简介：Theabsenceofsub-gridscale(SGS)motionsleadstosevereerrorsinparticlepairdynamics,whichrepresentsagreatchallengetothelargeeddysimulationofparticle-ladenturbulentflow.Inordertoaddressthisissue,datafromdirectnumericalsimulation(DNS)ofhomogenousisotropicturbulencecoupledwithLagrangianparticletrackingareusedasabenchmarktoevaluatethecorrespondingresultsoffilteredDNS(FDNS).ItisfoundthatthefilteringprocessinFDNSwillleadtoanonmonotonicvariationoftheparticlecollisionstatistics,includingradialdistributionfunction,radialrelativevelocity,andthecollisionkernel.Thepeakofradialdistributionfunctionshiftstothelarge-inertiaregionduetothelackofSGSmotions,andtheanalysisofthelocalflowstructurecharacteristicvariableatparticlepositionindicatesthatthemosteffectiveinteractionscalebetweenparticlesandfluideddiesisincreasedinFDNS.Moreover,thisscaleshiftinghasanobviouseffectontheodd-ordermomentsoftheprobabilitydensityfunctionofradialrelativevelocity,i.e.theskewness,whichexhibitsastrongcorrelationtothevarianceofradialdistributionfunctioninFDNS.Asawhole,theradialdistributionfunction,togetherwithradialrelativevelocity,cancompensatetheSGSeffectsforthecollisionkernelinFDNSwhentheStokesnumberbasedontheKolmogorovtimescaleisgreaterthan3.0.However,itstillleavesconsiderableerrorsforStk<3.0.