基于精细复合多元多尺度散布熵和深度残差收缩网络的轴向柱塞泵故障诊断  

Fault Diagnosis of Axial Piston Pump Based on Refined Composite Multi-variate Multi-scale Dispersion Entropy and Deep Residual Shrinkage Network

作　　者：储焰 常远 汤何胜[1] CHU Yan;CHANG Yuan;TANG Hesheng(School of Mechanical Engineering,Wenzhou University,Wenzhou Zhejiang 325025,China)

机构地区：[1]温州大学机电工程学院,浙江温州325035

出　　处：《机床与液压》2025年第6期142-147,共6页Machine Tool & Hydraulics

基　　金：国家自然科学基金项目(52275064);温州市重大科技创新攻关项目(ZG2021019)。

摘　　要：为了克服单传感器振动信息不能全面表达柱塞泵故障特征信息的问题,提出一种新的轴向柱塞泵故障诊断方法,将精细复合多元多尺度散布熵(RCMMDE)嵌入深度残差收缩网络(DRSN)框架中,更精确地提取轴向柱塞泵非线性故障特征。通过RCMMDE全面表征轴向柱塞泵故障信息,构建故障特征集;利用DRSN对轴向柱塞泵的故障进行分类;最后,通过轴向柱塞泵故障模拟实验,获取典型故障信号特征,并与其他智能诊断算法进行对比,验证模型的泛化能力,实现柱塞泵故障特征的精准识别。结果表明:随着尺度因子的增大,RCMMDE可实现轴向柱塞泵微弱故障特征的有效分离;DRSN模型提高了对高噪声振动信号的特征学习能力,故障诊断精度达到96.21%,明显优于其他分类算法。In order to solve the problem that a single sensor cannot fully express the fault characteristic of the piston pump,a new fault diagnosis method was proposed.The refined composite multi-variate multi-scale dispersion entropy(RCMMDE)was integrated into a deep residual shrinkage network(DRSN)framework to accurately extract nonlinear fault characteristics of axial piston pumps.The fault information of axial piston pump was comprehensively characterized through RCMMDE and a fault feature set was constructed.DRSN was used for fault classification of axial piston pump.Finally,through fault injection experiment of axial piston pump,typical fault characteristics were obtained and it was compared with other intelligent diagnostic algorithms to verify the generalization ability of the proposed model,achieving accurate recognition of piston pump fault characteristics.The results indicate that the RCMMDE can effectively separate the various weak fault characteristic information of the axial piston pump with the increase of the scale factor.The DRSN model enhances its ability to learn features from high noise vibration signals,and the fault diagnosis accuracy reaches 96.21%,which outperforms other classification algorithms significantly.

关 键 词：轴向柱塞泵 故障诊断 精细复合多元多尺度散布熵(RCMMDE) 深度残差收缩网络(DRSN) 

分 类 号：TH137.5[机械工程—机械制造及自动化]