基于改进蚁群神经网络的牵引逆变器故障诊断

Traction Inverter Fault Diagnosis Based on Improved Ant Colony Algorithm and Neural Network

牵引逆变器是各类地铁列车牵引传动系统的关键部件之一,在实际运行中其功率管极易发生各类故障。针对传统故障诊断方法无法准确识别相应的故障类型和故障部位的问题,基于改进蚁群神经网络对牵引逆变器功率管故障诊断方法进行了研究。通过提取牵引逆变器输出三相电压的频域故障特征作为神经网络的输入,以功率管的开路故障类型作为输出,采用改进蚁群算法训练神经网络的权值和阈值,对牵引逆变器的功率管开路故障进行了有效诊断。仿真和测试结果表明,改进蚁群算法神经网络具有较高的故障诊断准确性,收敛性好,可以快速有效地实现故障定位。

The traction inverter is the key component of traction drive system of all types of trains,which is easily to lead to all kinds of failure for the power transistors in the actual operation.While the corresponding fault type and location can't be distinguished with current diagnosis methods,the fault diagnosis method of power transistor for the traction inverter based on the improved ant colony algorithm and neural network was presented.By extracting the fault characteristics in frequency domain of the three-phase voltage of traction inverter as the input of neural network,adopting the open-circuit fault type of the power transistor as output and using the improved ant colony algorithm to train the weights and threshold values of neural network,the power tube open-circuit faults of traction inverter are effectively diagnosed.The simulation and test results show that the improved ant colony algorithm and neural network have a relatively high accuracy of fault diagnosis and a fast convergence which help to locate the faults quickly and efficiently.