A New Method for Analysis on Geo-Stress Field Fitting

Abstract:

Geo-stresses distribution of mining areas was analyzed by means of the partial least squares regression method (PLSRM). Based on the data measured in Wushaoling tunnel ridge section, numerical simulation of initial geo-stress field around the area had been done using the commercial software called ANSYS. PLSRM was achieved through actual measurement data and calculation results of geo-stress, then fitting data was ready. Compared with the traditional method of least squares, it is suggested that, due to the advantage of principal component analysis in applying PLSRM, present method can effectively reduce regression error caused by correlation between response variables. It can more effectively reflect the features of discontinuous geo-stress distribution in comparing to the method of traditional least-squares analysis. Therefore, it can provide effective data support for the establish of geo-stress field and it is applicable in back analysis on initial geo-stress field.

• Periodical: Advanced Materials Research (Volumes 838-841)
• Pages: 2267-2274
• DOI: https://doi.org/10.4028/www.scientific.net/AMR.838-841.2267
• Online since: November 2013
• Authors: Ping Shi*, Zhang Hua Chen, Hong Guang Ji
• Keywords: Correlation, Geo-Stress Field, Partial Least Square, Regression Analysis

References

[1] Chen X M, Ren Y Q, Xu Z Y, et al. Initial geo-stress field of surrounding rock in tunnel based on partial least square method. Engineering geophysical journal, 2011, 8(6): 767.

[2] Yuan H P, Zhao K, Wang J L, et al. Research of geo-stress field inversion using multiple linear regression. Non-ferrous metal science and engineering, 2011, 2(1): 59.

[3] Zhang Y H, Wei Q, Sheng Q, et al. Geo-stress field inversion of underground plant in Da Gangshan Hydropower station. Geotechnical engineering, 2011, 32(5): 1523.

[4] Wang H W, et al. Partial least squares regression method and its application. National Defence Industry Press, 1999, 9: 200.

[5] Wold S , Trygg Johan, Berglund Anders, Antti Henrik, et al. Some recent developments in PLS modeling. Chemometrics and Intelligent Laboratory Systems, 2001 : 131.

DOI: 10.1016/s0169-7439(01)00156-3