摘要
由于潜艇振动噪声源存在频率相互耦合现象,常规方法难以有效地解决耦合噪声源分离与贡献量化问题。采用一种新型多元统计分析方法-偏最小二乘回归分析方法来实现频率耦合噪声源的分离量化,该方法可同时提取反映输入/输出中最大信息且相关性最大的主成分,并能够在变量间存在多重相关性的条件下进行回归建模。仿真与舱段模型试验表明:当多激励源之间存在频率耦合时,能对噪声源进行分离和贡献量化,从而实现了噪声源对耐压壳体观测点贡献以及噪声源对辐射声场观测点贡献的排序,验证了偏最小二乘回归用于频率耦合源分离量化的可行性,为主要噪声源的控制提供了依据。
Traditional methods do not effectively handle separation and quantification of coupled vibration noise sources in submarines. So a new multivariate statistical analysis method,partial least square regression（ PLS）,is presented,which can be used to separate and quantify frequency coupled noise sources. PLS has the characteristic of simultaneously extracting principal input / output components,including maximum information,correlation of input with output,and regression modeling with multiple correlations among variables. Simulation and cabin model experiments show that,when there is frequency coupling between multiple excitation sources,PLS is capable of sorting among the energy contributions of internal noise sources to submarine hull,submarine hull to underwater acoustic field,and noise sources to underwater acoustic field. The feasibility of PLS for frequency coupled source separation and quantification is proven. The method provides a basis for the control of the main noise sources.
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
北大核心
2016年第3期368-375,共8页
Journal of Harbin Engineering University
基金
长江学者和创新团队发展计划资助项目(IRT1228)
关键词
偏最小二乘回归
舱段模型
频率耦合
噪声源分离量化
耐压壳体
辐射声场
PLS regression
submarine cabin
frequency coupling
separation and quantification of noise source
pressure hull
radiated acoustic field
