Journal of Engineering Geology, Vol. 10, No. 4, Winter 2017 3793

Multivariate Estimation of Rock Mass

Characteristics Respect to Depth Using ANFIS

Based Subtractive Clustering- Khorramabad –

Polezal Freeway Tunnels

Moosavi S.H.; Engineer of Geo-Techniques and

Excavation Devision, Imensazan Engineering Consulting

Co, Water Conveyor Tunnel of Nosud Project,

126873790194

Downloaded from jeg.khu.ac.ir at 11:32 IRST on Saturday October 28th 2017 [ DOI: 10.18869/acadpub.jeg.10.4.3793 ]

Downloaded from jeg.khu.ac.ir at 11:32 IRST on Saturday October 28th 2017 [ DOI: 10.18869/acadpub.jeg.10.4.3793 ]

Sharifzadeh M.; Faculty of Mining Eng, Mining and Metallurgy, Amirkabir University of Technology

Received: 26 Aug 2013 Revised: 30 Dec 2014

Abstract

Combination of Adoptive Network based Fuzzy Inference System (ANFIS) and subtractive clustering (SC) has been used for estimation of deformation modulus (Em) and rock mass strength (UCSm) considering depth of measurement. To do this, learning of the ANFIS based subtractive clustering (ANFISBSC) was performed firstly on 125 measurements of 9 variables such as rock mass strength (UCSm), deformation modulus (Em), depth, spacing, persistence, aperture, intact rock strength (UCSi), geomechanical rating (RMR) and elastic modulus (Ei). Then, at second phase, testing the trained ANFISBSC structure has been perfomed on 40 data measurements. Therefore, predictive rock mass models have been developed for 2-6 variables where model complexity influences the estimation accuracy. Results of multivariate simulation of rock mass for estimating UCSm and Em have shown that accuracy of the ANFISBSC method increases coincident with development of model from 2 variables to 6 variables. According to the results, 3-variable model of ANFISBSC method has general estimation of both UCSm and Em corresponding with 20% to 30% error while the results of multivariate analysis are successfully improved by 6-variable model with error of less than 3%. Also, dip of the fitted line on data point of measured and estimated UCSm and Em for 6-variable model approaches about 1 respect to 0.94 for 3- variable model. Therefore, it can be concluded that 6-variable model of

ANFISBSC gives reasonable prediction of UCSm and Em.

126873790194

Downloaded from jeg.khu.ac.ir at 11:32 IRST on Saturday October 28th 2017 [ DOI: 10.18869/acadpub.jeg.10.4.3793 ]

Keywords: ANFIS, Subtractive clustering, Rock mass carachteristics, Deformation modulus, Rock mass compressive strength, Multivariate model, KhorramabadPolezal.

Introduction

Geotechnical survey and quantification of rock mass characteristics constitutes basic proportion of construction of underground openings. Therefore, recognition of project risks and parameters, which control the designation of tunnels, is important in geological studies [1]. These studies start before tunnel construction and respect to tunneling stage, will expand in several phases by desk study, laboratory tests and field measurements. Final geotechnical surveys will continue even after the tunnel construction to modify or certify aforementioned predictions [12].

For a given Rock mass property, measurements have different values instead of the same data values; suppose, X is real value of special quantity and α is variability of the measurements; therefore, data measurements distribution (y) could be calculated as [3]:

126873790194

Equation 1 conceptually determines the uncertainty of real value (x). Utilization of models that are contained of more predictive carachteristics which simulate the rock mass could increase the precision of predictions coincident with the variability of measured data set. Variability of rock mass characteristics could be estimated by several methods in the concept of multivariate analysis. Regression analysis included by linear and nonlinear methods constitutes procedure, which has been used in different branches of science, especially for estimation of uncertainty of rock mass model. However, for the sake of imperfection of estimation and imprecision results respect to measured data, researchers are decided to find other methods to substitute the regression methods that are capable to map the designation of input data set on the output data for nonlinear behavior of data. Therefore, combination of adoptive network based fuzzy inference system (ANFIS) and subtractive clustering (SC) has utilized for estimation of UCSm and Em. This code possesses both neuro-fuzzy and subtractive clustering preferences [4].

Geology and engineering geology of KhorramabadPolezal freeway

Khorramabad-polezal freeway in the southwest of Iran with the length of 105 km connects the khorramabad city to the polezal, is substituted to the tortuous, rising and falling old route with 165 km length (Figure 1). Khorramabad-Polezal freeway has the advantage of reduction in travel time from 2:30 min to 1 hour [5]. This project includes of 12 twin tunnels, which all have the horseshoe section [5]. Geological formations of the project region are Talezang (Tz), Sarvak (Ks), Asmari (Om), Gurpi (Gu) and Ilam (Ku) having limestone, argillaceous limestone and shale rocks.

126873790194

Data survey

Preconstruction site investigation has been performed on the outcrops and slopes. Also geotechnical surveys have been continued during the construction of tunnels in tunnel route (Figure 2). Joint characteristics have been studied through the use of scan line method and categurized by ISRM suggested methods and rating.

According to the geological surveys, Uni-axial compressive strength (UCSi) and elastic modulus of rocks are determined by the using schmith hammer tests; therefore, following empirical equations are employed to transform schmith hammer rebound number to the intact rock strength and elastic modulus [6]:

126873790194

Also, Deformation modulus and rock mass strength are calculated by average of several empirical relations with the given RMR and GSI (Table 1). So, collected data have been categurized to traning and testing data sets. According to depth dependent variability of rock mass characteristics [7] trining data set of 125 measurements of 9 properties of rock mass consist of depth, spacing, persistence, aperture, UCSi, Ei, Em, UCSm and rock mass rating (RMR), constitute the space of 2 to 6 variable models of rock mass to map the structure of UCSm and Em by ANFISBSC. Then, testing data set of 40 measurements has been utilized to calculate the predictibility of trained structure for estimation of UCSm and Em.

Figure1. Geographical position of Khorramabad-Polezal freeway [8]

126873790194

Figure2. Geological survey at preconstruction stage on autcrops and slopes (right) at construction stage in tunnel route (left) [8]

Table1. Empirical equations to calculate UCSm and Em [8]

Rock mass strength

(MPa) Proposed by Deformation modulus

(GPa) Proposed by

قیمت: تومان