Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104094

Contents lists available at ScienceDirect

International Journal of
Rock Mechanics and Mining Sciences
journal homepage: www.elsevier.com/locate/ijrmms

Technical Note

A new empirical formula for evaluating uniaxial compressive strength using T


the Schmidt hammer test
Min Wanga,∗, Wen Wanb
a
School of Mechanical Engineering, Hunan University of Science and Technology, Xiangtan, China
b
School of Resource Environment and Safety Engineering, Hunan University of Science and Technology, Xiangtan, China

A R T I C LE I N FO A B S T R A C T

Keywords: The uniaxial compressive strength (UCS) of rock is an important geotechnical parameter for engineering ap-
Uniaxial compressive strength (UCS) plications. However, how to determine the UCS simply and accurately has drawn the attentions of may re-
Schmidt hammer test searchers. To date, different kinds of indirect methods have been invented to determine the UCS of rocks, and
Empirical formula among these methods, estimation of the UCS based on the Schmidt hammer rebound value (Hr) was commonly
Simulated annealing-gene expression
adopted. In this paper, an insightful analysis of the literature related to UCS estimation using the Schmidt
programming(SA-GEP)
hammer test was conducted, and three stages for the development of UCS estimation using Hr were classified.
The drawbacks and merits of different kinds of techniques were analyzed in detail. Then, a data set containing
the data for different rock types was collected from references, and to obtain an objective empirical formula, the
simulated annealing-gene expression programming (SA-GEP) method was employed to establish the correlation
between UCS and Hr. Based on the calculation results, the L-type Schmidt hammer was suggested for use in UCS
estimation, and the corresponding empirical formula was established. To confirm validity of the empirical
formula, the Schmidt hammer tests and uniaxial compressive tests were conducted separately, the experimental
results were in a good agreement with the proposed empirical formula, implying that the proposed empirical
formula can be applied in engineering practice.

1. Introduction investigations. The indirect methods for estimating the UCS of rocks
have attracted increasing attentions1,7–14 due to several advantages,
Uniaxial compressive strength (UCS) is an important mechanical such as little or no specimen preparation required for the test. Among
parameter for rock classification, rock mass engineering design and these indirect methods, the Schmidt hammer test has been widely used
rock mass stability analysis, and it can also be used to assess rock to determine the UCS of rock materials.15–25
material strength in Rock Mass Rating (RMR).1 Conventionally, the UCS The Schmidt hammer is a portable compact, lightweight, cost-ef-
of rocks can be determined by the uniaxial compressive tests, as sug- fective, and nondestructive device. The Schmidt hammer rebound test
gested by the International Society for Rock Mechanics (ISRM)2 and the was originally developed for concrete testing in the late 1940s, and for
American Standards for Testing Materials (ASTM).3 However, the many years, it has also been applied in a field testing of rock hard-
testing procedure is expensive, time-consuming, expensive, involves ness,26–32 it has been increasingly used worldwide due to its rapidity,
destructive tests,4 and requires quality machined specimens.5,6 In fact, simplicity and portability.33 The Schmidt hammer test is quite easy to
it is not always possible to extract sufficient quantities of high-quality implement because it requires less, no sample preparation and the
cores from weak, highly fractured, weathered and thinly bedded rocks. testing equipment is unsophisticated so that it can be easily applied in
Hence, the conventional method for the testing uniaxial compressive the field. The procedure for using the Schmidt hammer is as follows:
strength of rocks is limited to the physical and mineralogical properties first, the hammer should be regularly checked for calibration on a steel
only. Additionally, measurement of the UCS by using the uniaxial anvil according to the instructions provided by manufacturer. Then, the
compressive tests is not convenient for the in situ testing. Schmidt hammer is pressed against a surface of rock materials, and the
Therefore, various indirect estimation methods have been so far piston in the Schmidt hammer is then released on the plunger. Some of
developed, the UCS of rocks can be estimated from different kinds of the piston's energy is transformed into sound and heat. The remaining
measured data obtained in both laboratory tests and in situ energy represents the impact penetration resistance of the surface,


Corresponding author.
E-mail address: michaelwong307@outlook.com (M. Wang).

