基于多传感器数据融合和深度残差收缩网络的轴向柱塞泵故障诊断

Fault Diagnosis of Axial Piston Pump Based on Multi-sensor Data Fusion and Deep Residual Shrinkage Network

为了解决单传感器振动信息不能全面表达柱塞泵故障特征信息的问题,提出了一种基于多传感器数据融合深度残差收缩网络学习的轴向柱塞泵故障诊断方法。首先,采用多传感器对振动信号进行采集,完善振动信号的故障特征信息。其次,针对振动信号的非平稳、非线性等特征,提出基于多元多尺度散布熵的多通道融合方法,获取一维故障特征向量,从而达到增强故障冲击特征的目的。然后,将故障特征向量输入到深度残差收缩网络模型,通过注意力机制,利用软阈值函数降低样本噪声及无关特征干扰,实现轴向柱塞泵故障特征识别。最后,通过轴向柱塞泵故障诊断试验验证所提方法的有效性。试验结果表明,该方法可有效提取振动信号的故障特征,识别正确率明显高于典型的深度学习方法。

In order to solve the problem that the vibration information of a single sensor cannot fully express the fault feature information and the early fault diagnosis of the piston pump,a method based on multi-sensor data fusion and deep residual shrinkage network is proposed.It is applied to the feature learning and fault diagnosis of the vibration signal of the axial piston pump.Firstly,multi-sensors are used to collect the vibration signal,and the multi-channel signal is obtained to improve the fault feature information of the vibration signal.Secondly,a multichannel fusion method based on multivariate multi-scale dispersion entropy is proposed.This method can extract the non-stationary and nonlinear characteristics of each channel signal,so as to enhance the fault impulses characteristics.Then,the fault feature vector is input into the deep residual shrinkage network model.Through the attention mechanism,the soft threshold function is used to reduce the sample noise and irrelevant feature interference,and the fault feature recognition of the axial piston pump is realized.Finally,the effectiveness of the proposed method is verified by an axial piston pump fault diagnosis test.The experimental results show that the method can effectively extract the fault features of the vibration signal,and the recognition accuracy is significantly higher than that of the typical deep learning method.