摘要
针对传统模糊形态学联想记忆网络(FMAM)不能自适应选择结构元素的形状和大小,在对样本进行分类时存在错分问题,以量子的叠加、坍塌性质为基础,提出量子模糊形态学联想记忆网络(QFMAM),用量子位系统构造结构元素,量子位概率代表相应的隶属度,获取具有自适应特性的结构元素,在分类前先对样本进行处理从而降低无用干扰信息对分类精度的影响。利用QFMAM分别对仿真数据和齿轮箱台架实验信号进行分类,并与FMAM、支持向量机(SVM)、朴素贝叶斯分类器(NBC)的分类性能作比较,验证了提出的QFMAM训练效率高、学习能力强、分类精度高,是一种很好用的智能分类器。
The traditional fuzzy morphological associative memories network(FMAM)can not be adaptive to choose the shape and size of structural elements,so there is a problem of misclassification when classifying samples by FMAM.Based on the properties of quantum superposition and collapse,quantum fuzzy morphological associative memories network(QFMAM)was proposed,in which the structural elements were constructed by the qubit system,and the qubit probability represented the corresponding membership degree to obtain adaptive structural elements.The samples were processed with the adaptive structural elements before classified using QFMAM,thereby reducing the influence of useless interference information on classification accuracy.QFMAM was used to classify the simulation data and experimental signals of gear box respectively.It is verified that compared with the classification performance of FMAM,support vector machine(SVM)and naive bayes classifier(NBC),the proposed QFMAM has the advantages of higher training efficiency,stronger learning ability and higher classification accuracy,therefore QFMAM is a very useful intelligent classifier.
作者
吕纯
张培林
李兵
吴定海
张云强
LV Chun;ZHANG Peilin;LI Bing;WU Dinghai;ZHANG Yunqiang(Shijiazhuang Campus,Army Engineering University,Shijiazhuang Hebei 050003,China;College of Engineering, Shantou University,Shantou Guangdong 515063,China)
出处
《润滑与密封》
CAS
CSCD
北大核心
2021年第2期87-92,共6页
Lubrication Engineering
关键词
形态学联想记忆
量子
智能分类器
模糊形态学
齿轮箱
morphological associative memories
quantum
intelligent classifier
fuzzy morphology
gearbox