https://doi.org/10.1016/j.ijrmms.2019.104094
Received 27 July 2018; Received in revised form 17 December 2018; Accepted 3 September 2019
Available online 20 September 2019
1365-1609/ © 2019 Elsevier Ltd. All rights reserved.
M. Wang and W. Wan International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104094

which enables the piston to rebound. The rebound height of the piston Stage 1 (1966–1980) This stage began with the publication by Deere
is called the Schmidt hammer rebound value.34 The Schmidt hammer et al.15 who used different rock types, such as basalt, diabase, and do-
should be applied perpendicular to the core samples on five different lomite, and performed the corresponding Schmidt hammer test and the
representative points spaced by least twice the plunger's diameter, be- uniaxial compressive test. Based on the experimental data, the em-
sides, and the rebound value is illustrated by a pointer on a scale (in the pirical formula was then obtained. Because the Schmidt hammer is
10–100 range) on the side of the instrument.35 portable, compact and lightweight, the UCS can be quickly obtained by
The Schmidt hammer models are designed with different levels of using empirical formulas in situ, which is quite convenient in en-
impact energy, with the types L and N are commonly used for de- gineering practice. Therefore, estimation of UCS by using the Schmidt
termining the rock properties. The L-type Schmidt hammer (impact hammer test was gradually accepted by researchers.16,17,61,62
energy of 0.735 Nm) is commonly used for rock testing, while the N- Stage 2 (1981–1993) Owing to the merits of the Schmidt hammer
type hammer (impact energy of 2.207 Nm) is used for concrete test, this test was officially recommended by ISRM44 for obtaining the
testing.36,37 Due to the wide application of the Schmidt hammer in the hardness of rock materials, and the application of the Schmidt hammer
field of rock mechanics, the Schmidt hammer method has been ac- test was further accepted by many researchers and developed for other
cepted by the International Society for Rock Mechanics (ISRM) and the applications. An increasing number of empirical formulas for evalu-
American Society for Testing and Materials (ASTM). ating the UCS using the Schmidt hammer test were used,47,55–59,63 and
With the development of rock mechanics, it was reported that the UCS estimation using the Schmidt hammer test was widely applied,
Schmidt hammer test can be used for measuring different properties, with this method drawing intense attention.
such as the penetration degree,38–42 discontinuities in rock,43 wall Stage 3 (1994-present) With the development of computer science,
strength of rock discontinuities,44 mine-roof control,45 rock mass ex- different kinds of artificial algorithms were invented, and these tech-
cavatability classification,46 Young's modulus,15,47 and the UCS of rock niques have become increasingly attractive in many research fields
materials.48–54 The Schmidt hammer test was applied to estimate UCS because a wide range of uncertainty can be tolerated. Additionally, soft
in many studies, such as those carried out by Signgh et al.55 and Shorey computation techniques, such as the ANN, FIS, SVR, etc., were applied
et al.56 that established the correlation between UCS and the Schmidt to predict the UCS of rocks.11,13,14,79–87 These artificial algorithms
hammer rebound value for different lithological units. Haramy et al.57 promoted the development of the Schmidt hammer test for estimating
performed the Schmidt hammer test using coal blocks from different US UCS. Meanwhile, in this stage, it should be noted that additional
locations, and the Schmidt hammer test rebound value was used to parameters, such as dry density and ultrasonic P-wave velocity etc.,
predict the UCS of the rocks, while Ghose et al.58 proposed an empirical were utilized to estimate the UCS, clearly improving the accuracy of
relation between the UCS and Schmidt hammer rebound value for In- UCS estimation to some extent.
dian coal that is convenient for practical use. O'Rourke59 obtained a Despite the many improvements in the UCS estimation using the
correlation with the regression coefficient (R) of 0.60 between the UCS Schmidt hammer test, some problems that should be addressed properly
and Schmidt hammer rebound values based on the samples of sand- still remain. The drawbacks of empirical formulas and soft computation
