摘要
针对传统故障诊断方法特征提取不充分的问题,该文提出了基于变分模态分解(variational mode decomposition,VMD)的混合域特征和深度信念网络(deep belief network,DBN)的故障诊断方法。先采用VMD将故障轴流泵压力脉动信号分解成若干本征模态函数(IMF),计算了相关性较大的各IMF分量的能量值和熵值特征,并与原始信号提取的时域和频域特征组合成混合域特征。再引入遗传算法优化故障特征组合,去除冗余特征,并将优化后的故障特征输入到DBN中进行识别和分类。结果表明:与单一特征集相比,基于混合域特征的故障诊断效果更好;与支持向量机(SVM)相比,深度信念网络在故障诊断中更具优势,分类正确率可达95%。从而表明泵壳测点的压力脉动信号能够较好地反映轴流泵模型的故障特性,泵壳测点可作为故障监测点。
Since traditional fault diagnosis methods cannot fully capture the fault features,this paper proposes a fault diagnosis method based on mixed domain features of variational mode decomposition(VMD)and deep belief network(DBN).The pressure pulsation of the faulty axial flow pump is decomposed into several intrinsic mode functions using VMD,the energy and entropy characteristics of each IMF component with large correlation are calculated,and combined with the time domain and frequency domain features extracted from the original signal to form a mixed domain feature.Genetic algorithm is introduced to optimize the combination of fault features,remove redundant features,and input the optimized fault features into DBN for recognition and classification.The results show that compared with the single feature set,the fault diagnosis based on the mixed domain features offers better performance;Compared with support vector machine,deep belief network has more advantages in fault diagnosis,the classification accuracy can reach 95%.This indicates that the pressure pulsation signal of the pump housing can better reflect the fault characteristics of the axial flow pump model,the pump housing measurement point can be used as the fault detection point.
作者
吴咏
陈红勋
马峥
Wu Yong;Chen Hongxun;Ma Zheng(Shanghai Institute of Applied Mathematics and Mechanics,School of Mechanics and Engineering Science,Shanghai University,Shanghai 200072,China;Shanghai Branch,Chinese Ship Scientific Research Center,Shanghai 20001l,China)
出处
《水动力学研究与进展（A辑）》
CSCD
北大核心
2024年第2期244-251,共8页
Chinese Journal of Hydrodynamics
关键词
轴流泵
故障诊断
变分模态分解
混合域特征
遗传算法
深度信念网络
Axial flow pump
Fault diagnosis
Variational mode decomposition
Mixed domain feature
Genetic algorithm
Deep belief network
