为提升船舶电机轴承故障诊断精度，确保船舶航行的安全性，研究对抗神经网络算法在船舶电机轴承故障诊断中的应用。采用集合经验模态分解方法求取船舶电机轴承振动信号各固有模态函数分量的能量熵，并使用相关系数法清除虚假分量，将筛选后的有效数据作为船舶电机轴承故障特征。利用对抗神经网络构建轴承故障诊断模型，通过引入条件对抗损失函数解决模型训练过程中的不确定性问题，利用二人零和博弈问题能够描述对抗神经网络的训练过程，将船舶电机轴承故障特征向量与实际船舶电机轴承故障的标签信息作为诊断模型的输入，输出船舶电机轴承故障类别。实验结果显示该方法能够准确提取轴承故障振动信号，故障诊断精度高达99.7%。
In order to improve the accuracy of ship click bearing fault diagnosis and ensure the safety of ship navigation, the application of anti neural network algorithm in ship motor bearing fault diagnosis is studied. The energy entropy of each inherent modal function component of the vibration signal of ship motor bearing is obtained by using the set empirical mode decomposition method, and the false components are removed by using the correlation coefficient method. The screened effective data are used as the fault characteristics of ship motor bearing. The bearing fault diagnosis model is constructed by using the antagonistic neural network. The uncertainty problem in the model training process is solved by introducing the conditional antagonistic loss function. The training process of the antagonistic neural network can be described by using the two person zero sum game problem. The characteristic vector of the ship motor bearing fault and the label information of the actual ship motor bearing fault are used as the input of the diagnosis model, output ship motor bearing fault category. The experimental results show that this method can accurately extract the bearing fault vibration signal, and the fault diagnosis accuracy is as high as 99.7%.
2022,44(12): 108-111 收稿日期：2022-01-21
DOI：10.3404/j.issn.1672-7649.2022.12.021
分类号：TM356
基金项目：江苏省高等教育教改研究立项课题（2021JSJG410）
作者简介：徐向荣(1979-)，男，硕士，副教授，研究方向为船舶轮机工程几船舶电气工程
参考文献：
[1] 李远军, 孙继炫. 基于特征提取与识别两阶段的汽车电机轴承故障诊断[J]. 电子测量与仪器学报, 2019, 33(2): 56–63
[2] 肖雄, 王健翔, 张勇军, 等. 一种用于轴承故障诊断的二维卷积神经网络优化方法[J]. 中国电机工程学报, 2019, 39(15): 4558–4568
[3] 张习习, 顾幸生. 基于集成学习概率神经网络的电机轴承故障诊断[J]. 华东理工大学学报:自然科学版, 2020, 46(1): 68–76
[4] 李兵, 韩睿, 何怡刚, 等. 改进随机森林算法在电机轴承故障诊断中的应用[J]. 中国电机工程学报, 2020, 40(4): 1310–1319+1422
