基于PCA-RBF的高校学生宿舍火灾安全评价及应用  

Fire safety assessment of college dormitory based on PCA-RBF

作　　者：张立宁 范良琼 安晶[3] 苟鹏飞 ZHANG Li-ning;FAN Liang-qiong;AN Jing;GOU Peng-fei(Architecture Engineering College,North China Institute of Science and Technology,Sanhe 065201,Hebei,China;Safety Engineering College,North China Institute of Science and Technology,Sanhe 065201,Hebei,China;Telecommunications College,North China Institute of Science and Technology,Sanhe 065201,Hebei,China)

机构地区：[1]华北科技学院建筑工程学院,河北三河065201 [2]华北科技学院安全工程学院,河北三河065201 [3]华北科技学院电信工程学院,河北三河065201

出　　处：《安全与环境学报》2021年第3期921-926,共6页Journal of Safety and Environment

基　　金：中央高校基本科研业务费项目(3142018070);中央基本科研业务费项目(3142020043);河北省自然科学基金项目(EE2017508093);河北省教育厅基金项目(Z2020117);河北省科技厅软科学项目(19456107D);河北省高等教育学会课题(GJXH2019-175)。

摘　　要：为减少我国高校学生宿舍火灾隐患,采用PCA-RBF神经网络模型进行高校学生宿舍火灾安全评价。首先,获取高校学生宿舍火灾发生的主要影响因素,建立火灾安全评价指标体系。进而引入智能化评价方法,构建了基于径向基函数神经网络(Radial Basis Function Neural Network,RBF)的高校学生宿舍火灾安全评价模型。同时,为避免评价指标的冗余性和神经网络评价中维数爆炸的局限性,引入主成分分析法(Principal Component Analysis,PCA)对指标体系进行降维处理。最后,通过实例分析验证该评价模型的可行性和有效性。结果表明,引入主成分分析方法有效解决了RBF神经网络在高校学生宿舍火灾安全评价上存在的多种局限性问题。This paper is to study how to reduce the incidence of dormitory fire in China.PCA-RBF neural network model is first used to assess the safety of university dormitory to reduce the incidence of fire accidents.Firstly,the current study utilized Accident-Causing theory to analyze the main influencing factors of college dormitory fire,and then an objective fire safety assessment system was established.The system has 4 first-level indicators and 16 second-level indicators.Secondly,an intelligent evaluation method was introduced to build a fire safety evaluation model.The model is based on a radial basis function neural network(RBF).The dimension index system was reduced by principal component analysis(PCA)to realize the scientific coupling of PCA and RBF.It avoids the redundancy of the evaluation index and overcomes the limitation of dimension disaster in neural network evaluation.Finally,the feasibility and effectiveness of the evaluation model were verified by case analysis.There are 40 groups of samples,including 35 training samples,4 test samples and 1 support sample.The results indicate that the college dormitory fire is a synergy effect.The main four influencing factors include unsafe behavior of people,unsafe state of objects,defects in management and the influence of surrounding environment.Compared the evaluation results of the PCA-RBF model with those based only on RBF neural network model,the dispersion degree of the first model is reduced by 63.43%.From this point,we can see that PCA can effectively overcome the limitations of the RBF neural network in the safety assessment of college dormitory fire.PCA-RBF model can be fully used in the comprehensive evaluation of college dormitory fire.In this case,the actual value of the sample to be tested is finally calculated as 8.7882.Therefore,the fire risk level of the sample is high.This result indicates that the fire hazard of the school building is high,and it is necessary to further strengthen the fire emergency management of the school building.This study pro

关 键 词：安全工程 高校学生宿舍 校园火灾 主成分分析 径向基函数神经网络 火灾安全评价 

分 类 号：X932[环境科学与工程—安全科学]