基于改进多尺度残差网络的工业机器人旋转部件故障研究

Research on Rotating Parts Fault of Industrial Robots Based on Improved Multi-scale Residual Networks

针对传统残差神经网络未考虑不同尺度故障特征差异的问题。提出一种基于改进多尺度残差神经网络的工业机器人旋转部件故障诊断模型(Improved Multiscale Residual Neural Network,IMRNN)。首先使用不同大小的宽卷积层提取故障信号中的宏观特征,并将其合并为初始特征向量;其次构建多尺度自适应选择卷积块提取不同尺度的特征;然后引用残差架构到多个尺度层中提高模型的泛化性;同时应用通道注意力机制对特征进行加权融合,从而完成故障诊断。结果显示,该模型有更高的准确率和良好的抗干扰能力。

The traditional residual neural network does not consider the difference of fault characteristics at different scales.An Improved Multiscale Residual Neural Network(IMRNN) based fault diagnosis model for rotating parts of industrial robots is proposed.First,wide convolution layers of different sizes are used to extract the macro features of the fault signals and combine them into initial feature vectors.Secondly,multi-scale adaptive selection convolution blocks are constructed to extract features of different scales.Then,the residual architecture is applied to multiple scale layers to improve the generalization of the model.At the same time,the channel attention mechanism is used to perform weighted fusion of features to complete fault diagnosis.The results show that the model has higher accuracy and better anti-jamming ability.