of statistical learning[M]. Beijing: World Publishing Corporation, 2009: 1-745.
[17]	杨欢, 吴震, 张鹏, 等. 侧信道多层感知机攻击中基于贝叶斯优化的超参数寻优[J]. 计算机应用与软件, 2021, 38(5): 323-330. YANG H, WU Z, ZHANG P, et al. Bayesian optimization based hyperparameter search in side-channel attacks against multilayer perceptron[J]. Computer Applications and Software, 2021, 38(5): 323-330.
[18]	RASMUSSEN C E, WILLIAMS C K I. Gaussian processes for machine learning[M]. Cambridge: MIT Press, 2006: 1-248.
[19]	SNOEK J, LAROCHELLE H, ADAMS R P. Practical Bayesian optimization of machine learning algorithms [J]. Advances in Neural Information Processing Systems, 2012: 4.