SHIPENGINEERING船舶工程Vol.45No.12023总第45卷，2023年第1期—91—基于改进t-SNE和RBFNN的柴油机故障诊断尚前明，黄兴烨，沈栋，朱仁杰，胡秋芳，邱天（武汉理工大学船海与能源动力工程学院，武汉430063）摘要：针对柴油机故障诊断问题，提出一种基于改进t分布的随机邻域嵌入（t-SNE）和径向基函数神经网络（RBFNN）的柴油机故障诊断方法。针对t-SNE算法对振动信号的实际降维效果不够理想的问题，进行自适应加权优化；引入遗传算法（GA）解决果蝇优化算法（FOA）陷入局部最优的问题，将GA-FOA应用于RBFNN参数选取中；采用改进后的RBFNN模型对经自适应加权t-SNE降维的数据进行故障识别。研究结果表明，改进后的算法能明显改善聚类效果，提高故障识别的正确率，具有良好的应用前景。关键词：柴油机；振动信号；故障诊断；t分布的随机邻域嵌入（t-SNE）；径向基函数神经网络（RBFNN）中图分类号：TK428文献标志码：A【DOI】10.13788/j.cnki.cbgc.2023.01.14FaultDiagnosisofDieselEngineBasedonImprovedt-SNEandRBFNNSHANGQianming,HUANGXingye,SHENDong,ZHURenjie,HUQiufang,QIUTian(SchoolofNavalArchitecture,OceanandEnergyPowerEngineering,WuhanUniversityofTechnology,Wuhan430063,China)Abstract:Aimingattheproblemofdieselenginefaultdiagnosis,afaultdiagnosismethodbasedonthecombinationofimprovedt-distributedstochasticneighborembedding(t-SNE)andradialbasisfunctionneuralnetwork(RBFNN)isproposed.Adaptiveweightingoptimizationiscarriedouttosolvetheproblemthattheeffectoft-SNEalgorithmisnotidealwhenreducingthedimensionofvibrationsignal;geneticalgorithm(GA)algorithmisintroducedtoimprovetheproblemoffruitflyoptimizationalgorithm(FOA)fallingintolocaloptimization,andGA-FOAalgorithmisappliedtotheselectionofRBFNNparameters;Thedatareducedbyadaptiveweightedt-SNEareusedforfaultidentificationwiththeimprovedRBFNNmodel.Theresultsshowthattheimprovedalgorithmsignificantlyimprovestheclusteringeffect,improvestheaccuracyoffaultidentification,andhasagoodapplicationprospect.Keywords:dieselengine;vibrationsignal;faultdiagnosis;t-distributedstochasticneighborembedding(t-SNE);radialbasisfunctionneuralnetwork(RBFNN)0引言当前，柴油机通常采用各类传感器对自身的运行状态进行实时监测，...
