摘要
提出一种神经网络与符号推理集成系统ＩＮＮＳＩ（ＩｎｔｅｇｒａｔｉｏｎｏｆＮｅｕｒａｌＮｅｔ－ｗｏｒｋｓａｎｄＳｙｍｂｏｌｉｃＩｎｆｅｒｅｎｃｅ）．它包括两部分：根据领域规则构造神经网络；神经网络根据样本进行学习．该系统的特点是网络的拓扑结构在学习过程中可以动态地变化．通过齿轮箱故障诊断试验，可知应用该方法诊断齿轮箱故障是有效的，并可较大幅度地提高诊断的速度和精度．
A new system of integration of neural networks and symbolic inference (INNSI) is put forward. It includes two parts: 1) building neural networks on the basis of the rules; 2) training the neural networks by the samples. The strong point of the system is that the topologic structure of neural networks can be dynamically changed in learning process. Through the experiments of fault diagnosis for a gear train, the new system is proved to be effective and to be able to increase the speed and precision of fault diagnosis greatly.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
1997年第6期39-43,共5页
Journal of Southeast University：Natural Science Edition
基金
中国博士后科学基金
关键词
神经网络
符号推理
集成系统
算法
故障诊断
neural networks
symbolic inference
integrated systems
algorithms
fault diagnosis