stone, siltstone, limestone and anhydrite, and Kilic et al.60 found strong techniques should be analyzed separately. Empirical formulas have
relationships between the Schmidt hammer rebound value and the UCS been widely applied for in situ measurements due to the simplicity of
values for 19 different rock samples, including sedimentary, igneous, this approach. However, since empirical formulas were established on
and metamorphic rocks, and constructed the corresponding soft com- the basis of limited experimental data sets and limited rock types, these
puting models were constructed to predict the UCS. Obviously, the empirical formulas cannot be extensively applied in engineering prac-
methods for estimating the UCS using the Schmidt hammer test could tice. Another important factor that should be taken into consideration is
be divided into two kinds, namely, the empirical formula and the soft the empirical formula type. The empirical formula type (Table 1) is
computation techniques, both of which are useful methods of esti- chosen subjectively and ultimately determined trial and error. Fre-
mating the UCS. quently, some common function types, such as linear, exponential,
In this paper, to evaluate the UCS of rocks simply and accurately, power, and logarithmic functions, were selected, and then the best
the references related to the UCS estimation using the Schmidt hammer empirical formula was determined based on the fitting coefficient.
tests were reviewed in detail, the stages of development of UCS esti- Evidently, the trial and error method for determining the empirical
mation using the Schmidt hammer rebound value were classified, and formula type is not sufficiently scientific or objective.
the advantages and the disadvantages of the techniques used to esti- Soft computation techniques also have some limitations. First, the
mate the UCS were analyzed. Then, a large amount data set was col- soft computation techniques depend on the training data set, so if the
lected from the references, and a simulated annealing-gene expression training data set is not sufficient large, the UCS prediction will be either
programming algorithm was applied to establish objective empirical inaccurate or even be wrong. Second, it should be noted that many
formulas. Subsequently, a new empirical formula for estimating the prediction models listed in Table 2 are based on the ANN method that
UCS using the L-type Schmidt hammer test was established. has drawbacks that should be taken into consideration. On the one
Furthermore, to verify the validity of the proposed empirical formula, hand, the hidden layer in the artificial neural network is a black box,
the Schmidt hammer tests and the uniaxial compressive tests were the data processing procedure is not clear or sufficiently transparent,
performed, and the results obtained by the empirical formula were and the neural network may misunderstand the purpose of the re-
compared to the experimental data, confirming the validity of the searchers, so the prediction result should be verified frequently. On the
proposed empirical formula. other hand, the artificial neural network is a much more expensive
technique than the conventional method (empirical formulas), parti-
2. Development of Schmidt hammer tests for calculating UCS cularly with respect to the computational time. The computational time
of an ANN depends on the number of hidden layers and neurons, and
Due to the limitation of the uniaxial compressive test, the Schmidt the number of hidden layers and neurons can not be determined ea-
hammer test was commonly applied to evaluate the UCS of rocks. Many sily.88–90 Additionally, the fuzzy inference system was also frequently
attempts have been made to predict the correlation between UCS and used to predict the UCS. Fuzzy logic and fuzzy inference systems in-
Hr. To summarize, the research methods can be classified into two ca- volve too many fuzzy rules that are difficult to deal with in practical
tegories: empirical formulas (Table 1) and soft computation techniques cases where variability exits. Additionally, the soft computation tech-
(Table 2). Through analysis of Tables 1 and 2, three research stages for niques are not convenient for use in engineering practice.
the development of the Schmidt hammer tests for calculating UCS can Comparison to soft computation techniques shows that the em-
be observed. pirical formula is an easy-to-implement, time-saving technique for

