基于人工神经元网络和多特征参数的预警震级估算*

(1.云南省地震局,云南 昆明 650224; 2.昆明学院,云南 昆明 650214; 3.中国铁道科学研究院铁道科学技术研究发展中心,北京 100081; 4.中国地震局地球物理研究所,北京 100081)

地震预警; 震级估算; 人工神经元网络; 特征参数; 线性拟合

Study on Magnitude Estimation of Earthquake Early Warning Based on Various Characteristic Parameters and Artificial Neural Networks
YANG Liwei1,LIN Guoliang1,QIU Zhigang2,JIANG Wenxiang3,WANG Yushi4

(1.Yunnan Earthquake Agency,Kunming 650224,Yunnan,China )(2. Kunming Vniversity,Kunming 650214,Yunnan,China)(3. Railway Science and Technology Research and Development Centre,China Academy of Railway Sciences,Beijing 10081,China)(4. Insititute of Geophysics,China Earthquake Administration,Beijing 10081,China)

earthquake early warning; magnitude estimation; artificial neural networks; characteristic parameters; linear fitting

备注

预警震级测定是地震预警的关键技术环节之一。在满足地震预警系统时效要求的前提下,以国内现有的人工神经元网络构架为基础,考虑采用更多的特征参数,对实时持续计算确定预警地震震级的方法进行研究。通过对日本部分实际强震数据进行持续估算预警震级与实际震级间的偏差情况,对预警震级和实际震级进行线性拟合,提出对预警震级结果的修正公式,进一步完善本方法快速估算预警震级的准确程度。

The magnitude determination is the critical technology of earthquake early warning(EEW). With satisfying the timeliness requirements of the EEW system,this paper focuses on the research of magnitude real-time continuous determination considering adopting more characteristic parameters based on the current Artificial Neural Networks(ANN). We continuously calculate the deviation between estimation magnitude and the real magnitude by using partial strong motion records from Japan,and then suggest the revised model for estimation magnitude of EEW after linear fitting of those two magnitudes to further improve the accuracy of this fast magnitude estimation method.