基于性能衰退评估的轴承寿命状态识别方法研究

Bearing life state recognition method based on performance degradation evaluation

针对滚动轴承退化性能难以评估、寿命状态难以识别的难题,提出一种基于性能衰退评估的轴承寿命状态识别新方法,该方法基于卷积自编码器(convolutional autoencoder,CAE)与多维尺度分析(multidimensional scaling,MDS)算法构建轴承性能衰退指标,再根据构建指标和改进卷积神经网络(convolutional neural network,CNN)建立轴承寿命状态识别模型,实现轴承寿命状态识别。将轴承信号样本输入CAE,实现轴承寿命状态特征的自动提取与表达,再将所提取的特征通过MDS算法进行约简获得低维特征,在低维特征空间构造欧氏距离作为轴承性能衰退指标,依据指标实现轴承数据标签化。使用标签化的轴承数据训练CNN,建立轴承寿命状态识别模型。在训练过程中,为抑制过拟合,对原始训练样本进行加噪处理,为提高模型抗干扰能力,将Leaky ReLU(LReLU)函数和dropout作为激活函数。运用轴承全寿命试验数据对识别模型进行检验,通过对比验证,结果表明所提出的轴承寿命状态识别方法能更准确的实现轴承寿命状态识别。

Aiming at the problem of rolling bearing degradation performance being difficult to evaluate and its life state being difficult to recognize,a new method of bearing life state identification based on performance degradation evaluation was proposed.Based on convolutional auto-encoder(CAE)and multi-dimensional scaling(MDS)algorithm,this method was used to construct the bearing life state recognition model,and realize the bearing life state identification.Firstly,bearing signal samples were input into CAE to automatically extract and express bearing life state features.Then,the extracted features were reduced with MDS algorithm to obtain low-dimensional features.Euclidean distance was constructed in the low-dimensional feature space as the bearing performance degradation index,and the bearing data were labeled according to the index.Furthermore,a CNN was trained with the labeled bearing data to establish the bearing life state recognition model.In training process,in order to suppress over-fitting,the original training samples were added with noise.In order to improve the anti-interference ability of the model,Leaky ReLU(LReLU)function and dropout function were taken as activation functions.Finally,the bearing full life test data were used to test the bearing life state identification model.The results showed that the proposed method can more accurately realize the bearing life state identification.