【摘要】 针对机床主轴承的故障诊断,为解决传统方法仅由单一传感器数据分析准确性低的问题,提出基于主元小波包、递归神经网络以及振动及噪声信号多源数据融合的轴承故障诊断方法,实现对锻压机床主轴承的故障诊断。将振动及噪声传感器采集的信号,经主元小波包处理提取特征值,再利用递归神经网络进行局部故障诊断,得到不同传感器对轴承故障互相独立的故障证据,然后采用基于数据修正D-S证据理论将振动及噪声诊断结果融合,发现基于递归神经网络及数据修正D-S证据理论的诊断方法。该方法解决了单一传感器的不稳定性和局限性以及传统D-S证据理论冲突证据失效的问题,使故障诊断具备容错能力,提高了传统故障诊断的精确度。更多还原

【Abstract】 Aiming at the fault diagnosis of the main bearing of machine tool, the paper presents a fault diagnosis method based on the principal wavelet packet, recursive neural network and multi-source data fusion of the vibration and noise signals to solve the problem of low accuracy of the traditional method by sensor data analysis, and the fault diagnosis of the main bearing of the machine tool is realized. The signal characteristic values collected by vibration and noise sensors are extracted by the principal wavelet packet, and the partial fault diagnosis is carried out by using the recursive neural network, and the fault evidence of the bearing failure is obtained by different sensors. Then, the D-S evidence theory based on data revision is adopted to integrate the vibration and noise diagnosis results. This method solves the problem of the instability and the limitation of single sensor and the failure of traditional D-S evidence theory under the conflict evidences, makes diagnosis have fault tolerance ability, and improves the precision of the traditional fault diagnosis.更多还原