遗传神经网络在烈度评估中的研究与应用  

Application Research of Genetic Neural Network in Seismic Intensity Evaluation

作　　者：韶丹[1] 高贞贞[2] 田勤虎[1] 张炜超 任浩 Shao Dan;Gao Zhenzhen;Tian Qinhu;Zhang Weichao;Ren Hao(Shaanxi Earthquake Agency,Xi’an 710068,China;Xi’an Jiaotong University,School of Information and Communication Engineering,Xi’an 710049,China)

机构地区：[1]陕西省地震局,西安710068 [2]西安交通大学,信息与通信工程学院,西安710049

出　　处：《震灾防御技术》2020年第4期749-756,共8页Technology for Earthquake Disaster Prevention

基　　金：中国地震局地震应急青年重点任务(CEAEDEM201915)。

摘　　要：准确判定极震区烈度是震后应急工作高效开展的重要基础。收集1966—2017年发生在中国大陆地区MS 5.0以上有详细烈度记录的地震事件322例,选取与极震区烈度有关的7个因子进行主成分分析,将提取的主成分确定为BP神经网络的输入,极震区烈度为输出,在遗传算法优化的基础上,构建用于极震区烈度预测的BP神经网络模型。结果显示,与传统模型相比,神经网络模型在预测误差分布、精度和预测结果正确率等方面都具有明显的优越性。Accurate and rapid determination of seismic intensity in meizoseismal area is an important basis for efficient post-earthquake emergency work.In this paper,322 earthquake events of MS 5.0 or more occurred in the mainland of China are collected.Seven factors related to the intensity of the epicenter are selected and principal component analysis is carried out.The extracted principal component is determined as the input of BP neural network when the intensity of the epicenter is the output of the network.Based on the optimization of genetic algorithm,a model for intensity prediction in epicentral area is constructed.Finally,the new model is compared with three traditional ones,and the results show that the neural network model constructed in this paper has obvious advantages in prediction error distribution,accuracy,as well as correctness of prediction.

关 键 词：主成分分析 遗传算法 BP神经网络 极震区烈度 模型 

分 类 号：P315.9[天文地球—地震学] TP183[天文地球—固体地球物理学]