基于Inception-LSTM的退火窑辊道系统轴承故障诊断

Bearing Fault Diagnosis of Annealing Kiln Roller Table Systems Based on Inception-LSTM

玻璃生产线退火窑辊道系统轴承运行状态显著影响玻璃品质和生产效率,实时监测各轴承运行状态对确保退火窑系统的平稳运行具有重要意义,提出结合Inception模块和长短期神经网络(Long Short-term Memory,LSTM)的迁移诊断方法,对退火窑辊道系统中的辊道轴承和通轴轴承运行状态进行监测、诊断。首先,使用集合经验模态分解(Ensemble Empirical Mode Decomposition,EEMD)对轴承信号进行分解和重构降噪,并利用直方均衡化增强重构信号小波时频图的聚集性。然后,针对样本充足的辊道轴承,建立Inception-LSTM网络,提取多尺度特征并学习其中的时间依赖关系,实现状态诊断。再次,针对转速不同且样本量少的通轴轴承,以辊道轴承信号为源域,以通轴轴承信号为目标域,以Inception-LSTM网络为基础,使用多核最大均值差异(Multi-kernel Maximum Mean Discrepancies,MKMMD)减小分布差异,实现故障样本不平衡条件下的跨转速域不变特征提取和迁移诊断。最后,利用实验数据和实测数据验证本算法的有效性,结果表明,该方法能有效诊断出退火窑辊道系统轴承故障,且具有较高的准确率。

The operating status of bearings of the roller table system of the annealing kiln in the glass production line significantly affects the glass quality and production efficiency.Real-time monitoring of the operating status of each bearing is of important significance to maintain the smooth operation of the annealing kiln system.This paper proposes an Inception module and the long-term neural network(LSTM)combined migration diagnosis method to monitor and diagnose the operating states of the roller bearing and the through-shaft bearing in the roller table system of the annealing kiln.Firstly,ensemble empirical mode decomposition(EEMD)is used to decompose and reconstruct the bearing signal,and histogram equalization is employed to enhance the aggregation of the wavelet time-frequency map of the reconstructed signal.Then,for the roller bearing with sufficient samples,an Inception-LSTM network is established to extract multi-scale features and learn the temporal dependencies to achieve state diagnosis.Thirdly,for the through-shaft bearings with different rotational speeds and a small sample size,the roller bearing signal is used as the source domain,and the through-shaft bearing signal is used as the target domain.Based on the Inception-LSTM network,the multi-kernel maximum mean difference(MK-MMD)is used to reduce distribution differences and achieve invariant feature extraction and migration diagnosis across the rotational speed domain under the condition of unbalanced fault samples.Finally,the validity of the algorithm is verified by the experimental data and the measured data.The results show that this method can effectively diagnose the bearing fault information of the roller table system of the annealing kiln,and has a high accuracy.