摘要
为有效滤除舰船动力电路信号内的噪声和干扰,提取出有用的故障特征,并在复杂多变的运行环境中,准确诊断各种未知故障类型,研究基于神经网络的舰船动力电路故障诊断方法。利用栈式稀疏自编码器在舰船动力电路信号内,提取有用的电路故障特征,利用其稀疏性滤除电路信号内的噪声和干扰,减少故障特征之间的冗余;通过K-means算法,优化概率神经网络结构;在优化后的概率神经网络内,输入有用的故障特征,输出舰船动力电路故障诊断结果,依据其强大的在线学习能力,提升其对未知故障诊断的适应性。实验证明该方法可有效提取舰船动力电路故障特征;在不同噪声强度的运行环境下,该方法均可精准诊断电路故障。
In order to effectively filter out the noise and interference in the signal of ship power circuit,extract the useful fault characteristics,and accurately diagnose various unknown fault types in the complex and variable operating environment,a fault diagnosis method of ship power circuit based on neural network is studied.The stack sparse auto encoder is used to extract useful circuit fault features from ship power circuit signals,and the noise and interference in circuit signals are filtered by its sparsity to reduce the redundancy between fault features.The structure of probabilistic neural network is optimized by K-means algorithm.In the optimized probabilistic neural network,useful fault features are input and the fault diagnosis results of ship power circuit are output.According to its powerful online learning ability,its adaptability to unknown fault diagnosis is improved.Experimental results show that this method can extract the fault characteristics of ship power circuit effectively.The method can accurately diagnose circuit faults under different noise intensity operating environment.
作者
霍艳飞
张福燕
HUO Yanfei;ZHANG Fuyan(Applied Technology College,Dalian Ocean University,Dalian 116300,China)
出处
《舰船科学技术》
北大核心
2024年第18期118-121,共4页
Ship Science and Technology
基金
辽宁省教育科学“十四五”规划2022年度立项课题(JG22EB040)。