2
M. Wang and W. Wan International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104094

Table 1
Empirical formulas for estimating the UCS using Hr.
Reference Rock type Schmidt hammer Empirical formula
type

Deere and Miller (1966)15 Basalt, diabase, dolomite, gneiss, granite, limestone, marble, Not given UCS = 1246Hr − 34890
quartzite, rock salt, sandstone, schist, siltstone, tuff
Aufmuth (1973)16 25 different lithology Not given UCS = 0.33(Hr ·ρ)1.35
Dearman and Irfan (1978)17 Granite L-type UCS = 0.0016Hr3.47
Beverly (1979)61 20 different rocks Not given UCS = 12.74e 0.185Hr
Kidybinski (1980)62 Different rock types and coal Not given UCS = 0.447e 0.45(Hr + 3.5) + γ
Singh et al. (1983)55 30 sedimentary units L-type UCS = 2Jr
Sheorey et al. (1984)56 20 lithological units N-type UCS = 0.4Hr − 3.6
Haramy and DeMarco 10 different lithologies L-type UCS = 0.99Hr − 0.38
(1985)57
Ghose (1986)58 Coal Not given UCS = 0.88Hr − 12.11
O'Rourk (1989)59 Sandstone, siltstone, anhydrite L-type UCS = 4.85Hr − 76.18
Sachpazis (1990)47 Carbonates L-type UCS = 4.29Hr − 67.52
Xu et al. (1990)63 Mica-schist Not given UCS = 2.98e 0.06Hr
Aggistalist and Alivizatos Gabbro, basalt Not given UCS = 1.31Hr − 2.52
(1996)18
Tugrul and Zarif (1999)64 Granite Not given UCS = 8.36Hr − 416
Katz et al. (2000)65 Chalk, limestone, marble, granite N-type UCS = 2.208e 0.067Hr
Kahraman (2001)66 Dolomite, sandstone, limestone, marble, granite, diabase, N-type UCS = 6.97e 0.014Hr
serpentine, hematite
51
Yilmaz and Sendir (2002) Gypsum Not given UCS = e (0.818 + 0.06Hr )
Yasar and Erdogan (2004)51 13 samples of various carbonate rock types Not given UCS = 0.000004Hr4.29
Dincer et al. (2004)52 Andesite, basalts, tuffs Not given UCS = 2.75Hr − 36.83
Aydin and Basu (2005)67 Granite L-type UCS = 1.45e 0.07Hr
Fener et al. (2005)68 11 different rock samples N-type UCS = 4.24e (0.059Hr )
Shalabi et al. (2007)53 Shale, anhydrite, dolomite L-type UCS = 3.201Hr − 46.59
Kilic and Teymen (2008)60 19 different rock samples N-type UCS = 0.0137Hr2.2721
Cobanoglu and Celik Sandstone, mortar samples L-type UCS = 6.59Hr − 212.63
(2008)54
Gupta (2009)69 Granite Not given UCS = 1.15Hr − 15
Yagiz (2009)35 Carbonates, metamorphic Not given UCS = 0.28Hr2.584
Torabi et al. (2010)22 Siltstone, sandstone, shale, argyle L-type UCS = 0.0465Hr − 0.1756PLI + 27.682
Minaeian and Ahangari Stuff, limestone, marl L-type UCS = 0.678HrUCS = 0.056Vp + 0.315Hr − 0.46
(2013)70
Bruno et al. (2013)71 Sedimentary carbonate rocks Not given UCS = e 2.28Hr − 4.04
Saptono et al. (2013)72 Wlarukin formation sandstone, mudstone (Turkey) L-type UCS = 0.308Hr1.327
Kadir and Kesimal (2015)73 Sedimentary rocks, igneous rocks L-type UCS = 0.1383Hr1.743
UCS = 0.097Hr1.8776
UCS = 4.2423Hr − 81.92
Armaghani et al. (2016)74 Granite, metamorphic, sedimentary rocks L-type UCS = 11.42e 0.0297Hr + 0.001Vp1.178 + 22.297Is(50) − 35.051
Liang et al. (2016)75 Sandstone Not given UCS = 43.36DD + 11.161Is(50) + 1.039Hr − 112.46
Azimian (2017)76 limestone L-type UCS = 2.664Hr − 35.22UCS = 1.530Hr + 0.11Vp − 24.673
Hebib et al. (2017)77 limestone, sandstone, Dolomite, Calcareous tuff L-type UCS = 2.855e 0.0632Hr
Kong and Shang (2018)78 Magnesian limesonte, woodkirk sandstone L-type UCS = 1.80 × 10−5Hr − 5.5(L− type)
N-type
UCS = 0.30Hr1.43 (N− type)

UCS : uniaxial compressive strength; Hr : Schmidt hammer rebound value; Vp : ultrasonic P-wave velocity; DD : dry density; PLI : point load index; γ : unit weight.

