Development and application of neural network technology in solving geodynamic problems

Abstract

The present study is aimed at searching the most effective, easy-to-use method for seismic zoning, in order to consider the ongoing environmental change in real time. The study demonstrates the relevance and necessity of tackling this problem, provides characteristics and peculiarities of current microzonation methods, and discusses the problem of seismic microzoning. The paper highlights traditional and neural network approaches as two directions in solving the problem, and determines their advantages and disadvantages. In addition, the paper provides significant arguments for applying the neural network approach and defines the perspectives for the solution. An intermediate problem of geophysics is suggested to be solved using the neural network approach. The current study results involved searching networks able to provide sufficient accuracy for obtaining pictures of geological sections and conducting analysis of predicted results for a number of neural networks. Multilayer perceptron is considered to give the most reliable results. The future work is supposed to develop an algorithm for building a map of seismic microzonation by means of neural network technology.