摘要
为有效提取往复泵工作时非平稳时变信号中的故障特征和将故障特征准确分类,提出以泵缸内的压力作为系统特征信号来提取故障特征向量的方法.将小波包分解的"频率-能量-故障识别"模式诊断方法引入泵阀工作状态监测中,通过改进的BP神经网络进行故障诊断.试验确定了网络的初始值,即选择学习率初始值为1.5、惯性因子为0.6、网络结构为3层的BP网络,其中隐含层的节点数为19个,即网络的结构是8-19-3.结果表明,该法降低了对原始信号处理的难度,且各阀箱内的压力之间无相互影响.该技术已应用于某船载系统的往复泵实时故障诊断中,实验验证了其有效性.
Two key issues of fault diagnosis for the pump valves of reciprocating pump are extracting the fault feature information of non-stationary time variation process efficiently from system feature signals and classifying the specific faults correctly. A diagnosis method based on frequency-energy-fault identification pattern recognition diagnosis approach was adopted to detect the fault on pump valves of reciprocating pumps taking the pressure signals in pump cylinder as system feature ones. The improved BP neural net- work was used to diagnose various faults of pump valves by the feature vectors constructed. The network starting value was determined through the experimental confirmation method, namely selecting 1.5 as the starting value and 0.6 as the inertia factor. The network structure is of 3-layer BP network and the number of node is 19 in the concealment level, namely the network structure is 8-19-3. The results show that the approach deals with the primitive pressure signal simply and acquires fault feature vectors easily, and the pressures in different valve boxes have no influence on each other.
出处
《大连海事大学学报》
EI
CAS
CSCD
北大核心
2007年第3期22-25,31,共5页
Journal of Dalian Maritime University
基金
黑龙江省教育厅科学技术研究基金资助项目(11511021)
关键词
往复泵
小波包
神经网络
故障诊断
特征向量
reciprocating pumps
wavelet packet
neural network
fault diagnosis
feature vectors