基于DAE-BP的炉辊轴承外圈裂纹识别

Crack Identification of Furnace Roller Bearing Outer Ring Based on DAE-BP

为提高对炉辊轴承外圈裂纹识别的准确率,提出了基于DAE-BP的炉辊轴承外圈裂纹识别方法。先通过深度自编码器(DAE)对炉辊轴承振动信号的时域指标进行特征提取并重构,然后融合重构数据与原始时域指标数据,最后利用融合数据训练BP神经网络。实验结果表明,该提出方法对炉辊轴承外圈裂纹识别的准确率达到了99.61%,优于BP诊断方法,有效提高了识别准确率。

In order to improve the accuracy of crack identification on the outer ring of furnace roll bearing,a method based on DAE-BP model was proposed.Firstly,the time-domain characteristics of the vibration signals of the roller bearings were extracted and reconstructed by deep auto-encoder.Then,the reconstructed data is fused with the original time-domain index data.Finally,the BP neural network is trained with fusion data.The experimental results show that the accuracy of the proposed method is up to 99.61%,which is superior to the BP diagnosis method and improves the accuracy effectively.