Rolling element bearings are key components of mechanical equipment. The bearing fault characteristics are a ected by the interaction in the vibration signals. The low harmonics of the bearing characteristic frequenci...Rolling element bearings are key components of mechanical equipment. The bearing fault characteristics are a ected by the interaction in the vibration signals. The low harmonics of the bearing characteristic frequencies cannot be usually observed in the Fourier spectrum. The frequency loss in the bearing vibration signal is presented through two independent experiments in this paper. The existence of frequency loss phenomenon in the low frequencies, side band frequencies and resonant frequencies and revealed. It is demonstrated that the lost frequencies are actually suppressed by the internal action in the bearing fault signal rather than the external interference. The amplitude and distribution of the spectrum are changed due to the interaction of the bearing fault signal. The interaction mechanism of bearing fault signal is revealed through theoretical and practical analysis. Based on mathematical morphology, a new method is provided to recover the lost frequencies. The multi-resonant response signal of the defective bearing are decomposed into low frequency and high frequency response, and the lost frequencies are recovered by the combination morphological filter(CMF). The e ectiveness of the proposed method is validated on simulated and experimental data.展开更多
基金Supported by National Natural Science Foundation of China(Grant Nos.51675178,51475164)
文摘Rolling element bearings are key components of mechanical equipment. The bearing fault characteristics are a ected by the interaction in the vibration signals. The low harmonics of the bearing characteristic frequencies cannot be usually observed in the Fourier spectrum. The frequency loss in the bearing vibration signal is presented through two independent experiments in this paper. The existence of frequency loss phenomenon in the low frequencies, side band frequencies and resonant frequencies and revealed. It is demonstrated that the lost frequencies are actually suppressed by the internal action in the bearing fault signal rather than the external interference. The amplitude and distribution of the spectrum are changed due to the interaction of the bearing fault signal. The interaction mechanism of bearing fault signal is revealed through theoretical and practical analysis. Based on mathematical morphology, a new method is provided to recover the lost frequencies. The multi-resonant response signal of the defective bearing are decomposed into low frequency and high frequency response, and the lost frequencies are recovered by the combination morphological filter(CMF). The e ectiveness of the proposed method is validated on simulated and experimental data.
文摘针对不同型号滚动轴承监测信号之间特征分布差异大、故障数据样本少,导致轴承故障精度低的问题,提出了一种基于改进交替迁移学习的滚动轴承故障诊断算法。为了充分发挥卷积神经网络(convolutional neural network, CNN)对二维数据优秀的特征提取能力,首先将一维振动信号转化为二维图像,输入到深度卷积神经网络中学习;其次,为了减少源域与目标域数据间的特征分布差异,提出了改进的交替迁移学习(improved alternately transfer learning, IATL),通过交替计算域间的CORAL损失函数和最大均值差异(maximum mean discrepancy, MMD)损失函数,并反向传播更新各层网络权重与偏置参数,以实现变工况、跨轴承型号和小故障样本条件下轴承特征迁移适配;最后,在全连接层使用Softmax函数对目标域数据进行故障诊断。为了验证该算法的有效性,采用凯斯西储大学(Case Western Reserve University, CWRU)的滚动轴承数据集进行了迁移试验验证。结果表明,与仅计算CORAL损失函数和MMD损失函数等算法对比可知,该算法有效地减少了领域数据之间的特征分布差异,具有较高的故障分类准确率。
