Abstract

The aim of this article is representing of approach to solve the inverse exploration seismilogy problems with use of methods of machine learning. The two-dimensional problem of fracture placement and spacial orientation detection is cosidered in this article. To solve this problem the neural network is used. The training of network was produced by the direct exploration seismology problems with different placement and orientation of fracture solutions with use of mathematical modeling by grid-characteristic method on regular meshes. The use of such numerical method takes into consideration the characteristic physical properties of describing processes and give us possibility to construct correct algorithms on boundaries and contact boundaries in integrational domain.

中文翻译：

---

机器学习方法检测地质裂缝

摘要

本文的目的是代表一种使用机器学习方法解决反勘探地震学问题的方法。本文讨论了二维的裂缝布置和空间定向检测问题。为了解决这个问题，使用了神经网络。网络的训练是由直接探索地震学问题产生的，该问题具有不同的裂缝解决方案的位置和方向，并使用规则网格上的网格特征方法通过数学建模。这种数值方法的使用考虑到了描述过程的特征物理特性，并为我们提供了在积分域的边界和接触边界上构造正确算法的可能性。