Neural network for seismic shaking intensity modeling

Neural network for seismic shaking intensity modeling

Saved in:
Bibliographic Details
Date:2010
Main Authors: Lazarenko, M., Korolev, V.
Format: Article
Language:English
Published: Інститут геофізики ім. С.I. Субботіна НАН України 2010
Series:Геофизический журнал
Online Access:https://nasplib.isofts.kiev.ua/handle/123456789/101762
Tags: Add Tag
No Tags, Be the first to tag this record!
Journal Title:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Cite this:Neural network for seismic shaking intensity modeling / M. Lazarenko, V. Korolev // Геофизический журнал. — 2010. — Т. 32, № 4. — С. 84-86. — англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
id nasplib_isofts_kiev_ua-123456789-101762
record_format dspace
spelling nasplib_isofts_kiev_ua-123456789-1017622025-02-09T14:57:04Z Neural network for seismic shaking intensity modeling Lazarenko, M. Korolev, V. 2010 Article Neural network for seismic shaking intensity modeling / M. Lazarenko, V. Korolev // Геофизический журнал. — 2010. — Т. 32, № 4. — С. 84-86. — англ. 0203-3100 https://nasplib.isofts.kiev.ua/handle/123456789/101762 en Геофизический журнал application/pdf Інститут геофізики ім. С.I. Субботіна НАН України
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
collection DSpace DC
language English
format Article
author Lazarenko, M.
Korolev, V.
spellingShingle Lazarenko, M.
Korolev, V.
Neural network for seismic shaking intensity modeling
Геофизический журнал
author_facet Lazarenko, M.
Korolev, V.
author_sort Lazarenko, M.
title Neural network for seismic shaking intensity modeling
title_short Neural network for seismic shaking intensity modeling
title_full Neural network for seismic shaking intensity modeling
title_fullStr Neural network for seismic shaking intensity modeling
title_full_unstemmed Neural network for seismic shaking intensity modeling
title_sort neural network for seismic shaking intensity modeling
publisher Інститут геофізики ім. С.I. Субботіна НАН України
publishDate 2010
url https://nasplib.isofts.kiev.ua/handle/123456789/101762
citation_txt Neural network for seismic shaking intensity modeling / M. Lazarenko, V. Korolev // Геофизический журнал. — 2010. — Т. 32, № 4. — С. 84-86. — англ.
series Геофизический журнал
work_keys_str_mv AT lazarenkom neuralnetworkforseismicshakingintensitymodeling
AT korolevv neuralnetworkforseismicshakingintensitymodeling
first_indexed 2025-11-27T02:15:44Z
last_indexed 2025-11-27T02:15:44Z
_version_ 1849908004856528896
fulltext ������� ��� ����� ���� �� ������������� ��� !" # �$ %& '($ ()*) Neural network for seismic shaking intensity modeling M. Lazarenko, V. Korolev, 2010 Institute of Geophysics, National Academy of Sciences of Ukraine, Kiev, Ukraine misha@nds.org.ua The seismic zonation realizes one of major eco- nomic but most complicated tasks of seismology — numerical estimation (assessment) of seismic hazard (on the patch of earth surface), actually per- forming the prediction of shaking intensity distribu- tion, using ill-found observations and conceptual regulations. The discussed approach to the prob- lem is an attempt of filling the methodological gaps of this prediction by means of networks of artificial neurons — a powerful instrument of statistical anal- ysis, allowing to build the behavior model of shak- ing intensity, using the limited set of observed ex- amples of this behavior. A neural-net model was formed on multi-layer, feedforward, fullconnected, controlled neuron net- work, using for training the error back propagation method and the set of parameter vectors — acces- sible to us geological, physical and morphological properties of seismic waves propagation medium. For the target the macroseismic estimation of sha- king intensity was used. A highest seismic hazard is threatening Ukraine from the Romanian earthquakes of Vrancea zone. "Domestic" sources do not represent a serious dan- ger, except, possibly, the Crimea, where 1927 event and specificity of region make it a target of zona- tion. Vrancea source. The seismic waves propaga- tion media may be regarded as a filter, distorting the wave field, and the projection of elements of this media on an Earth’s surface as a digital image or raster, which is shown on Fig. 1. Every element, or pixel, is described by the vec- tor of parameters, which components are the num- ber of accessible to us descriptions of physical pro- perties of geological media, functionally related to the response of this media on excitation by seis- mic waves. Factors that govern the seismic field distortion and control a diversity of shaking intensi- ty distributions over the Earth’s surface can condi- tionally be divided into three groups: regional or deep, local, and shallow (Tab. 1). Table 1. Components of training set for the model of intensity of Ukraine ���������� ��� �� � ������� �� �� � ����� � ������������������ � ��� ��� � � �������� ! "� � � � �� � #$$���������������� ����% �� &� ���� ��� � ����� $���'(� �� �� �� )�� � (���� �� ���� ��� ���� ����� �� ���*� +��� ����(�$�� , � *� -�'���(������ ��$�.