To improve the accuracy of rolling bearing fault diagnosis,this paper puts forward a multi-sensor fault diagnosis method based on neural network and D-S evidence theory,and test the validity of model with three sensors monitoring data.First,two acceleration sensors and a acoustic sensor are used to collect vibration signals and noise signals of rolling bearing.Then,by using Ensemble Empirical Mode Decomposition(EEMD) decompose the vibration signals of two acceleration sensors and get each Intrinsic Mode function(IMF) component,the energy characteristics of each IMF component was extracted as the input vector of thesubnet 1 and subnet 2 respectively;meanwhile using WP(wavelet packet) extract noise signals energy spectrum feature and the result was taken as the input parameters of the subnet3;Finally,The local diagnostic results of three sub-networks are normalized processing and obtained each independent evidence,applying weighted correction method adjuste the conflict evidences and obtain the final fault diagnostic results by using D-S evidence theory to fuse the information of each evidence.The experimental results show that the method can effectively enhance the accuracy and reduce the uncertainty in rolling bearing fault diagnosis.%为提高滚动轴承故障诊断的准确性, 本文提出了一种基于多传感器神经网络和D-S证据理论的故障诊断方法, 并通过包含3个传感器的监测数据融合对模型进行了验证.首先, 利用两个加速度传感器和一个声传感器采集滚动轴承的振动信号和噪声信号.其次, 分别对两个加速度传感器的振动信号进行总体经验模态分解(Ensemble Empirical Mode Decomposition, EEMD)得到各固有模态函数(Intrinsic Mode Function, IMF)分量, 并提取各IMF分量的能量特征作为子网络1和子网络2的输入参数;同时, 对声传感器的噪声信号进行小波包分解提取各频段能量特征作为子网络3的输入参数;3个子网络的局部诊断结果归一化处理得到各自独立的证据体, 对冲突证据加权修正并运用D-S证据理论进行决策级的信息融合得出最终的故障诊断结果.实验结果表明: 该方法可有效提高滚动轴承故障诊断的准确率, 降低故障诊断的不确定性.
展开▼
