【摘要】 针对动车组设备中滚动轴承故障预测实时性不强的问题，提出了基于增强节点更新的宽度神经网络方法。首先，采用宽度神经网络的方法对预处理之后的滚动轴承原始振动数据进行模型训练；其次，在训练过程中通过增加增强节点进行权值更新；最后，使用宽度网络对设置滑动窗口的数据进行预测并输出最终结果。动车组模拟实验台采集的滚动轴承故障数据的实验结果表明：模型训练时间得以缩短，预测时间控制在30 ms以内，达到实际工业设备预测要求；与传统深度学习相比，基于增强节点更新的宽度神经网络其预测准确性得以保障，且预测实时性优于其他方法。更多还原

【Abstract】 In order to solve the problem of poor real-time prediction of rolling bearings fault in multiple units, a width neural network method based on enhanced node update is proposed. Firstly, the width neural network is used to train the rolling bearings original vibration data after preprocessing, then the weights are updated by adding enhanced nodes in training process, finally, the width network is used to predict the data set in the sliding window and output the final results. The experimental results of the fault data from multiple units’ rolling bearings show that the training time of the model can be shortened, and the prediction time can be controlled within 30 ms, which meets the requirements of the actual industrial equipment prediction; the prediction accuracy of width neural network based on enhanced node update is also guaranteed, and the prediction real-time performance is better than other methods.更多还原