基于VMD-LSTM的液压泵健康状态识别研究  

Research on Health Status Identification of Hydraulic Pump Based on VMD-LSTM

作　　者：梁泽慧 LIANG Ze-hui(Institute of Physical and Chemical Engineering of Nuclear Industry,Tianjin 300180,China)

机构地区：[1]核工业理化工程研究院,天津300180

出　　处：《液压气动与密封》2024年第6期65-72,共8页Hydraulics Pneumatics & Seals

摘　　要：常规的液压泵健康状态识别方法,主要采用信号幅值特征识别出健康状态,忽略了共振频率的影响,导致识别结果状态相关度较低,因此,提出了基于可变记忆深度长短时记忆网络(Variable Memory Depth Long Short-Term Memory,VMD-LSTM)的液压泵健康状态识别研究。利用VMD-LSTM结合的算法,通过3个步骤对液压泵信号进行去噪、预加重以及分帧处理,分析了处理信号与液压泵材料之间的共振频率,由此提取出排除共振频率影响的时域特征,将该特征代入到算法中识别得出液压泵的退化率健康状态。实验结果表明:方法能够实现对液压泵健康状态的识别,并且状态相关度较高,识别结果较为准确,满足了液压泵在实际应用中的安全运维需求。The conventional method for identifying the health status of hydraulic pumps mainly uses signal amplitude features to identify the health status,ignoring the influence of resonance frequency,resulting in low correlation between the recognition results and the state.Therefore,a study on hydraulic pump health status recognition based on Variable Memory Depth Long Short-Term Memory(VMD-LSTM) network was proposed.Using the algorithm combined with VMD-LSTM,the hydraulic pump signal was denoised,pre emphasized,and frame processed in three steps.The resonance frequency between the processed signal and the hydraulic pump material was analyzed,and time-domain features that exclude the influence of resonance frequency were extracted.This feature was then applied to the algorithm to identify the degradation rate and health status of the hydraulic pump.The experimental results show that the method proposed in this paper can achieve the recognition of the health status of hydraulic pumps,and the state correlation is high.The recognition results are relatively accurate,meeting the safety operation and maintenance needs of hydraulic pumps in practical applications.

关 键 词：液压泵 VMD-LSTM 健康状态 状态识别 退化率 

分 类 号：TH137[机械工程—机械制造及自动化] U284[交通运输工程—交通信息工程及控制]