基于改進(jìn)bp神經(jīng)網(wǎng)絡(luò)的圍巖自穩(wěn)能力評估模型

《基于改進(jìn)bp神經(jīng)網(wǎng)絡(luò)的圍巖自穩(wěn)能力評估模型》由會員上傳分享，免費在線閱讀，更多相關(guān)內(nèi)容在行業(yè)資料-天天文庫。

1、基于改進(jìn)BP神經(jīng)網(wǎng)絡(luò)的圍巖自穩(wěn)能力評估模型摘要:指揮防護(hù)工程是國家防護(hù)工程體系的重要組成部分。為提高其建設(shè)水平，采用改進(jìn)的前饋(bp)神經(jīng)網(wǎng)絡(luò)，對指揮防護(hù)工程圍巖自穩(wěn)能力進(jìn)行評估。結(jié)合指揮防護(hù)工程圍巖的特點，設(shè)計評估網(wǎng)絡(luò)拓?fù)浣Y(jié)構(gòu)。針對bp網(wǎng)絡(luò)原始模型的缺陷改進(jìn)，引入動量項、自適應(yīng)調(diào)節(jié)學(xué)習(xí)率、陡度因子、可變隱層節(jié)點等，并采用遺傳算法（ga）尋找最優(yōu)的初始權(quán)值和閾值。最后結(jié)合實例對算法進(jìn)行驗證。結(jié)果表明，該模型科學(xué)可靠，具有較好的工程應(yīng)用價值。關(guān)鍵詞:前饋神經(jīng)網(wǎng)絡(luò)；遺傳算法；評估；圍巖；自穩(wěn)能力；指揮防護(hù)工程self.stabilityevaluationmodelofsurr

2、oundingrockbasedonimprovedbpneuralnetworkwangduo.dian1,2*,qiuguo.qing1,daiting.ting3,wangyue11.engineeringinstituteofcorpsofengineers，plauniversityofscienceandtechnology，nanjingjiangsu210007，china;2.unit66081ofpla,huailaihebei050083,china;3.chinasatellitemaritimetrackingandcont

3、rollingdepartment,jiangyinjiangsu214431,chinaabstract:commandprotectionengineeringistheimportantcomponentofnationalprotectionengineeringsystem.toraisethelevelofconstructionofcommandprotectionengineering,thebackpropagation(bp)neuralnetworkisimprovedtogiveresearchonself-stabilityevaluationofi

4、ts’surroundingrock.firstly,thenetworktopologyisdevised，basedonthepointofsurroundingrock.secondly,themodelisimprovedaccordingtoitsdisadvantages,byintroducingthemomentum,self-adaptiveadjustinglearnrate,variablehiddennodesandsteepfactor,furthermore,geneticalgorithm(ga)isimportedtoseekitsbestin

5、itialweightandthresholdvalue.finally,beusedtoacertaincommandprotectionengineering，themodelisprovedtobecredibleandprecise.commandprotectionengineeringistheimportantcomponentofnationalprotectionengineeringsystem.toraisethelevelofconstructionofcommandprotectionengineering,thebackpropagation(bp

6、)neuralnetworkwasimprovedtogiveresearchonself.stabilityevaluationofitssurroundingrock.firstly,thenetworktopologywasdevised，basedonthecharacteristicsofsurroundingrock.secondly,themodelwasimprovedaccordingtoitsdisadvantages,byintroducingthemomentum,self.adaptiveadjustinglearnrate,variablehidd

7、ennodesandsteepfactor;furthermore,geneticalgorithm(ga)wasimportedtoseekitsbestinitialweightandthresholdvalue.finally,aninstancewasgiventovalidatethealgorithm.theresultsshowthatthemodelisscientificallyreliableandofbettervalueinengineering.keywords:backpr