摘要
在模糊逻辑与神经网络融合的基础上,引入补偿运算单元,构成补偿模糊神经网络,使网络从初始定义的模糊规则进行训练,再动态的优化模糊规则,提高网络的容错率和稳定性.针对网络训练的不同阶段对学习速率的不同要求,提出一种具有分段可变学习速率的补偿模糊神经系统,可以提高网络的整体性能,实现动态的、全局优化的运算.故障诊断仿真研究表明:模型具有更好的收敛特性,能够大大的缩短训练时间,减少训练步数,提高误差精度.
On the basis of the combination of fuzzy logic and neural network,a compensatory unit is introduced to make up a compensatory fuzzy neural network,which makes the network being training from the fuzzy rules defined initially and dynamically optimizes the fuzzy rules to improve the network fault-tolerant rate and stability.In order to meet the different requirements of the learning rate for the different stages of the training,a compensation fuzzy neural network based on segmentation variable learning rate is proposed to improve the overall performance of the network and realize dynamic and globally optimal calculation.The simulation about fault diagnosis has shown the model has better convergent characteristics,greatly reduce the training time and training steps and improve the error precision.
出处
《华侨大学学报（自然科学版）》
CAS
北大核心
2012年第1期1-5,共5页
Journal of Huaqiao University(Natural Science)
关键词
故障诊断
模糊逻辑
神经网络
分段可变
学习速率
fault diagnosis
fuzzy logic
neural network
segmentation variable
learning rate
