摘要
根据双桥串联可控整流电路输出故障电压具有周期性、平移性、随着控制角大小变化波形伸展的特点,按晶闸管序号进行了故障分类,共有25大类、294小类,提出了实时采样的快速傅立叶变换+概率神经网络的故障诊断方法.首先对整流输出电压实时采样值进行快速傅立叶变换得到幅值和相位,然后根据幅值信息利用概率神经网络进行故障大类识别,利用相位信息进行故障小类识别.仿真结果表明,该方法诊断结果正确、实时性好、硬件实现简单,对复杂电力电子主回路的故障诊断具有普遍适用性.
According to the double-bridge series controlled rectifier circuit＇s output fault voltage having the characteristics of periodicity and translation and the characteristic of waveform extending along with the control angle change, faults are classified by the circuit＇s number as 25 big classifications and 294 small classifications. A new fault diagnosis method is proposed based on real-time sampling and fast Fourier transform and probabilistic neural network. By fast Fourier transform with the rectified output voltage＇s real-time sampling values, the fault＇s amplitude and phase can be obtained. The big classification fault based on the amplitude information is recognized using probabilistic neural network, and the small classification fault is recognized based on the phase information. The simulation result indicates that the method has the advantages of high accuracy, real-time performance and simple hardware, and it has the universal applicability to the fault diagnosis of complex power electronic circuits.
出处
《测试技术学报》
2008年第2期171-174,共4页
Journal of Test and Measurement Technology
关键词
故障
电力电子
电压
概率神经网络
傅立叶变换
fault
power electronic
voltage
probabilistic neural network
Fourier transform
