摘要
研究鼠龙异步电机转子故障诊断技术,先运用小波包分析技术对电机电流信号进行了特征值的提取,将信号进行3层分解,然后将噪声信号分离并对噪声信号进行FFT能量分析,根据噪声能量含量的大小来判断断条故障的严重程度。然后利用BP神经网络对电机的电流信号能量特征值分别进行训练和检验,并初步得出诊断结果。再运用D-S证据理论对BP神经网络的输出结果进行了决策层的信息融合故障诊断,并得出了最终的诊断结果。实验表明,在一定程度上采用多证据的融合能进一步提高诊断的准确度。
For the rotor fault diagnosis of squirrel induction motor, characteristic values in current signal is abstracted by wavelet package analysis technique, then signal is decomposed in three level, the noise signal is separated and analyzed by FFT energy analyses, which is used to judge the severity of broken bars. Preliminary diagnosis results can be obtained by using of the BP neural network for the training and inspection of the current signal energy characteristic values. Final result is obtained by utilizing the D-S evidence theory into the decision-making information fusion for the preliminary result. Experiment indicates that using multiply evidences fusion can improve the degree of accuracy to some extent.
出处
《华北电力大学学报（自然科学版）》
CAS
北大核心
2012年第3期53-58,共6页
Journal of North China Electric Power University：Natural Science Edition
关键词
小波
信息融合
故障诊断
神经网络
证据
wavelet
information fusion
fault diagnosis
neural networks
evidence
