金川二矿区地表移动预测分析-prediction and analysis of surface movement in jinchuan no.2 mining area.docx

摘要利用金川二矿区多年的地表岩移GPS监测数据，分析了岩移变化规律，得出了岩移具有变形量逐年递增，变形速率逐渐变大，地表变形仍在持续发展、且有向下盘发展的趋势等特征。针对传统的矿山开采沉陷理论和预测方法并不适用于像金川这样厚大、急倾斜、地质构造复杂的金属矿床的问题。本文探讨利用时间序列来对金川二矿地表岩移进行预测。利用金川二矿区多年的地表岩移GPS监测数据，建立合理的ARMA时间序列模型，进行地表岩移预测。结果表明，具有较高的预报精度。因此，时序预测方法可以弥补金属矿山岩移预测方法不足的缺憾，并可在一定程度上指导生产。此外，本文根据Elman神经网络能够逼近任意非线性函数的特点和具有反映系统动态特性的能力，建立了金属矿山开采地表移动角预测的Elman神经网络预测模型。模型输入层有七个神经元，输出层有两个神经元。从网络训练结果看，选取的因素是合理的，输入因素与输出因素之间有着密切的联系，从理论上证明了可用于预测由于开挖引起的岩层和地表移动角问题。通过对金川矿区1#矿体进行岩体移动角的预测，并与传统的经验公式计算结果进行对比，结果表明所建模型预测结果比较准确。说明Elman神经网络模型用于金属矿山开采地表移动角预测的可行性，具有很好的应用价值。关键词：金川二矿;地表岩移;ARMA模型;Elman神经网络;地表移动角ABSTRACTBasedonGPSmonitoreddataofgroundsurfacedeformationintheNo.2MiningArea,thepaperanalyzedrockmovementlaw,obtainedtheviewpointsoftherockmovementincludingthedeformationvaluewasincreasingyearbyyear,deformationratewasgraduallyaccelerating,surfacedeformationwascontinuousdeveloping,andhadthetrendofdevelopingtothefootwall.Mostoftraditionaltheoriesandforecastingmethodsofsubsidencewerebasedoncoalmining,whicharenotsuitableformetaldepositbeingofthick,steepandcomplicatedgeologicalstructure.Inviewofthis,timeseriesmethodwaschosentopredictsurfacemovementintheNo.2MiningArea,JinchuanGroupCo.Ltd..BasedonGPSmonitoreddataofgroundsurfacedeformationintheNo.2MiningArea,theARMAmodelwasestablishedandthenusedforpredictionsofdeformations.Itshowedthatthemethodwasofhigheraccuracy.Therefore,theproposedmethodcanofferasolutiontotheshortageofpredictionmethodofsurfacemovementinmetalmine,andcanguidetheexploitationtoacertainextent.Inaddition,anElmanneuralnetworkpredictionmodelwasestablishedbaseonitscharacterofwellapproachinganynonlinearcontinuousfunctionandabilitytoreflectdynamicfeaturesofthesystem,andthenemployedforpredictingsurfacemovementangleinmetalmine.Themodelhadseveninput-layersandtwooutput-layers.Thetrainedresultsshowedthattheselectedfactorswerereasonable,andthereweregoodrelationshipbetweentheinputandoutputfactors.Itwasprovedthatthemodelcouldbeusedforpredictingthescopeofstrataandsurfacemovementinducedbyminingintheories.Basedonthemodel,t