基于改进DBN的发电机旋转整流器故障特征提取技术  

Generator Rotating Rectifier Fault Feature Extraction Technique Based on Improved DBN

作　　者：崔江[1] 郭瑞东 张卓然[1] 王莉[1] 孟飒飒 CUI Jiang;GUO Ruidong;ZHANG Zhuoran;WANG Li;MENG Sasa(Nanjing University of Aeronautics and Astronautics,College of Automation Engineering,Nanjing 211106,Jiangsu Province,China)

机构地区：[1]南京航空航天大学自动化学院,江苏省南京市211106

出　　处：《中国电机工程学报》2020年第7期2369-2376,共8页Proceedings of the CSEE

基　　金：国家自然科学基金项目(51777092);国家自然科学基金优秀青年科学基金项目(51622704);中央高校基本科研业务费专项资金资助(NS2017019)。

摘　　要：该文研究一种基于深度置信网络(deep belief network,DBN)的改进方法,并将其用于航空发电机旋转整流器故障特征提取及诊断操作中。首先,采集发电机交流励磁机励磁电流信号。其次,利用粒子群算法(particle swarm optimization,PSO)对深度置信网络进行训练,用于优化和确定深度置信网络的结构。最后,对所研究的方法进行仿真模型和实际平台的算法验证,并设计一个基于数字信号控制器(digital signal controller,DSC)的紧凑型实时诊断系统,成功实现了算法的移植工作,并取得了理想的诊断效果。This paper proposed a feature extraction technique based on an improved deep belief network(DBN).This method can be applied to fault feature extraction and diagnosis of rotating rectifiers in aerospace generators.Firstly,the excitation current signal of the AC exciter was collected.Secondly,the deep belief network was improved by the particle swarm optimization(PSO)algorithm to determine the network structure of the final deep belief network.Finally,the presented method was verified with simulations and physical experiments.We also designed a compact real-time diagnostic system based on the digital signal controller(DSC),which successfully realized the transplantation of the algorithm.The experiment results show that ideal diagnostic performance can be achieved with this DSC system.

关 键 词：航空发电机 旋转整流器 特征提取 深度置信网络 数字信号控制器 

分 类 号：TM461[电气工程—电器]