基于LM-BP神经网络的地震直接经济损失快速评估方法研究*

(1.天津市地震局,天津 300201; 2.天津理工大学 理学院,天津300384; 3.天津工业大学 电子与信息工程学院,天津 300387)

地震灾害; 灾害评估; 直接经济损失; LM-BP神经网络

Research on Rapid Evaluation Method of Earthquake Direct Economic Loss based on LM-BP Neural Network
ZHAO Shida1,ZHANG Nan1,ZHANG Siwen2,SUN Xiaodong3

(1.Earthquake Administration of Tianjin Municiple, Tianjin 300201, China)(2.School of Science,Tianjin University of Technology, Tianjin 300384, China)(3.Colledge of Electronics and Information Engineering,Tianjin Polytechnic University, Tianjin 300387, China)

earthquake disaster; disaster assessment; direct economic losses; LM-BP neural network

备注

在综合考虑地震致灾因子、抗震设防因子、经济指标因子的基础上,选取地震震级、震源深度、受灾面积、受灾人口、设计基本地震加速度、人均GDP和产业机构比例等7个因素作为主要评价指标,运用神经网络分析方法,建立了基于LM-BP神经网络的地震直接经济损失评估模型。从历史地震事件中提取相关数据作为样本,并使用该样本对网络进行训练。最后对模型输出结果的误差率和模型的泛化能力进行分析,认为该模型可以有效评估地震直接经济损失,并具有较高的稳定性。

Firstly, on the basis of comprehensive considering the earthquake induced disaster factor, seismic fortification factor, economic index factor, we selected seven factors as the main evaluation indicators to evaluate the direct economic losses caused by earthquake, such as earthquake magnitude, focal depth, disaster area, affected population, seismic fortification basic acceleration, per capita GDP of disaster area, industrial structure ratio. Secondly, we constructed the model of seismic economic loss assessment based on the LM-BP neural network by using neural network analysis method. Thirdly, we extracted the relative data from historical earthquake events as the sample, and used it to training. Finally, we analyzed the error rate of the model output result and the generalization ability of the model, and concluded that the model can effectively evaluate the earthquake direct economic loss and had the high stability.