Elsevier

Engineering Geology

Volume 96, Issues 3–4, 1 February 2008, Pages 141-158
Engineering Geology

Prediction of uniaxial compressive strength of sandstones using petrography-based models

https://doi.org/10.1016/j.enggeo.2007.10.009Get rights and content

Abstract

The uniaxial compressive strength of intact rock is the main parameter used in almost all engineering projects. The uniaxial compressive strength test requires high quality core samples of regular geometry. The standard cores cannot always be extracted from weak, highly fractured, thinly bedded, foliated and/or block-in-matrix rocks. For this reason, the simple prediction models become attractive for engineering geologists. Although, the sandstone is one of the most abundant rock type, a general prediction model for the uniaxial compressive strength of sandstones does not exist in the literature. The main purposes of the study are to investigate the relationships between strength and petrographical properties of sandstones, to construct a database as large as possible, to perform a logical parameter selection routine, to discuss the key petrographical parameters governing the uniaxial compressive strength of sandstones and to develop a general prediction model for the uniaxial compressive strength of sandstones. During the analyses, a total of 138 cases including uniaxial compressive strength and petrographic properties were employed. Independent variables for the multiple prediction model were selected as quartz content, packing density and concavo–convex type grain contact. Using these independent variables, two different prediction models such as multiple regression and ANN were developed. Also, a routine for the selection of the best prediction model was proposed in the study. The constructed models were checked by using various prediction performance indices. Consequently, it is possible to say that the constructed models can be used for practical purposes.

Introduction

Mining engineers request the uniaxial compressive strength more often than any other rock material property (Bieniawski, 1974). The second most sought property is the triaxial strength, although the uniaxial compressive strength is requested nearly nine times more often (Cargill and Shakoor, 1990). The other important intact rock property for rock engineering projects is the modulus of elasticity. Several researchers (for ex. Sachpazis, 1990, Katz et al., 2000, Yilmaz and Sendir, 2002, Sonmez et al., 2006) emphasised on the importance of the modulus of elasticity. However, high quality core samples of regular geometry are needed for the determination of the uniaxial compressive strength and the modulus of elasticity. Such cores cannot always be extracted from weak, highly fractured, thinly bedded, foliated and/or block-in-matrix rocks. To cope with this difficulty, prediction of the some mechanical properties of intact rock became a practical approach. Especially, in the early stages of the engineering designs, such prediction models are highly useful because they are cost effective and provide reproducible data.

Krynine (1948) estimated a 40% share for sandstones in all sedimentary rocks. Ronov (1968) suggested that 24% of world platforms comprises sandstones. Boggs (1993) gives a rather gross values 20–25% for sandstone abundance in stratigraphic record. Considering these abundances declared by various researchers, to encounter a type of sandstone is a usual situation during a tunnel, slope or foundation excavation. When considering these reasons, it is evident that understanding the key petrographical parameters governing the uniaxial compressive strength of sandstones and development of some prediction models for the uniaxial compressive strength of sandstones have crucial importance.

It is expected that various petrographical properties have a control on the uniaxial compressive strength of sandstones. However, a general prediction model for the uniaxial compressive strength of sandstones does not exist in the literature although some rock type-specific empirical models and strength-petrography relationships for different types of sandstones have been proposed. In addition, although there are several petrographic parameters of sandstones, a parameter selection routine for the purposes of strength prediction is not discussed in the literature. For this reason, the main goals of the present study are to discuss the key parameters influencing the uniaxial compressive strength of sandstones, to construct a database as large as possible, to develop some general prediction models for sandstones and to check the validity of the developed prediction models. For the purposes of the study, after compilation of the data from the previous studies, the Ankara sandstone was selected in the first stage in accordance with the purpose of the study, and the uniaxial compressive strength tests and the thin-section analyses were performed. Employing the results of these analyses and the data obtained from the literature, the key parameters were selected and some multiple prediction models were developed using multiple regression analyses and ANN.

Section snippets

Previous studies

The uniaxial compressive strength of sandstones is controlled by several inherent and environmental parameters. The inherent parameters can be characterized by petrographical properties. Petrographic characteristics known to affect mechanical properties of sandstones include grain size, packing density, packing proximity, degree of grain interlocking, void space, and mineral composition (Shakoor and Bonelli, 1991). However, conflicting results have been reported relating to the influence of

Laboratory studies

In this study, the sandstones collected from Ankara and its southern vicinity were used. These sandstones belong to the Karakaya Complex of Triassic age and extend from west to the east throughout the North Anatolia in patchy outcrops (Okay and Goncuoglu, 2004). Karakaya complex is supposed to have been formed in a short-lived extensional trough (Kocyigit, 1987, Genc and Yılmaz, 1995) and mainly comprises gravity-laden siliciclastics, intercalated volcanogenic rocks and reef limestones. This

Construction of database

A total of 61 samples including uniaxial compressive strength and petrographical parameters of the Ankara sandstone were prepared. In addition, a database including the data of different authors (Bell, 1978, Shakoor and Bonelli, 1991, Ulusay et al., 1994, Bell and Lindsay, 1999) was compiled to apply statistical analyses and to construct ANN model. Bell (1978) performed both the uniaxial compressive strength tests and thin-section petrographical studies on 29 samples taken from the Fell

Parameter selection routine

When constructing a prediction model having multiple inputs, the selection of input parameters has a crucial importance. Standard statistical procedures consider only redundancy among the input parameters. In this study, not only redundancy was taken into account but also a logical procedure including physical meanings of the petrographical parameters and ease of determining them were considered. The petrographical parameters were divided into two main groups such as mineralogy and microfabric.

