Back Analysis of Initial Ground Stress Based on Back-Propagating Neural Network

 

Ruixuan Tang*, Echuan Yan, Jingsen Cai, Kun Lv, Feifei Lv

Faculty of Engineering, China University of Geosciences, Wuhan, China
*Corresponding author, e-mail: arzkama23@126.com

 

 ABSTRACT

In this paper, the location area of an underground cavern has been studied. Ground stress measurement values of limited points in this area were obtained using hydraulic fracturing technique, typical measurement values were selected as output samples of Back-Propagating (BP) neural network model; output samples were being analyzed using 3DEC software to get corresponding input samples; a 3-layer BP neural network model contains a hidden layer was build and trained using MATLAB software; ground stress of underground cavern location area was obtained using back analysis method though trained BP neural network model. After comparative analysis between inversion values and measurement values the results were confirmed: average error of inversion values is less than the measurement values (statistical information from home and abroad); the accuracy of this method can meet the requirements.

Keywords: Ground stress inversion; BP neural network; 3DEC; Numerical simulation

Get the entire paper (pdf)    Go back to the TOC