中文摘要：

采用主成分聚类分析理论对正常磨粒、滑动磨粒、切削磨粒和疲劳磨粒等4种不同磨粒表面纹理灰度共生矩阵进行分析,得到4个不同纹理方向的能量、熵、惯性矩、局部平稳性、相关、最大概率和方差7个参量.利用主成分理论求解,构造主成分方程,得到磨粒纹理参量主成分的二维分布,最后,通过聚类分析,将4种磨粒纹理参量的主成分划分为不同的聚类区域.结果表明：主成分聚类分类方法可以对磨粒的多参量进行聚类分析;主成分聚类分类方法对冗余数据具有较强地处理能力;多组主成分聚类分析方法具有互补性,特别适合多参量磨粒分类的信息融合处理.

英文摘要：

The character of inhomogeneity wear particles surface texture is different from each others in wear particles identify. A texture analysis method based on gray level co-occurrence matrix is presented for judging the sort of wear particles which include rubbing, sliding, cutting and fatigue wear particles. The processing methods produced seven parameters, including energy, entropy, moment, local Uniformity, correlation, Maximum probability, and mean square error at four directions. Then, through gray level co-occurrence matrix and principal component analy- sis, two equations were constructed for seven parameters for principal component analysis. Two-dimensional distribution figure for different texture of four kind wear particles were obtained. Finally, two-dimensional distribution figure was divided into four portions through solution of three equations, which was classified as four kind wear particles. This method was used to analyze and identify the type of wear particles. The advantages of this method includes （ 1 ） Component Classification Analysis method can be used to classify different kind of wear particles, （2） the method deals with redundancy data, （ 3 ） More sets of Component Classification Analysis are mutually related each other, （4） this method is adaptable for information fusion of other parameters for classification and identification of wear particles.