摘要
将支持向量机方法用于铁谱磨粒模式识别,以磨粒样本的圆形度、细长度、散射度和凹度4个形态特征量作为支持向量机分类器的输入,以滑动磨损、切削磨损、正常磨损和疲劳点蚀4种磨损形式作为分类器的输出,建立基于支持向量机的磨粒分类器;研究支持向量机中误差惩罚系数和核参数对磨粒分类器的性能影响;通过实验比较了基于支持向量机与基于BP神经网络的磨粒分类器的性能,结果表明,基于支持向量机的磨粒分类器分类准确率为96%,基于BP神经网络的磨粒分类器分类准确率为90%。
Support vector machine （SVM） was used in the pattern recognition of ferrograghy wear particle. Taking roundness, slenderness, dispersity and concavity as SVM classifier＇s inputs, and four wear patterns （severe wear, cutting wear, normal wear, fatigue wear） were outputs of the SVM classifier. Influences of error punish modulus and kernel function over wear particle classifier were presented. Authors compared the performances of wear particle classifier based on SVM with that of artificial neural network classifier under experiments. The results show that classification accuracy rate of wear particle classifier based on SVM is 96%, while that based on BP neural network is 90%.
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2006年第13期1391-1394,共4页
China Mechanical Engineering
基金
国家自然科学基金资助项目(50375141)
关键词
磨损
铁谱技术
模式识别
BP神经网络
支持向量机
磨粒识别
ferrography
pattern recognition
BP neural network
support vector machine （SVM）
wear particle recognition