Multiple regression analyses

Most of the problems in geology involve complex and interacting forces, which are impossible to isolate and study individually (Davis, 1973). For this reason, the multivariate regression analysis was applied for the generalized model in this study because they allow us to consider changes in several properties simultaneously.

Fahy and Guccione (1979), Ulusay et al. (1994) and Zorlu et al. (2004) developed some multiple regression equations (Eqs. (3), (4), (5), respectively) for the prediction of

Construction of artificial neural network model

In the last decade, some soft computing methods such as fuzzy inference systems and ANN models have been used in engineering geology literature to construct some multiple prediction methods for estimation of some rock parameters. den Hartog et al. (1997), Alvarez Grima and Babuska (1999), Finol et al. (2001), Gokceoglu (2002), Gokceoglu and Zorlu (2004), Nefeslioglu et al., 2003, Nefeslioglu, H.A. et al., 2006, Tutmez and Hatipoglu (2007) employed some fuzzy inference systems for predicting

Assessment of the prediction performances and selection of the best models

Total 10 prediction models (5 of them are multiple regression and 5 of them are ANN) were developed in the present study. When considering the testing data sets as well, total 20 coefficients of correlations were obtained. The high performance for the training data sets shows good learning of the prediction model while that for the testing data sets indicates good generalization ability of the models. For this reason, when selecting the best model among the models developed herein, some

Conclusions

The conclusions obtained from the present study can be drawn as follows:

Considering the previous empirical equations among the uniaxial compressive strength and petrographic parameters of sandstones, some contradictions were detected. One of the important causes of these contradictions is sourced from selection of input parameters for multiple prediction models. In this study, to eliminate this problem, a new logical parameter selection routine for petrographic parameters is introduced.

References (63)

  • KatzO. et al.

    Evaluation of mechanical rock properties using a Schmidt hammer

    International Journal of Rock Mechanics and Mining Sciences

    (2000)
  • MeulenkampF. et al.

    Application of neural networks for the prediction of the unconfined compressive strength (UCS) from Equotip hardness

    International Journal of Rock Mechanics and Mining Sciences

    (1999)
  • NefesliogluH.A. et al.

    Indirect determination of weighted joint density (wJd) by empirical and fuzzy models: Supren (Eskisehir, Turkey) marbles

    Eng. Geol.

    (2006)
  • PrikrylR.

    Assessment of rock geomechanical quality by quantitative rock fabric coefficients: limitations and possible source of misinterpretations

    Engineering Geology,

    (2006)
  • SinghV.K. et al.

    Prediction of strength properties of some schistose rocks from petrographic properties using artificial neural Networks

    International Journal of Rock Mechanics and Mining Sciences

    (2001)
  • SinghT.N. et al.

    Prediction of thermal conductivity of rock through physico-mechanical properties

    Building and Environment,

    (2007)
  • SonmezH. et al.

    Estimation of rock modulus: for intact rocks with an artificial neural network and for rock masses with a new empirical equation

    International Journal of Rock Mechanics and Mining Sciences,

    (2006)
  • TutmezB. et al.

    Spatial estimation model of porosity

    Computers and Geosciences,

    (2007)
  • UlusayR. et al.

    Prediction of engineering properties of a selected litharenite sandstone from its petrographic characteristics using correlation and multivariate statistical techniques

    Engineering Geology

    (1994)
  • YilmazI. et al.

    Correlation of Schmidt hardness with unconfined compressive strength and Young's modulus in gypsum from Sivas (Turkey)

    Engineering Geology

    (2002)
  • ZorluK. et al.

    Predicting intact rock properties of selected sandstones using petrographic thin-section data

    International Journal of Rock Mechanics and Mining Sciences

    (2004)
  • AkyurekB. et al.

    Ankara-Elmadag-Kalecik dolayinin temel jeoloji özellikleri

    Jeoloji Mühendisliği

    (1984)
  • BaheerI.

    Selection of methodology for modeling hysteresis behavior of soils using neural networks

    Journal Computer Aided Civil Infrastructure Engineering

    (2000)
  • BellF.G. et al.

    Petrographic and engineering properties of sandstones from the Sneinton Formation, Nottinghamshire, England

    Quarterly Journal of Engineering Geology

    (1998)
  • BieniawskiZ.T.

    Estimating the strength of rock materials

    Journal of the South African Institute of Mining and Metallurgy

    (1974)
  • BlattH.
  • BoggsS.

    Principles of Sedimentology and Stratigraphy

    (1993)
  • DavisJ.C.

    Statistics and Data Analysis in Geology

    (1973)
  • DemuthH. et al.

    MATLAB Version 7.3.0.267; Neural Network Toolbox for Use with Matlab

    (2005)
  • ErolO.

    Ankara güneydogusundaki Elmadagi ve çevresinin jeolojisi ve jeomorfolojisi üzerinde bir arastirma, MTA Publication, Serie

    (1956)
  • FahyM.P. et al.

    Estimating strength of sandstones using petrographic thin-section data

    Bulletin of the Association of Engineering Geologists

    (1979)
  • Cited by (405)

    View all citing articles on Scopus
    View full text