Modeling Engineering-Geological Layers by k-nn and Neural Networks

Abstract

In this paper a novel approach for solving task of engineering geological layers approximation is described. The task is confined in performing smooth geological surface from the heterogenic array of points in 3-dimensional space. This approach is based on statistical learning systems, as: k-nn (k-nearest neighbors) algorithm and neural networks (multilayer perceptron) which allows to separate points belonging to different geological layers. Our method enables also modeling convex 3-dimensional intrusions. The main advantage of our approach is the possibility of the surface model scaling without increasing the calculation complexity. © Springer International Publishing Switzerland 2014.