基于 LSTM-ICNN的烟草包装机传动系统滚动轴承状态预测研究  

Research on rolling bearing state prediction of tobacco packaging machine transmission system based on LSTM-ICNN

作　　者：江逸斐 陈忠华 兰志超 王少禹 张乐 Jiang Yifei;Chen Zhonghua;Lan Zhichao;Wang Shaoyu;Zhang Le(Xiangyang Cigarette Factory of Hubei China Tobacco Industry Co.,Ltd.,Hubei Xiangyang,441021,China)

机构地区：[1]湖北中烟工业有限责任公司襄阳卷烟厂,湖北襄阳441021

出　　处：《机械设计与制造工程》2024年第3期97-101,共5页Machine Design and Manufacturing Engineering

基　　金：襄阳市卷烟厂易地技术改造项目(THZBHB-21102)。

摘　　要：为提高烟草包装机传动系统滚动轴承状态预测精度,提出一种基于长短时记忆(LSTM)卷积神经网络结合改进卷积神经网络(ICNN)的轴承状态预测方法。首先通过LSTM提取滚动轴承的时序特征;然后在卷积神经网络(CNN)全连接层中嵌入局部最大均值差异函数,从而提取域不变特征,并通过回归损失函数输出传动系统滚动轴承状态预测结果;最后对以上预测方法进行试验验证。试验结果表明,在不同工况下,网络预测模型的RMSE和MAE都较小,且在实际在线监测系统应用中,RMSE和MAE分别为0.082和0.065。由此说明,提出的网络预测模型具有良好的预测精度,可用于烟草设备的在线故障监测。In order to improve the prediction accuracy of rolling bearing state,a bearing state prediction method based on long short-term memory(LSTM)convolutional neural network combined with improved convolutional neural network(ICNN)is proposed.Firstly,the time-sequenced features of the rolling bearing are extracted by LSTM;then the local maximum mean difference function is embedded in the full connection layer of the convolutional neural network CNN network,so as to extract the domain invariant features,and the regression loss function is used to output the prediction results of the rolling bearing status of the transmission system;finally,the above prediction method is tested.The test results show that the RMSE and MAE of the network prediction model are small under different working conditions,and in the actual application of online monitoring system,the RMSE and MAE are 0.082 and 0.065 respectively.This indicates that the proposed network prediction model network has good prediction accuracy and can be used for online fault monitoring of tobacco equipments.

关 键 词：烟草设备 包装机 状态预测 在线监测 长短时记忆卷积神经网络 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]