基于变点分析思想的高维机械噪声数据本征维估计方法

利用变点分析思想,针对高维数据协方差矩阵的特征值曲线特点,提出一种本征维估计方法:基于均值估计的特征值变点法,并应用于某双层圆柱壳体的机械噪声数据处理中。通过与其他现有常用方法的比较分析,验证了基于均值估计特征值变点法的可行性。结果显示:基于均值估计的特征值变点法理论简单、计算简便,计算结果客观、合理,对某双层圆柱壳体机械噪声数据的处理效果较好。

A method of latent characteristic variables dimensionality evaluating of High-dimensional mechanical noise data based Change-Point analysis

According to eigenvalue-curve’s characteristic of covariance matrix from a High-dimensional data, a new method of latent characteristic variables dimensionality evaluating is proposed based change-point analysis technology, which called Eigenvalue Change-point method based mean estimating, and then apply it to deal with the mechanical noise data of ribbed cylindrical double-shells. Through the analysis compar-ing with other methods, the proposed method was validated in its feasibility. It is also verified that Eigen-value Change-point method based mean estimating is simply in theory and convenient in computing; its com-pute result is impersonal and reasonable; meanwhile, it has attained a preferable result when dealing with the mechanical noise data of a ribbed cylindrical double-shells.