摘要
针对故障诊断信号特征提取问题,提出了小波包分析的改进算法,该算法通过对小波包分解系数的重新排序,解决了小波包分析的频带混叠问题,给出了应用改进小波包分析进行故障诊断特征提取的算法,并在此基础上提出了基于改进小波包分析预处理的神经网络故障字典法.通过仿真比较,该方法剔除了样本信号的冗余成分,大幅度地减少了神经网络的规模,加快了网络的收敛速度,为导航设备故障诊断的特征提取提供了行之有效的手段.
Aiming at the problem of signal feature extraction in fault diagnosis,the improved wavelet packet algorithm is put forward. The problem of block overlap of frequency bands is solved by the recomposition of wavelet packet decomposing coefficient in the algorithm, then, the method of its application to feature extraction in fault diagnosis is also presented, and a new system based on BEP networks with the improved wavelet packet algorithm is given. By simulation and comparison, due to eliminating the redundancy of signal, the system can sharply reduce the network size and have a faster training speed, and it offers an effective way to feature extraction in fault diagnosis of navigation equipment.
出处
《光电技术应用》
2005年第4期56-58,共3页
Electro-Optic Technology Application
关键词
特征提取
故障诊断
小波包
神经网络
导航设备
feature extraction
fault diagnosis
wavelet packet
neural network
navigation equipment