文档名:低压货舱多参数火灾探测集成模型的优化选择
摘要:针对低压飞机货舱环境下单模型的烟雾探测算法不灵敏的问题,给出了一种基于二次优化选择(QuadraticOptimizationChoice,QOC)策略的集成分类模型.首先,对多种火灾特征参数包括CO体积分数、温度、湿度、红外和蓝光波长光散射功率信号、烟颗粒索特平均直径及对应增长率进行增益评估,筛选出关联度高的参数作为属性,通过特征工程和性能对候选分类器进行排序,然后采用QOC策略和软投票法集成机制确定次级分类器,最后指定多层感知器(MultilayerPerceptron,MLP)作为元分类器的模型集成方法,以提高烟雾探测模型在真实飞行环境的准确性和鲁棒性.模型性能将基于精确率、召回率和F1、F2、F3指标进行比较.结果表明,集成模型应用于60kPa低压环境烟雾探测结果优于K邻近算法(KNearestNeighbor,KNN)和MLP,对榉木和航空汽油可燃物分别具有0.9724和0.9601的分类精确率,较原始算法KNN分别提高了0.0872和0.0626,较原始算法MLP分别提高了0.0368和0.1822,集成模型具有更好的性能.
Abstract:AnensembleclassificationmodelbasedonQuadraticOptimizationChoice(QOC)isproposedtotargettheproblemofinsensitivesmokedetectioninlow-pressureaircraftcargocompartmentenvironments.Themodelleveragesacomprehensiveevaluationofmultiplefirefeatureparameters,includingCOconcentration,temperature,humidity,dual-wavelengthPTR,Sautermeandiameterandgrowthrates.ThroughtheWekaattributeevaluator,highlycorrelatedattributesareidentifiedandselectedasinputfeaturesforthesubsequentsteps.Inthefeatureengineeringstage,thecandidateclassifiersarerankedbasedontheirperformanceandsuitabilityforthegivenproblem.Tocombinethestrengthsofdifferentclassifierseffectively,asoftvotingmechanismisadopted.Thismechanismallowsthesecondaryclassifierstocontributetheirpredictionstothefinaldecision-makingprocessbasedontheirrespectiveconfidences.Therefore,theensemblemodelcanleveragethediverseperspectivesofitscomponentclassifiers,resultinginimprovedaccuracyandrobustnessinreal-flightenvironments.Themeta-classifierchosenforthemodelintegrationisMultilayerPerceptron(MLP)whoseflexibilityallowsittoadaptandgeneralizewelltovariousscenarios.Toassesstheperformanceoftheensemblemodel,precision,recall,F1,F2,andF3scoresareusedforcomparisonwithKNNandMLP.ExperimentalresultsdemonstratethattheproposedmethodoutperformsbothKNearestNeighbor(KNN)andMLPwhenappliedtosmokedetectionina60kPalow-pressureenvironment.Overall,theensembleclassificationmodelbasedonQOC,combinedwithsoftvotingandMLPasameta-classifier,offersasolutionforenhancingtheaccuracyandrobustnessofsmokedetectioninlow-pressureaircraftcargocompartments.Thisensembleclassificationmodelhas0.9724and0.9601classificationaccuracies,whichare0.0872and0.0626higherthanKNNand0.0368and0.1822higherthanMLP.Thecomprehensiveevaluationoffirefeatureparametersandthethoughtfulselectionofclassifierscontributetothemodel'sabilitytohandlecomplexitiesandvariations,makingitavaluabletoolforensuringflightsafetyandmitigatingpotentialfirehazards.
作者:邓力 吴丹丹 周进 贺元骅 刘全义 王海斌 Author:DENGLi WUDandan ZHOUJin HEYuanhua LIUQuanyi WANGHaibin
作者单位:中国民航飞行学院民航安全工程学院,四川广汉618307;民机火灾科学与安全工程四川省重点实验室,四川广汉618307中国民航飞行学院民航安全工程学院,四川广汉618307
刊名:安全与环境学报 ISTICPKU
Journal:JournalofSafetyandEnvironment
年,卷(期):2024, 24(5)
分类号:X928.7
关键词:安全工程 火灾探测 飞机货舱 多参数 集成模型
Keywords:safetyengineering firedetection aircraftcargocompartment multi-parameter ensemblemodel
机标分类号:TP277TP391.41X928.7
在线出版日期:2024年6月12日
基金项目:国家自然科学基金,德阳市科技局重点研发项目,四川省院省校合作项目,四川省重点实验室项目低压货舱多参数火灾探测集成模型的优化选择[
期刊论文] 安全与环境学报--2024, 24(5)邓力 吴丹丹 周进 贺元骅 刘全义 王海斌针对低压飞机货舱环境下单模型的烟雾探测算法不灵敏的问题,给出了一种基于二次优化选择(QuadraticOptimizationChoice,QOC)策略的集成分类模型.首先,对多种火灾特征参数包括CO体积分数、温度、湿度、红外和蓝光波长光散...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
低压货舱多参数火灾探测集成模型的优化选择 Low pressure environments multi-sensor fire detection algorithm ensemble model of quadratic optimization choice
低压货舱多参数火灾探测集成模型的优化选择.pdf
- 文件大小:
- 7.25 MB
- 下载次数:
- 60
-
高速下载
|
|