摘要
为识别出柱塞泵的不同故障模式,提出一种柱塞泵故障诊断方法。首先,将柱塞泵的振动加速度信号数据通过改进的自适应噪声完备集成经验模态分解(Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise,ICEEMDAN)方法分解为多个固有模态分量(Intrinsic Mode Function,IMF),分别计算筛选后的IMF分量的时域、频域特征并拼接为特征向量;其次,使用K近邻算法计算各特征向量之间的相似度,依据相似度生成一种故障样本特征图,实现一维时序数据到图结构数据的转换,刻画不同故障样本之间的联系;最后,将故障样本特征图输入至基于初始残差和恒等映射的图卷积网络(Graph Convolutional Network via Initial Residual and Identity Mapping,GCNII)诊断模型中,准确的识别出柱塞泵的不同故障类型。经实验验证,该方法在5种柱塞泵故障类型上的诊断率达到了98.67%。
In order to identify different failure modes of plunger pump,a fault diagnosis method of plunger pump is proposed.Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(ICEEMDAN) method is decomposed into multiple Intrinsic Mode Functions(IMF),and the time-domain and frequency-domain features of the filtered IMF components are calculated respectively and spliced into feature vectors.Secondly,K-nearest neighbor algorithm is used to calculate the similarity between each feature vector,and a fault sample feature map is generated according to the similarity,which realizes the transformation of one-dimensional time series data to the graph structure data,and describes the relationship between different fault samples.Finally,the fault sample feature map is input into the Graph Convolutional Network via Initial Residual and Identity Mapping(GCNII) diagnostic model.Accurately identify the different fault types of the plunger pump.The experimental results show that the diagnosis rate of 5 kinds of piston pump faults is 98.67%.
作者
杨喜旺
YANG Xi-wang(Shanxi University of Electronic Science and Technology,Linfen 041000,China)
出处
《液压气动与密封》
2024年第6期21-29,共9页
Hydraulics Pneumatics & Seals
基金
山西省重点研发计划(国际科技合作)(201903D421008)
山西省自然科学基金(201901D111157)
山西省回国留学人员科研教研资助项目(2022-141)
山西省基础研究计划资助项目(202203021211096)。
