基于运行工况和多分类支持向量机的柴油机共轨系统诊断方法黄英，王拓，裴海俊，王健，王绪（北京理工大学机械与车辆学院，北京100081）摘要：对于柴油机共轨系统计量阀复位弹簧松弛和喷油器针阀偶件磨损两类典型故障，提出了基于运行工况和多分类支持向量机的故障诊断策略.该策略考虑到运行工况对故障诊断精度的影响，根据车速将实车采集数据划分为3个子工况集.通过卡方检验和机理分析选择和故障情况相关度较高的状态参数，并使用主成分分析法提取特征参数，并根据轮廓系数筛选出对故障敏感度最高的子工况集.采用层次定比采样方法划分训练集，通过粒子群算法优化支持向量机的惩罚参数和径向基函数参数.最后通过实车测试数据集对模型进行验证，实验结果表明该方法在在突出工况的诊断正确率基本达到90%以上，满足故障诊断要求.关键词：共轨系统；故障诊断；支持向量机；工况划分；主成分分析；粒子群算法中图分类号：TK428文献标志码：A文章编号：1001-0645(2023)07-0719-07DOI：10.15918/j.tbit1001-0645.2022.179DiagnosisMethodofDieselCommonRailSystemBasedonOperatingConditionsandMulti-CategorySupportVectorMachineHUANGYing，WANGTuo，PEIHaijun，WANGJian，WANGXu（SchoolofMechanicalEngineering,BeijingInstituteofTechnology,Beijing100081,China）Abstract：Afaultdiagnosisstrategybasedonoperatingconditionsandmulti-classificationsupportvectorma-chinewasproposedfortwotypicalfaultsofdieselcommonrailsystemwithloosemeteringvalveresetspringandworninjectorneedlevalvecoupling.Thestrategywasarrangedasfollows.Firstly,theinfluenceofoperat-ingconditionsonthefaultdiagnosisaccuracywastakenintoaccountandthedatacollectedfromrealvehiclesweredividedintothreesub-conditionsaccordingtothevehiclespeed.Thenthestateparameterscorrelatedhighlywiththefaultconditionwereselectedbycardinalitytestandmechanismanalysis,andthefeatureparamet-erswereextractedusingprincipalcomponentanalysis,andthesub-conditionsetwiththehighestsensitivitytothefaultwasfilteredaccordingtothecontourcoefficientS.Ahierarchicalfixed-ratiosamplingmethodwasusedtodividethetrainingset,andthepenaltyparametercandradialbasisfunction(RBF)parametergofthesupportvectormachinewereoptimizedbytheparticleswarmalgorithm.Fin...
