文档名:基于神经网络的二元混合液体自燃温度预测
摘要:自燃温度(Auto-IgnitionTemperature,AIT)是防火防爆安全设计的关键临界参数之一.为解决目前多数采用试验方法测量混合物AIT费时费力且有一定危险性的问题,运用定量结构-性质关系方法,使用反向传播神经网络(BackPropagationNeuralNetwork,BPNN)和一维卷积神经网络(one-DimensionalConvolutionalNeuralNetwork,1DCNN)技术建立二元混合液体AIT预测模型.以二元混合液体的分子描述符为输入、试验测得的AIT为输出,经多种方法对模型的拟合性、稳定性和预测能力评价验证.结果表明,BPNN模型和1DCNN模型均有良好的预测能力,其均方根误差分别为4.780℃和9.603℃,拟合度与5折交叉验证拟合度差值分别为0.058和0.040,表明BPNN模型有更好的拟合能力,1DCNN模型有良好的稳定性.
Abstract:Auto-IgnitionTemperature(AIT)isoneofthecrucialparametersinthedesignoffireandexplosionsafetymeasures.However,thecurrentexperimentalmethodsusedtomeasuretheAITvaluesofmixedliquidsaretime-consuming,labor-intensive,andhazardous.ThisstudyemploystheQuantitativeStructure-PropertyRelationship(QSPR)approachandutilizesaBackPropagationNeuralNetwork(BPNN)andaone-DimensionalConvolutionalNeuralNetwork(1DCNN)toestablishapredictivemodelforAITvaluesofbinarymixedliquids.Theinputparametersoftheexperimentweremoleculardescriptorsofthebinarymixedliquids,andtheoutputparametersweretheexperimentallydeterminedAITvalues.Themodel'sfittingdegree,stability,andpredictiveabilitieswereassessedandvalidatedusingvariousmethods,followedbyadeterminationofitsapplicabilityrangeandacomprehensiveinterpretation.Accordingtotheresults,theBPNNand1DCNNmodelsinthetrainingsethaverootmeansquareerrorsof4.780℃and9.603℃,respectively.Thecorrespondingaverageabsoluteerrorsare3.775℃and7.842℃,andtheaverageabsolutepercentageerrorsare18.202%and18.488%.Thedifferencebetweenthegoodnessoffitandthe5-foldcross-validationgoodnessoffitare0.058and0.040,respectively.ThesefindingsindicatethattheBPNNmodelexhibitsexcellentfittingcapabilities,the1DCNNmodeldemonstratesgoodstability,andbothmodelsdisplaysatisfactorypredictiveabilities.Theleveragemethodisusedtodeterminethemodels'applicabilityrange,anditisfoundthattheleveragevaluesintheapplicationdomainanalysisdiagramallfellwithintheapplicablerange(withinthestandardresidualrangeof±3andtotheleftofthestandardleveragevalue).TheshapleyadditiveexplanationmethodisutilizedtoassesstheimpactofnineatomtypesontheAITvaluesofbinarymixedflammableliquids.Theresultsrevealthatbothmodelsexhibitthehighestpredictiveaccuracyforbinarymixedliquidsofalkanesandalcohols.
作者:胡双启 郭丙宇 程泽会 吴薇 Author:HUShuangqi GUOBingyu CHENGZehui WUWei
作者单位:中北大学环境与安全工程学院,太原030051中北大学软件学院,太原030051
刊名:安全与环境学报 ISTICPKU
Journal:JournalofSafetyandEnvironment
年,卷(期):2024, 24(5)
分类号:X932X937
关键词:安全工程 反传播神经网络(BPNN) 一维卷积神经网络(1DCNN) 二元混合液体 自燃温度
Keywords:safetyengineering BackPropagationNeuralNetwork(BPNN) one-DimensionalConvolutionalNeuralNetwork(1DCNN) binarymixedliquids auto-ignitiontemperature
机标分类号:TP391TP273TP183
在线出版日期:2024年6月12日
基金项目:基于神经网络的二元混合液体自燃温度预测[
期刊论文] 安全与环境学报--2024, 24(5)胡双启 郭丙宇 程泽会 吴薇自燃温度(Auto-IgnitionTemperature,AIT)是防火防爆安全设计的关键临界参数之一.为解决目前多数采用试验方法测量混合物AIT费时费力且有一定危险性的问题,运用定量结构-性质关系方法,使用反向传播神经网络(BackPropa...参考文献和引证文献
参考文献
引证文献
本文读者也读过
相似文献
相关博文
基于神经网络的二元混合液体自燃温度预测 Neural network-based prediction of auto-ignition temperature of binary mixed liquids
基于神经网络的二元混合液体自燃温度预测.pdf
- 文件大小:
- 4.4 MB
- 下载次数:
- 60
-
高速下载
|
|