�� �� � ��/ ����������� ����� � , �, � � *����* 0�������������� ����� � ����� �� ��$ ������� (������� , �� �)� ����� +��,��/� '�� 1����� % ����� �* ������'(� 2������ *�%�%�� 3 '����(�$�� , � � � 3��,������������� �345*� ����%� A near zone, that envelopes the southwest of the country, is exposed to body waves, macroseis- mic effect of which is controlled to a considerable degree by the local inhomogeneties and source mechanism. Shaking intensity in a distant zone is governed by the surface waves, for which defining are regional features of geological structure func- tionally related to integral characteristics of the sta- tionary physical fields. Macroseismic measure- ments in metric of MSK-type scale are based on the reactions of standard sensors, which are large- ly determined by character of coupling with the Earth’s surface. The mediated characteristic of such a contact can serve for the level assessment of its degeneration under influence of endogenous and exogenous factors, determined by the horizontal and vertical relief sections. ������������� ��� !" # �$ %& '($ ()*) �+ ,-� �./�-0�1 ��/ ��� �2��- ��1/ 1� 1���-� ��� ����� �� As targets the mean values of the macroseis- mic estimations for a given pixel were considered. For Vrancea earthquakes five training sets were formed: in three of them as targets macroseismic evaluations of ’77,’86, and ’90 earthquakes were used. In fourth, for the pessimistic scenario, the worst of the first three were selected. In fifth the first three and 26 pixels of known macroseismic evalua- tions of ’40 Vrancea earthquake were integrated. As an example on Fig. 2 the izoseismal map for Ukraine of Vrancea ’86 earthquake and pessimistic scenario for XX century quakes are presented. When compiling the training set for the neural networks for simulation of shaking intensity caused by the single event, such parameters as hypocen- tral depth and magnitude are not valid as parame- ters, because they are identical for all vectors-ob- jects of training set and therefore not informative. Using the macroseismic estimations of four earth- quakes a neural model allows to integrate these parameters and, due to a capacity for generaliza- tion, to model the shaking intensity, caused by a source with characteristics, different from such, used for training, what is shown on Fig. 3 for the extreme parameters of Vrancea source known from histori- cal retrospections. Fig. 1. Raster image of Ukraine. Fig. 2. Shaking intensity distribution in Ukraine in grades of MSK- 64 scale for Vrancea1986 earthquake (a�������� � � � �� ��� �� ���� ���������� ������ ����������������� � �b). ������� ��� ����� ���� �3 ������������� ��� !" # �$ %& '($ ()*) Microzonation of Yalta. Big scale seismic zo- nation, obligatory for construction of civil and indus- trial objects in seismoactive regions, must be rea- Fig. 3. Prediction of shaking intensity distribution in Ukraine in grades of MSK-64 scale for Vrancea, M=8.1, h=150 km. Fig. 4. Prediction of shaking intensity distribution in Great Yalta in grades of MSK-64 scale for ’27 Yalta earthquake type. Parameters Units Range Magnitude Richter 3.7—6.8 Source depth km 17—22 Range km 120—170 Back azimuth Degree 18.2—26.8 Soil type Boolean 0.1 Sediment thickness m 2—20 Water table m 0—20 N-W extent Fault # N-E extent Fault # Elevation m 25—300 Tilt Degree 15—30 Exposition Degree 90—315 Shaking intensity MSK-64 scale 3.6—8.3 lized with due regard for number of factors con- cerning not only physical and mechanical prop- erties of soil, their geological and morphological characteristics but also using some standard as- sessments relating these characteristics with responses to possible seismic force. The neural network can solve this problem. Tab le 2. Components of vector parameters for Yalta neural model Based on macroseismic evaluations of Yalta ’27 and ’80 earthquakes, according to the algorithm used for Vrancea case, for parameters describing the geological medium and listed in Tab. 2, the neural network was trained and used for imitation of the shaking intensity behavior on Great Yalta terrain (Fig. 4).

Similar Items