深度神经网络在滚动轴承故障诊断领域得到广泛应用,而将深度神经网络应用于滑动轴承-转子系统故障诊断的研究较少。其中大多数研究假设轴承故障的训练数据与测试数据分布相同,基于该训练数据得到的神经网络能较好地对轴承故障进行描述,...深度神经网络在滚动轴承故障诊断领域得到广泛应用,而将深度神经网络应用于滑动轴承-转子系统故障诊断的研究较少。其中大多数研究假设轴承故障的训练数据与测试数据分布相同,基于该训练数据得到的神经网络能较好地对轴承故障进行描述,但当轴承转子系统的结构和工况发生变化,原神经网络就不能对故障进行准确诊断。提出一种基于改进型联合分布差异(Improved Joint Distribution Discrepancy,IM-JDD)方法的深度卷积迁移学习框架(Deep Convolutional Transfer Learning Network,DCTLN),该框架采用二维振动图像作为网络输入,通过深度卷积神经网络提取图像的可迁移特征,提出的改进型联合分布差异方法实现了不同结构及工况下滑动轴承-转子系统故障特征的迁移学习。最后在结构不同的滑动轴承-转子实验台上进行测试,结果表明,本框架在不同工况下和不同机器间对无标记故障样本具有较强的诊断能力,并优于其他竞争方法。展开更多
In view of an entire dynamic model of tilting-pad journal bearing(TPJB) in which the pads swing and vibrate along geometric direction of preload, a TPJB of elastic and damped pivots was designed and manufactured. Vibr...In view of an entire dynamic model of tilting-pad journal bearing(TPJB) in which the pads swing and vibrate along geometric direction of preload, a TPJB of elastic and damped pivots was designed and manufactured. Vibration experiments were carried out under the conditions of different rotor bending stiffness and oil supply pressure to find out the relationship between the new bearing's vibration depression effect and other dynamic parameters of the rotor. The result shows that critical amplitudes can be efficaciously reduced while system's stability can be remarkably improved by this bearing. Besides, the bearing's effect of vibration depression weakens as the rotor bending stiffness increases, but heightens it as the oil supply pressure increases.展开更多
文摘深度神经网络在滚动轴承故障诊断领域得到广泛应用,而将深度神经网络应用于滑动轴承-转子系统故障诊断的研究较少。其中大多数研究假设轴承故障的训练数据与测试数据分布相同,基于该训练数据得到的神经网络能较好地对轴承故障进行描述,但当轴承转子系统的结构和工况发生变化,原神经网络就不能对故障进行准确诊断。提出一种基于改进型联合分布差异(Improved Joint Distribution Discrepancy,IM-JDD)方法的深度卷积迁移学习框架(Deep Convolutional Transfer Learning Network,DCTLN),该框架采用二维振动图像作为网络输入,通过深度卷积神经网络提取图像的可迁移特征,提出的改进型联合分布差异方法实现了不同结构及工况下滑动轴承-转子系统故障特征的迁移学习。最后在结构不同的滑动轴承-转子实验台上进行测试,结果表明,本框架在不同工况下和不同机器间对无标记故障样本具有较强的诊断能力,并优于其他竞争方法。
基金Project(2012CB026000)supported by the National Basic Research Program of China(973 Program)
文摘In view of an entire dynamic model of tilting-pad journal bearing(TPJB) in which the pads swing and vibrate along geometric direction of preload, a TPJB of elastic and damped pivots was designed and manufactured. Vibration experiments were carried out under the conditions of different rotor bending stiffness and oil supply pressure to find out the relationship between the new bearing's vibration depression effect and other dynamic parameters of the rotor. The result shows that critical amplitudes can be efficaciously reduced while system's stability can be remarkably improved by this bearing. Besides, the bearing's effect of vibration depression weakens as the rotor bending stiffness increases, but heightens it as the oil supply pressure increases.
