基于VMD分解与卷积神经网络的液压泵故障诊断  

Fault diagnosis of hydraulic pump based on VMD decomposition and convolutional neural network

作　　者：李军 江水 徐启胜 李岩 LI Jun;JIANG Shui;XU Qisheng;LI Yan(Anhui 3H1 Information Technology Co.,Ltd.,Hefei 230123,Anhui China)

机构地区：[1]安徽三禾一信息科技有限公司,安徽合肥230123

出　　处：《锻压装备与制造技术》2022年第3期82-87,共6页China Metalforming Equipment & Manufacturing Technology

摘　　要：针对液压泵数据退化特征维数高以及故障诊断精度不高的问题,提出了一种基于变分模态分解(Variational mode decomposition,VMD)与卷积神经网络(Convolutional Neural Network,CNN)的液压泵故障诊断方法。利用VMD良好的分解能力处理高维度数据,进行数据扩展,提取详细特征;基于CNN良好的特征提取和分类性能,在不需要先验知识的情况下直接从数据中提取特征,实现高精度故障诊断。该方法因具有端到端特征学习能力,在实测液压泵数据上进行验证,具有较高的故障诊断精度和稳定性。Aiming at the problems of high feature dimension data degradation and low fault diagnosis accuracy of hydraulic pump,a new fault diagnosis method of hydraulic pump based on Variational mode decomposition(VMD) and Convolutional Neural Network(CNN) has been proposed.The good decomposition ability of VMD has been used to process high-dimensional data,perform data expansion,and extract detailed features;based on the good feature extraction and classification performance of CNN,features can be directly extracted from data without prior knowledge to achieve high-precision fault diagnosis.Due to its end-to-end feature learning ability,the method has been verified on the measured hydraulic pump data,which has high fault diagnosis accuracy and stability.

关 键 词：液压泵 变分模态分解 卷积神经网络 故障诊断 

分 类 号：TH137.9[机械工程—机械制造及自动化] V245.1[航空宇航科学与技术—飞行器设计]