Prediction of the deformation modulus of rock masses using Artificial Neural Networks and Regression methods

Static deformation modulus is recognized as one of the most important parameters governing the behavior of rock masses. Predictive models for the mechanical properties of rock masses have been used in rock engineering because direct measurement of the properties is difficult due to time and cost constraints. Using empirical methods, the deformation modulus is estimated indirectly from classification systems. This paper presents the results of the application of Artificial Neural Networks (ANN) technique and Regression models to estimate the deformation modulus of rock masses. A database, including 224 actual measured deformation modulus, Uniaxial Compressive Strengths of the rock (UCS), and Rock Mass Rating (RMR) was established. Data were collected from different projects. To predict Em by regression, a nonlinear regression method was used. This model showed the coefficient correlation of 0.751 and mean absolute percentage error (MAPE) of 9.911%. Also a three-layer ANN was found to be optimum, with an architecture of two neurons in the input layer, four neurons in the hidden layer and one neuron in the output layer. The correlation coefficient determined for deformation modulus predicted by the ANN was 0.786 and the quantity of MAPE was 6.324%. With respect to the results obtained from the two models, the ANN technique was shown to be better than the regression model because of its higher accuracy.

[1]  Kishan G. Mehrotra,et al.  Elements of artificial neural networks , 1996 .

[2]  Z. Bieniawski.  Determining rock mass deformability: experience from case histories , 1978 .

[3]  Jian Zhao,et al.  Development of a rock mass characteristics model for TBM penetration rate prediction , 2009 .

[4]  Evert Hoek,et al.  HOEK-BROWN FAILURE CRITERION - 2002 EDITION , 2002 .

[5]  Resat Ulusay,et al.  Deformation modulus of heavily jointed–sheared and blocky greywackes by pressuremeter tests: Numerical, experimental and empirical assessments , 2008 .

[6]  Håkan Stille,et al.  Classification as a tool in rock engineering , 2003 .

[7]  William S. Gardner,et al.  Design of Drilled Piers in the Atlantic Piedmont , 1987 .

[8]  Gholamnejad Javad,et al.  Application of artificial neural networks to the prediction of tunnel boring machine penetration rate , 2010 .

[9]  Byung-Sik Chun,et al.  Indirect estimation of the rock deformation modulus based on polynomial and multiple regression analyses of the RMR system , 2009 .

[10]  Reza Rahmannejad,et al.  THE ESTIMATION OF ROCK MASS DEFORMATION MODULUS USING REGRESSION AND ARTIFICIAL NEURAL NETWORKS ANALYSIS , 2010 .

[11]  Herbert H. Einstein,et al.  Using RQD to estimate the deformation modulus of rock masses , 2004 .

[12]  Z. Bieniawski,et al.  A nonlinear deformation modulus based on rock mass classification , 1990 .

[13]  Arild Palmström,et al.  The deformation modulus of rock masses — comparisons between in situ tests and indirect estimates , 2001 .

[14]  Evert Hoek,et al.  Practical estimates of rock mass strength , 1997 .

[15]  Qiuming Gong,et al.  Tunnelling through a frequently changing and mixed ground : a case history in Singapore , 2007 .

[16]  Candan Gokceoglu,et al.  Technical Note Indirect determination of the modulus of deformation of rock masses based on the GSI system , 2004 .

[17]  C. Gokceoğlu,et al.  Estimating the deformation modulus of rock masses: a comparative study , 2003 .

[18]  Abbas Majdi,et al.  Evolving neural network using a genetic algorithm for predicting the deformation modulus of rock masses , 2010 .

[19]  E. Hoek,et al.  Empirical estimation of rock mass modulus , 2006 .