基于油液-振动多维特征与PSO-LSTM的齿轮箱磨损状态监测  

Gearbox Wear Condition Monitoring Based on Multi-dimensional Features of Oil-vibration and PSO-LSTM

作　　者：郭向阳 王建国[1] 范斌 张超[1] 于航 GUo Xiangyang;WANG Jianguo;FAN Bin;ZHANG Chao;YU Hang(School of Mechanical Engineering,Inner Mongolia University of Science and Technology,Baotou Inner Mongolia 014010,China;School of Mechanical and Electrical Engineering,Inner Mongolia Agricultural University,Hohhot Inner Mongolia 010018,China)

机构地区：[1]内蒙古科技大学机械工程学院,内蒙古包头014010 [2]内蒙古农业大学机电工程学院,内蒙古呼和浩特010018

出　　处：《润滑与密封》2023年第12期117-123,共7页Lubrication Engineering

基　　金：国家自然科学基金地区科学基金项目(51965054,51865045);内蒙古自治区自然科学基金面上项目(2021MS05041);内蒙古农业大学高层次人才科研启动项目(NDYB2019-9)。

摘　　要：针对单独从振动特征、油液特征对齿轮箱进行磨损状态监测存在特征维度单一、准确率低的问题,提出基于油液-振动多维特征与粒子群优化算法-长短时记忆神经网络(PSO-LSTM)的齿轮箱磨损状态监测算法。对铁谱图像进行预处理,提取磨粒浓度特征、磨粒个数特征,对振动信号进行小波阈值去噪,并提取时域特征,得到油液振动十四维特征作为LSTM模型的输入;采用粒子群优化算法对LSTM模型进行参数寻优。实验验证:使用油液振动十四维特征的PSO-LSTM模型的识别准确率要优于单独使用振动和油液特征的PSO-LSTM模型,PSO-LSTM模型对于油液振动十四维特征数据的识别准确率全面优于未经优化的LSTM模型。In order to solve the problem of single feature dimension and low accuracy in monitoring gearbox wear condition from vibration and oil features alone,a gearbox wear condition monitoring algorithm based on oil-vibration multi-dimensional features and particle swarm optimization algorithm-long and short term memory neural network(PSO-LSTM)was proposed.The ferrography images were preprocessed to extract the features of wear particle concentration and the number of wear particles.The vibration signal was denoised by wavelet threshold,and the time-domain features were extracted.The 14-dimensional characteristics of oil-vibration were obtained as the input of LSTM model.Particle swarm optimization algorithm was used to optimize the parameters of LSTM model.The experimental results show that the recognition accuracy of the PSO-LSTM model using the 14-dimensional features of oil-vibration is better than that of the PSO-LSTM model using the vibration and oil characteristics alone.The recognition accuracy of the PSO-LSTM model is better than that of the unoptimized LSTM model in all aspects.

关 键 词：齿轮箱 多维特征 状态监测 粒子群优化算法 长短时记忆神经网络 

分 类 号：TH117.1[机械工程—机械设计及理论]