Page 65 - 2024年第55卷第9期
P. 65

Predictionmodelofgradationevolutionconsideringparticle
                                            breakageofcarbonaceousmudstone

                                  1
                                                                       1
                                                          2
                                                1
                                                                                    1
                      FUHongyuan,YANGHaitao,WUErlu,ZENGLing,ZHONGTao,JIANGYiyun                   1
                         (1.SchoolofCivilEngineering,ChangshaUniversityofScienceandTechnology,Changsha 410114,China;
                                  2.SchoolofCivilEngineering,ShaoxingUniversity,Shaoxing 312000,China)
                  Abstract:Inordertoaccuratelypredictthechangeofparticlebreakageandgradationcurveofcarbonaceousmud
                  stoneduringimpactloading,amathematicalmodelof“impactenergyandwatercontent - crushingindex - gradation
                  distribution” ofcarbonaceousmudstonewasestablished.Firstly, byintroducingasingleparametergrading
                  equationthatcan describethecontinuousgradingcurve , themathematicalrelationship between thesingle
                  parametergradingequationandtheparticlebreakagerate B w isestablished,andtheconversionof“crushingindex -
                  gradationdistribution ”isrealized.Aseriesofimpacttestswithdifferentwatercontentandimpactenergywerecar
                  riedoutbysettingthecontinuousgradationoftwotypicalcarbonaceousmudstonefillersofinverseStypeandhyper
                  bolictype.Theresultsshowthattheparticlebreakagerateofcarbonaceousmudstoneincreaseswiththeincreaseof
                  initialwatercontentwhentheimpactenergyisconstant ,regardlessofwhethertheinitialgradationofcarbonaceous
                  mudstoneishyperbolicdistributionorinverseS - typedistribution.Whenthewatercontentisconstant ,theparticle
                  breakagerateofcarbonaceousmudstoneincreaseswiththeincreaseofimpactenergy.Whentheimpactenergyrea
                  ches6.075kJ,theparticlebreakagerateremainsbasicallyunchanged,andtheparticlecontentofeachparticle
                  grouptendstobeastableproportionrelativetotheinitialcontent.Accordingtotherelationshipbetweentheparticle
                  breakagerate B w ofcarbonaceousmudstoneintheimpactprocessandtheinitialmoisturecontentandimpactener
                  gy ,themathematicalrelationshipbetweentheparticlebreakagerateandthemoisturecontentandimpactenergyis
                  established,andtheconversionof“impactenergyandmoisturecontent - crushingindex”isrealized,Amathemat
                  icalmodelof “impactenergyandwatercontent - crushingindex - gradationparameter”wasestablishedbyusingthe
                  crushingindexasthemedium,anditwasfoundthatthemodelcanbetterreflecttheevolutionlawofthegradation
                  curveofcarbonaceousmudstoneunderdifferentimpactenergyandwatercontentconditions.
                  Keywords:carbonmudstone;impacttest;gradationequation;particlebreakage;gradationevolution

                                                                                    (责任编辑:李 娜)

              (上接第 1057页)
                 Dam deformationpredictionmodelbasedonself - adaptivetemporaldecompositionscreening

                                                           4
                                                                                                 1,2
                                                                                     1,2
                                         1,2,3
                                                                      1,2
                            1,2
                      GUYu ,SUHuaizhi ,ZHANGShuai,YAOKefu ,LIUMingkai ,QIYining
                          (1.TheNationalKeyLaboratoryofWaterDisasterPrevention,HohaiUniversity,Nanjing 210098,China;
                          2.CollegeofWaterConservancyandHydropowerEngineering,HohaiUniversity,Nanjing 210098,China;
                        3.CooperativeInnovationCenterforWaterSafetyandHydroScience,HohaiUniversity,Nanjing 210098,China;
                                4.PowerChinaKunmingEngineeringCorporationLimited,Kunming 650051,China)
                  Abstract:Highprecisionanalysisandpredictionofdamdeformationisanimportantmeanstomasterdamwork
                  ingbehavioranddiagnosedamanomalies.Aimingattheproblemssuchasinsufficientinformationfeaturemining,
                  weakgeneralizationabilityanddifficultyinaccuratepredictionofexistingmodels,greyWolfalgorithmwasused
                  tooptimizethecompleteensembleempiricalmodedecompositionwithadaptivenoisetosolvethemultidimensional
                  parametercalibrationproblem,andthresholdevaluationindexeswereusedtoretaintheeffectiveinformationfea
                  turesofdeformationtimeseriesdata.Thecross - validationmethodiscombinedwithrecursivefeatureselection
                  method,andtheoptimalfactorsubsetisselectedbymultiplelearnerstoremoveredundantfeatures,extracteffec
                  tiveinformationandenhancetheinterpretabilityofthemodel.Consideringthecharacteristicsoftimeseriesdata ,
                  thenumberofstepsinthetimewindowofthebidirectionallongshortterm memoryneuralnetworkisoptimized,
                  andinordertoconstructdamdeformationanalysisandpredictionmodel,severalmethodssuchasnoisereduction
                  ofdamdeformationdataandinputofoptimalfeaturefactorsareused.Theresultsshowthatthemodelhasthea
                  bilityofaccuratelyminingnonlinearinformation,andthepredictionperformancehasbeensignificantlyimproved,
                  whichcanprovidereferencefordamsafetymonitoring.
                  Keywords:damdeformationprediction;greywolfalgorithm;thresholdnoisereduction;bidirectionallongshort
                  - termmemoryneuralnetwork ;completeensembleempiricalmodedecompositionwithadaptivenoise

                                                                                    (责任编辑:韩 昆)

                —  1 0 0 —
                     7
   60   61   62   63   64   65   66   67   68   69   70