基于全卷积层神经网络的轴承剩余寿命预测

传统的数据驱动的轴承剩余寿命预测方法需要基于知识和经验，通过人工建立性能退化指标，费时费力，为此，采用卷积神经网络对输入信号进行特征自学习和剩余寿命预测。将传统卷积神经网络中的全连接层全部更换为卷积层与池化层，以减少神经网络需训练的参数；采用加权平均方法对预测结果进行降噪处理。轴承加速寿命实验数据集验证了所提方法的有效性。

Bearing Remaining Life Prediction Based on Full Convolutional Layer Neural Networks

Traditional method of data-driven bearing remaining life prediction was based on knowledge and experience, and the degradation index was established manually which was time-consuming and labor-intensive. Therefore, the convolutional neural network (CNN) was used to perform feature self-learning and remaining life prediction. All connected layers in the traditional CNN were replaced with convolutional layers and pooling layers to reduce the training parameters of the neural network. Weighted average method was used to denoise the prediction results. Dataset of the accelerated life test of the bearings shows the effectiveness of the proposed method.