转台的粗糙神经网络故障诊断系统设计  

Design of rough-neural network fault diagnosis system for turntable

作　　者：赵佰亭[1] 贾晓芬[1] 曾庆双[2] 

机构地区：[1]安徽理工大学电气与信息工程学院,淮南232001 [2]哈尔滨工业大学空间控制与惯性技术研究中心,哈尔滨150001

出　　处：《中国惯性技术学报》2012年第4期501-504,共4页Journal of Chinese Inertial Technology

基　　金：国防科技预研基金项目(9140A17030207HT0150);安徽理工大学博士基金(11223);安徽理工大学青年基金(2012QNZ06)

摘　　要：针对转台故障的多样性与复杂性,设计了独立于专家的粗糙神经网络故障诊断系统。首先建立转台故障诊断决策表,然后用粗糙集方法约简冗余属性,最后设计了神经网络分类器和辨识器。实验结果显示,诊断系统能较好地区分和辨识具有相同故障现象的不同故障,诊断正确率达到96.7%。将粗糙集理论与神经网络相结合,简化了信息表达空间,减小了神经网络结构的复杂性,并具有强大的容错和抗干扰能力,工程实用性强。In view of the diversity and complexity of the turntable failure, an expert-independent fault diagnosis system was designed based on rough-neural network. Firstly, the fault diagnosis decision table was established, and then the attributes are reduced by a rough-set method. Finally, the neural network classifier and recognizer were designed. The experiment results show that the diagnosis system could distinguish and identify the different faults with the same failure phenomena, and the diagnostic accuracy is up to 96.7%. By combing the rough sets with neural network, the rough sets can reduce the attributes and delete the redundancy. The rough-neural network can simplify the training sets, reduce the complexity of the neural network structure, and has the powerful fault tolerance and anti-jamming capability. The system has strong engineering practicality.

关 键 词：故障诊断 粗糙集 神经网络 转台 

分 类 号：U666.1[交通运输工程—船舶及航道工程]