中文摘要：

为了提高无刷直流电动机中轴承故障检测的鲁棒性和可靠性,提出一种基于离散小波变换（DWT）和递归神经网络（RNN）的检测方法。通过传感器采集电机振动信号和定子电流信号,通过DWT将信号分解为6个频段,并计算各频段信号的能量作为特征。利用线性局部切空间排列算法（LLTSA）对特征进行降维,获得4个具有高分类率的特征。将特征向量作为输入,通过带有偏差单元的RNN分类器来识别故障类型。实验结果表明,在不同转速和负载下,该方法都能够准确检测出故障类型,具有可行性和有效性。

英文摘要：

In order to improve the robustness and reliability of bearing fault detection of BLDC, a detection method based on discrete wavelet transform （DWT） and recurrent neural network （RNN） was proposed. The motor vibration signal and stator current signal were collected by the sensor. The signal was decomposed into six frequency bands by DWT, and the energy of each frequency band signal was calculated as a feature. The features were reduced by using LLTSA, and four features with high classification rate were obtained. The feature vector was used as input, and the fault type was identified by RNN classifier with deviation unit. The experimental results show that the method can detect the fault type accurately and it is feasible and effective under different speed and load.