基于多特征融合的脉冲功率电源软故障诊断方法研究  

Research on soft fault diagnosis method of pulsed power supply based on multi-feature fusion

作　　者：周桐宇 罗红娥[1] 顾金良[1] 夏言[1] ZHOU Tongyu;LUO Honge;GU Jinliang;XIA Yan(National Key Laboratory of Transient Physics,Nanjing University of Science and Technology,Nanjing 210094,China)

机构地区：[1]南京理工大学瞬态物理国家重点实验室,南京210094

出　　处：《兵器装备工程学报》2024年第8期10-17,44,共9页Journal of Ordnance Equipment Engineering

摘　　要：脉冲功率电源作为电热发射、电磁发射等技术的核心器件,其稳定性对整个发射系统的性能起着决定性作用。针对脉冲功率电源软故障,提出一种融合多特征的BP神经网络故障诊断方法。通过建立脉冲功率电源仿真模型,获取放电电流故障数据样本;对故障样本进行时域分析和小波分析,提取时域参数及特定频带能量,以此构建融合了多种特征的特征向量;利用遗传算法对BP神经网络的初始权重和阈值进行优化,实现对脉冲功率电源故障模式的准确识别。实验结果与其他故障诊断方法进行对比,证实了本方法的有效性。As the cornerstone component in technologies such as electrothermal and electromagnetic emission,the stability of the pulse power supply is pivotal to the performance of the entire emission system.In addressing soft faults in pulse power sources,a fault diagnosis method based on the integration of multi-feature into a Back Propagation(BP)neural network is proposed.By constructing a simulation model of the pulse power supply,we gathered discharge current fault data samples.Time-domain analysis and wavelet analysis were applied to these samples to extract time-domain parameters and the energy within specific frequency bands,thereby creating a feature vector that encapsulates a spectrum of characteristics.The genetic algorithm was utilized to optimize the initial weights and thresholds of the BP neural network,thus achieving precise recognition of the fault patterns of the pulse power supply.Comparative experiments with other diagnostic methods have corroborated the efficacy of this approach.

关 键 词：脉冲功率电源 故障诊断 小波包变换 BP神经网络 多特征融合 

分 类 号：TM832[电气工程—高电压与绝缘技术]