基于相关系数方法的样本失效分布研究

在分析传统树叶图方法的基础上,指明了直观判断失效分布的弊端,提出了用量化理论曲线和试验曲线间相似度的方法来确定样本失效分布的思想,进而考虑用相关系数方法来量化曲线间的相似度。在分析数据平滑的重要意义的基础上,用小波分析提取了数据轮廓,进而首次将平滑数据和理论数据的相关系数用于样本失效分布的确定。实验结果表明,基于平滑数据相关系数的相似度方法能够区分曲线间形状的差异,该方法为失效数据分布的判定提供了量的标准,因而是对原有树叶图方法的进一步补充和完善。

Research on failure distribution of samples based on correlation coefficient method

The shortcoming of determining failure distribution in intuition is pointed out,and an idea of quantitating similarity between theoretic curve and testing curve to determine the failure distribution of samples is presented,which is based on analyzing the traditional leaf-figure method.Furthermore,the correlation coefficient method is considered to quantitate similarity between two curves.At the same time,the importance of smoothing data is analyzed.Then,the correlation coefficient between smoothing data and theoretic data is firstly used to determine the failure distribution of samples,which is based on extracting an outline of testing data by means of wavelet transformation.The experiment result shows that the correlation coefficient method based on smoothed data could distinguish the shape difference between two curves,and offers a quantitative standard for determining the distribution of failure data.In sum,the presented method is a further supplement to the previous leaf-figure method.