摘 要:以TWERD变频器为研究对象,分析了三相SPWM逆变器的工作原理和故障类型,研究了该类型变频器在正常运行和故障状态下的输出线电压波形。根据故障信息和输入及负载变化时的不同数据,对变频器输出线电压进行小波变换,提取其低频能量值作为特征向量,再利用BP神经网络建立特征向量与故障类型的映射关系,确定变频器故障桥臂和故障点。仿真结果表明经过46次训练后,诊断准确率达到96.5%以上,收敛速度快,精度高。 关键词:变频器;小波变换;BP神经网络;故障诊断
Abstract: Taking TWERD frequency converter as research object, the working principle and fault type of three-phase SPWM inverter were analyzed. Its output line voltage waveform in normal operation and fault condition was studied. According to the fault information and the different data of input and load changes, the frequency converter output line voltage waveform was decomposed by wavelet transform. The low frequency energy value is picked-up and regards it as eigenvector. The mapping relationship between eigenvector and fault type were established by BP neural network, the fault bridge and fault location of frequency converter were found. Simulation results show that the diagnosis accuracy is 96.5% after training 46 times, obtained the fast convergence speed and high precision. Key words: frequency converter; wavelet transform; BP neural network; fault diagnosis
|