资源描述:
《基于案例推理技术的立磨生料粉磨过程优化设定-研究》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库。
1、OPTIMALSETTINGMETHODFORRAWMEALGRINDINGPROCESSOFVERTICALMILLBASEDONCASE-BASEDREASONINGABSTRACTGrindingprocessofrawmealwithverticalmillisanimportantstepinthenewdrycementproduction.Atthepresentstage,thereisnomoreaccuratemodelinthegrindingprocess,andrawmealfinenessstilln
2、eedofflinemanualcheck.Atthesametime,thekeycontrolvariablesofverticalmillareoftensetmanuallybytheoperator,andcannotbetimelyadjustedwiththechangesofworkingconditions,withtheresultthatitcannottimelymaketheproductionprocessindexreachtotheexpectedvalue.Therefore,modelingf
3、orgrindingprocessofrawmealwithverticalmillandtimelyoptimizethesetvaluestowardthedynamicworkingconditions,whichisofgreatsignificanceforimprovingcementrawmealquality,guaranteeingthestableequipmentoperation,andreducingtheunitpowerconsumptionforproduction.Inthispaper,wec
4、arryouttheresearchfromtwoaspectsofmodelingandoptimizationsetting.Thispaperfirstdiscussestheresearchstatusofmodelingandoptimizeofrawmealgrindingprocess,andintroducestheresearchandapplicationstatusofcase-basedreasoningandparticleswarmoptimizationalgorithm.Second,wedete
5、rminethesubsequentkeyvariablesneedingmodelingandoptimizationbasedonanalyzingindetailsthespecifictechnicalprocessandparameterrequirementsforgrindingprocess.Itintroducesthestructureofwaveletneuralnetwork(WNN)andlearningalgorithm,andbyusingthedatacollectedfromthecementp
6、lant,itbuildsindexpredictionmodelsIIIofgrindingprocessseparatelybasedonWNNandBPnetwork,andcomparestheperformanceofthesetwomodels.Third,bycombiningcase-basedreasoningandparticleswarmoptimizationalgorithm,itproposesanintelligentoptimizingsettingmethodofkeycontrolvariab
7、lesofgrindingprocess.Itusesparticleswarmoptimizationalgorithmtooptimizetheoriginalcasedataandestablishestypicalcasedatabase.Whenanewworkingconditionoccurs,bysearchingandreusingthecases,itdeterminesthesetvaluesforthecurrentworkingcondition,andloadsthesetvaluestoWNNpre
8、dictionmodeltoverifyiftheproductionindexmeetstheobjectiveexpectation.Andaccordingtothedifferencebetweenexpectedvalueandpredictedval