在引入了一套磨粒形态学描述子来提取磨损颗粒的显微形态特征的基础上 ,采用人工神经网络技术 ,编制了用于磨损颗粒自动识别的 BP网络计算机模拟程序。在网络训练的过程中应用了本文引入的因子模糊化训练方法 ,使训练速度大大加快 ,以异或问题为例 ,速度可提高 5～ 1 0倍。应用此网络对磨粒测试库进行识别实验 ,识别正确率在 90 %以上 ,并且识别速度很快 ,大大优于传统的磨粒识别方法。 Based on the introduction of a set of abrasive morphological descriptors to extract the microscopic features of wear particles, a BP neural network simulation program for automatic identification of wear particles was developed by using artificial neural network technology. In the process of network training, the fuzzy fuzzification training method introduced in this paper is applied, so that the training speed is greatly accelerated. Taking the exclusive-or problem as an example, the speed can be increased by 5 to 10 times. Application of this network to wear abrasive test library recognition experiments, recognition accuracy rate of 90% or more, and recognition speed, much better than the traditional method of abrasive identification.