基于遗传算法改进的BP神经网络模型的磨损机制智能识别  

Intelligent Recognition of Wear Mechanism Based on Improved BP Neural Network Model by Genetic Algorithm

作　　者：盛晨兴[1,2] 程俊[2,3] 李文明[4] 段志和 马奔奔[1,2] 

机构地区：[1]高性能船舶技术教育部重点实验室(武汉理工大学),湖北武汉430063 [2]武汉理工大学能源与动力工程学院,湖北武汉430063 [3]西安交通大学润滑理论及轴承研究所,陕西西安710049 [4]交通运输部南海救助局,广东广州510310

出　　处：《润滑与密封》2014年第1期24-28,共5页Lubrication Engineering

基　　金：国家高技术研究发展计划(863计划)(2011AA110202);武汉理工大学自主创新研究基金项目

摘　　要：通过提取磨粒形状特征参数、颜色特征参数和表面纹理等特征参数对磨粒形态进行量化表征,并以此为输入矢量,引入遗传算法(GA)改进BP神经网络对磨粒进行自动分类识别,建立遗传算法改进的BP神经网络模型,并给出具体的算法实现过程。分别应用遗传算法改进的BP神经网络模型和未引入遗传算法改进的BP神经网络模型对磨粒图像进行智能识别。实验结果表明,遗传算法改进的BP神经网络综合了遗传算法的全局优化和BP算法局部搜索速度快的特点,网络识别率较高,具有较好的全局性。A improved back propagation（BP） neural network by Genetic algorithm was introduced to realize the auto- matic classification and recognition of wear debris, based on the qualitative characterization of the morphological features of the wear debris making use of the characteristic parameters of wear debris shape, color, and surface texture. A neural net- work model based on the improved back propagation （BP） neural network by Genetic algorithm was established to classify and recognize the wear debris using those parameters as the input vectors. The algorithm of the established model was de- tailed. By comparing the results of automatic recognizing the wear debris by the improved BP neural network and the pres- ented BP neural network, it shows that the improved back propagation （BP） neural network combines the global optimiza- tion feature of genetic algorithm and the fast speed feature in local search of BP algorithm, which has a high recognition rate and better global search feature.

关 键 词：特征提取 磨粒识别 遗传算法 BP算法 神经网络 

分 类 号：TH117.1[机械工程—机械设计及理论] TP301.6[自动化与计算机技术—计算机系统结构]