摘要
在机械设备的故障诊断中 ,常采用BP网络算法对故障进行诊断计算 ,但由于BP网络易于收敛于局部极小点 ,且在初始参数与网络结构选取不当时 ,网络将出现发散现象。为了克服这一缺陷 ,该文提出将神经网络优化算法应用于汽轮发电机组的故障诊断中 ,实现了神经网络权值和阈值的快速计算 ,并以汽轮发电机组的故障诊断为背景 ,将神经网络快速算法的结果与BP网络算法的结果进行比较 ,证明该方法不但比BP算法精度高且收敛速度快、可靠性好。
BP algorithm is often used to diagnose and calculate in faults diagnosis of the mechanical equipment.But BP network is apt to converge on the local minimum point. And,if the selected original parameter and network structure is not suitable,the divergent phenomenon will turn up in the network. In order to overcome this defect,this article advances that the neural network optimization algorithm is used in faults diagnosis of the steam electric generating set (rotating mechanical equipment),and the fast calculation in weights values and threshold values of neural network are gained. Based on the steam electric generating set,the neural network optimization algorithm result is compared with BP network algorithm result. It is shown that this method is more fast and higher accuracy than BP algorithm,and the neural network is absolutedly,convergent.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2002年第2期103-106,共4页
Proceedings of the CSEE
关键词
汽轮发电机组
鲍威尔法
BP法
Powell法
人工神经网络
powell fast optimization algorithm
neural network
steam electric generating set
faults diagnosis