This paper introduces a typical nonlinear dynamics technique , based on the fractal geometry to achieve quantitative description and automatic analysis of wear particles generated between running friction parts inside diesel locomotive , discusses the model and algorithms of fractus calculation of wear particles , and finally gives the correspondent experimental results , in which wear particle ' s automatic recognition is finished by means of the cooperation of fractal geometry and artificial neural network 本文引入分形几何这种非线性系统方法来完成对磨粒的定量描述和自动分析，给出分维数算法模型及相应实验结果，并结合人工神经网络来完成磨粒的分割与自动识别。
