基于深度学习算法的船用高频电路工作状态检测研究何一芥，王波(武汉晴川学院北斗学院，湖北武汉430204)摘要:为了提升船用高频开关电源的运行可靠性，提出基于深度学习算法的船用高频电路工作状态检测方法。采集船用高频电路工作状态信号，作为深度受限波尔兹曼机的输入，深度受限波尔兹曼机利用2层受限玻尔兹曼机，通过2次非线性映射，提取船用高频电路工作状态特征。设置所提取的高频电路工作状态特征，作为支持向量数据描述方法的输入，将输入样本映射至高维内积空间，判定样本是否存在于高维内积空间的最优超球体内，检测船用高频电路工作状态为正常或异常状态。实验结果表明，该方法可以精准检测船用高频电路工作状态，满足船舶高频开关电源的运行可靠性需求。关键词：深度学习算法；船用高频电路；工作状态检测；非线性映射；高维内积空间；最优超球体中图分类号：U665文献标识码：A文章编号：1672–7649(2023)12–0156–04doi：10.3404/j.issn.1672–7619.2023.12.031ResearchonworkingstatedetectionofmarinehighfrequencycircuitbasedondeeplearningalgorithmHEYi-jie,WANGBo(BeiDouSchool,WuhanQingchuanUniversity,Wuhan430204,China)Abstract:Adeeplearningalgorithmbasedworkingstatedetectionmethodformarinehighfrequencycircuitisstudiedtoimprovetheoperationalreliabilityofmarinehighfrequencyswitchingpowersupply.TheworkingstatesignalsofMarinehighfrequencycircuitsarecollectedandusedasinputofthedepthlimitedBoltzmannmachine.ThedepthlimitedBoltzmannmachineusesthetwo-layerlimitedBoltzmannmachinetoextracttheworkingstatecharacteristicsofmarinehighfrequencycircuitsthroughtwononlinearmapping.Theextractedworkingstatecharacteristicsofthehigh-frequencycircuitaresetastheinputofthesupportvectordatadescriptionmethod,whichmapstheinputsampletothehigh-dimensionalinnerproductspace,determineswhetherthesampleexistsintheoptimalhypersphereofthehigh-dimensionalinnerproductspace,andde-tectswhethertheworkingstateofthemarinehigh-frequencycircuitisnormalorabnormal.Experimentalresultsshowthattheproposedmethodcanaccuratelydetecttheworkingstateofmarinehigh-frequencycircuitandmeettheoperationalreli-abilityrequirementsofmarinehigh-frequencyswitchingpowersupply.Keywords:deeplear...
