摘要
提出一种基于注意力机制(Attention Mechanism,AM)的双向门控循环神经网络模型的齿轮故障识别系统。使用基于STM32的嵌入式主控制器分别采集正常齿轮、断齿齿轮、轮齿剥落齿轮等3种故障齿轮工作时的振动传感器数据,使用基于注意力机制的双向门控循环单元网络模型进行齿轮故障识别。双向门控循环神经网络模型添加了注意力机制,保留输入特征的重要信息,不随步长增加而消失。将采集到的原始数据集按7∶2∶1的比例划分为训练集、验证集和测试集。测试集模型的齿轮故障识别准确率达到了99.67%,与GRU和Bi-GRU等模型的结果对比证明该模型的正确率更高。本系统可用于汽车变速器的监测与故障诊断。
This paper proposes a gear fault identification system including a bidirectional gated recurrent unit(Bi-GRU)neural network model with attention machanism.Based on the STM32 MCU,the data acquisition system collects separately the vibration sensor data of three kinds of gears,i.e.the normal gears,the gears with broken teeth,and the gears with tooth peeling damage.The acquired data is divided into training,validation and test set according to the ratio of 7∶2∶1.The attention mechanism is added into the Bi-GRU model to retain the important information of the input features even with the increase of time series.Finally,the gear fault identification accuracy of the model on the test set reaches 99.67%,higher than the results obtained using GRU or Bi-GRU without the attention mechanism.The proposed system can be used for monitoring and fault diagnosis of automobile gearboxes.
作者
冯贤洋
何荇兮
符礼丹
陆彬春
陈鸣辉
FENG Xianyang;HE Xingxi;FU Lidan;LU Binchun;CHEN Minghui(The State Key Laboratory of Mechanical Transmission,Chongqing University,Chongqing 400044,China)
出处
《汽车工程学报》
2023年第1期111-117,共7页
Chinese Journal of Automotive Engineering
基金
中央高校基本科研业务费资助项目(2019CDQYJX008)。
关键词
嵌入式
齿轮故障识别
双向门控网络
注意力机制
embedded
gear failure recognition
bi-directional gated recurrent neural network
attention mechanism
