【摘要】 以三相逆变器-电动机驱动系统为例,研究基于神经网络的逆变器故障智能诊断方法。在三相逆变器易发的功率管开路故障、短路故障及短路故障后逆变器工作状态理论分析和仿真实验研究基础上,建立三相逆变器故障仿真模型。通过仿真实验获取逆变器故障状态下的电机定子电流信号数据,利用信号分段方法提取故障特征构造神经网络的学习样本,并以此训练神经网络,确定用于三相逆变器故障诊断的神经网络结构和参数。在感应电机矢量控制系统中进行仿真验证,仿真实验结果表明,该方法相对已有逆变器故障诊断方法,具有诊断迅速、可靠性高的优点,可在20ms内在线完成上述各种功率管故障的诊断。更多还原

【Abstract】 Using three-phase inverter-motor system as an example,an intelligent diagnosis approach was designed for the inverter faults based on artificial neural network.The more common faults of the inverter are single switch open-circuit,short-circuit as well as post-short-circuit conditions.Based on theoretical analysis and simulation on above fault conditions,various faulted models of the inverter were developed.The stator current signals were obtained by simulation experiments under the inverter fault conditions.These signals were segmented,diagnostic features were extracted from the signal segments.A neural network was then trained on these features.The structure and parameters of the neural network for fault diagnostics of three-phase inverter were developed.Simulation experiments were conducted on the field-oriented control(FOC) motor drive system.The results show that the proposed method is rapid,reliable comparing with the existing methods,and the diagnosis of fault mentioned above can be completed on line within 20ms or less.更多还原