摘要
给出一个BP神经网络,应用频响函数作为神经网络的输入参数,用来识别结构的状态信息,由于频响函数数据量大,直接作为神经网络输入参量容易造成网络训练收敛慢或不收敛。将频响函数在(0,1)范围进行归一化处理,通过用BP网络试验,新方法收敛速度提高了25倍,表明这是一种简单有效的方法。
This paper presented a Back-Propagation network method(BP network),using struc- tural frequency response function as the network input parameters.This method was proposed to i- dentify a structure's state information.Except many advantages such as simpleness,this method had also many shortages.For example,the networks training will be difficult because of its large data.In general,input parameters should be normalized between 1 to+1.The frequency response function data was normalized between 0 and 1.An example was given in details,in which a BP network was used.Two kinds of network are compared,the new method gives an answer much faster then tradi- tional ways for 25 times.And the results show that this method is more simple,and effective.
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2006年第S2期281-283,共3页
China Mechanical Engineering
关键词
频响函数
神经网络
故障诊断
输入参数
frequency response function
neural network
fault diagnosis
input parameter
