以输入电流谐波为特征量的SOM网络整流电路故障诊断方法

Fault Diagnosis Method of SOM Network Rectifier Circuit with Input Current Harmonics as Characteristic

针对故障诊断系统中最重要的故障特征提取和故障识别环节,提出了基于输入电流谐波变化,利用人工神经网络对整流电路进行故障诊断的方法。以三相桥式全控整流器为模型电路,以各类故障状态下的三相交流侧输入电流谐波数据为故障特征,选取自组织特征映射神经网络(SOM)为故障识别算法,对得到的谐波数据进行编程分析识别,通过多次学习和训练,对各类元器件故障进行精确分类,从而建立起故障诊断数据库和查询途径,完成对各类故障的检测。该算法训练速度快、诊断精度高,通过仿真和相关实验验证了理论分析的可行性和有效性。

Aiming at the most important fault feature extraction and fault identification of fauh diagnosis system, the method of fault diagnosis for rectifier circuit based on the input current harmonic is proposed using artificial neural network. Taking three-phase full bridge controlled rectifier as the circuit model, and a three-phase and side input current harmonic data as the fault feature under various fault conditions, this paper selects self-organizing feature map neural network （SOM） as the fault identification algorithm, and takes a program analysis of identification of the harmonic data obtained; through repeated learning and training, the faults of various types of components are accurately classified, so as to establish a fault diagnosis database and query ways to complete the detection of various types of faults. The training speed of the algorithm is high, and the diagnosis accuracy is high： The feasibility and effectiveness of the theoretical analysis are verified by simulation and experiment.