摘要
虚拟样本是解决小样本条件下模式识别问题的有效手段。在总结前人研究成果的基础上给出了虚拟样本的定义,利用VC维理论分析了虚拟样本的作用。针对双层圆柱壳体结构噪声源识别这一实际问题,提出了两种分别基于频响函数和傅里叶变换的虚拟样本生成算法。试验结果表明两种算法均能有效提高噪声源识别正确率。
On the base of summarizing the previous research results, the definition of virtual sample was given and the effect of virtual sample work was analyzed utilizing VC-dimension theory. In order to achieve the identification of noise sources in the ribbed cylindrical double-shells under limited samples condition, two algorithms based on the inherent invariance of FRF (frequency response function) and FT (Fourier transform) respectively were studied. The test results show that both algorithms can improve the identification rate effectively.
出处
《振动与冲击》
EI
CSCD
北大核心
2008年第5期32-35,共4页
Journal of Vibration and Shock
基金
国家自然科学基金(50775218,10674151)
国防科技预研基金(9140A10050506JB1113)资助
关键词
虚拟样本
双层柱壳结构
噪声源识别
频响函数
virtual samples
ribbed cylindrical double-shells
noise source identification
frequency response function
