基于分层模糊支持向量机的油液磨粒自动识别  

Automatic Recognition of Wear Debris in Oil Based on Hierarchical Fuzzy Support Vector Machines

作　　者：任松[1] 徐雪茹 欧阳汛 赵云峰[1] 王小书[1] REN Song;XU Xueru;OU Yangxun;ZHAO Yunfeng;WANG Xiaoshu(State Key Laboratory of Coal Mine Disaster Dynamics and Control,Chongqing University,Chongqing 400030,China)

机构地区：[1]重庆大学煤矿灾害动力学与控制国家重点实验室,重庆400030

出　　处：《润滑与密封》2019年第5期1-8,共8页Lubrication Engineering

基　　金：国家自然科学基金项目(51774057)

摘　　要：提出基于模糊支持向量机的机械设备在用油液磨粒自动识别方法。首先利用K-均值聚类算法对磨粒图像进行分割,提取磨粒的形状尺寸特征参数、边缘细节特征参数、表面纹理特征参数作为其量化表征,分别选择最能反映待识别磨粒特征的参数作为各个二分类器的输入向量;然后结合二叉树法和一对多法间接构造磨粒的分层多类别分类器模型,在训练过程中同时利用粒子群算法优化分类器的参数,建立一种参数自适应的模糊支持向量机分层多类别分类模型。将该模型应用到旋挖钻机在用油液的磨损颗粒识别中,识别率最高达90%。该模型结构简单、分类精度好,在磨粒识别领域较大的工程应用价值。A automatic recognition of wear debris extracted from mechanical equipment oil based on Fuzzy support vector machines was proposed.The wear debris image extracted from mechanical equipment oil was segmented by K-means clustering algorithm to extract the feature parameters as their quantitative characteristics,including shape feature,edge detail and surface texture feature.The parameters that best reflect the characteristics of the wear debris to be identified were selected as input vectors for each two-class classifier.Based on the decision tree and one-against-rest method,a new hierarchical multi-class classification model of fuzzy support vector machines was established.The particle swarm optimization was used to optimize parameters of the model during the training process.The proposed model was applied to identify wear particles extracted from lubricant and hydraulic fluid of a rotary drilling rig,the recognition rate was up to 90%.The proposed model has advantages of simple structure and good classification accuracy,and has a great engineering value in the field of wear debris recognition.

关 键 词：磨粒识别 模糊支持向量机 粒子群算法 特征参数 分层多类别分类 

分 类 号：TH117[机械工程—机械设计及理论] TP181[自动化与计算机技术—控制理论与控制工程]