determining the UCS, but the limitations of the empirical formulas expression objectively based on the data set.
should be addressed properly. Hence, in this paper, a soft computing The gene expression programming (GEP) algorithm was invented by
technique, namely, simulated annealing-gene expression programming, Ferreira in 2001.91,92 The GEP is a genotype-phenotype genetic algo-
was used to determine the empirical formulas objectively based on a rithm that has the simplicity of the genetic algorithm and the cap-
large data set. abilities of genetic programming,93 while overcoming the GEP the
drawbacks of the genetic algorithm and genetic programming ap-
3. Empirical formula determination using simulated annealing- proaches.94,95 The chromosomes in GEP consist of two different kinds of
gene expression programming symbols, namely, function symbols and terminal symbols. The function
symbols can be standard mathematical operations {+, -, *,/, sqrt, exp,
3.1. Gene expression programming log, ln, sin, cos, … }, logical operations {and, or, not, … } or user-
defined functions. While, the terminal symbols can be inclusive vari-
In the process of identifying the proper empirical formula relating ables or numerical constants.96,97 In the process of transformation from
UCS and Hr, the empirical formula type can not be determined a prior, the chromosomes, the function symbols have a different number of
it should be determined based on the data set. In fact, the empirical arguments, whereas the terminal argument is zero. The chromosomes in
formula function can be treated as an algebraic expression, thus GEP are linear symbolic strings that follow the Karva language that was
transforming the problem into that of finding an appropriate algebraic invented by Ferreira.92 The Karva language is applied to read the coded
expression that can reflect the data set well. Gene expression pro- programs in the chromosomes. For example, a GEP chromosome can be
gramming (GEP) is a good approach for determining an algebraic expressed as:

3
M. Wang and W. Wan International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104094

Table 2
Soft computation techniques for predicting the UCS.
Reference Rock type Predictive model variable

79
Garret (1994) Not given ANN Hr Vp Is(50) , n
Yilmaz and Yuksek (2008)80 Gypsum ANN Hr , ne Is(50) , Id
Yilmaz and Yuksek (2009)11 Gypsum ANN, ANFIS Hr Vp Is(50) , Wc
Dehghan et al. (2010)81 Travertine samples ANN Hr Is(50)Vp , n
Monjezi et al. (2012)13 Sandstone, limestone, dolomite, granite, chalk, geneiss, siltstone, tuff, gypsum, olivine, ANN-GA Hr ρ , n
granodiorite, slate, schist, conglomerate, quartzite, gabbro and amphibolite
Yagiz et al. (2012)14 Carbonates, metamorphic ANN Hr Vpn ρ , Id
Rezaei et al. (2012)82 Diabase, gabbro, olivine, amphibolite, dolerite, granodiorite, quartzite, basalt FIS Hr ρ , n
Yurdakul and Akdas (2013)83 marble, limestone, travertine, lacustrine, Breccia ANN Hr SH , Vp
Mishra and Basu (2013)24 Granite, sandstone and schist FIS Hr Vp Is(50) , BPI
Momeni et al. (2015)84 limestone, granite PSO-ANN Hr ρVp , Is(50)
Danial Jahed Armaghani et al. Sandstone ICA-ANN Hr Vp , Is(50)
(2016)85
Fattahi (2017)86 Basalt, Metabasalt SVR-ABC Hr
Heidari et al. (2018)87 grainstone, wackestone-mudstone, boundstone, gypsum, silty marl FIS Hr BPI , Is(50) , Vp

BPI : block punch index; Is(50) : point load strength; Hr : Schmidt hammer rebound value; Vp : ultrasonic P-wave velocity; Id : slake durability index; ne : effective porosity;
n : porosity; ρ : density; SH : shore hardness; FIS: fuzzy inference system; ANN: artificial neural network; PSO: particle swarm optimization; ICA: imperialist com-
petitive algorithm; ANFIS: adaptive neuro-fuzzy inference system; SVR: support vector regression; ABC: artificial bee colony algorithm; GA: genetic algorithm.

Q∗ + −a b c d (1)

where “Q” represents the square root function in the GEP individual, ‘*’
represents multiplication, ‘+’ represents addition, ‘-’ represents sub-
traction, and ‘a’, ‘b’, ‘c’, ‘d’ are the terminal symbols. Then, the term-
inals and functions are listed from top to bottom, and from left to right
on each line. The process is repeated until a line containing a terminal is
formed. Therefore, the chromosome (Eq. (1)) can be simply trans-
formed into an expression tree as illustrated in Fig. 1.
According to the expression tree (Fig. 1), an algebraic expression
can be obtained:

(a + b) × (c − d ) (2)

GEP have two advantages: first, GEP chromosomes can be easily


transformed into an algebraic expression, and this process is simple and
elegant; second, GEP chromosomes have fixed lengths, but the open
reading frames length may be equal to or less than the length of the GEP
chromosomes, which is convenient for the further operation in the GEP
algorithm.91
A chromosome in the GEP algorithm consists of two parts: a head
and a tail. A head can contain any function or terminal symbols,
whereas a tail part can only contain terminal symbols. The length of the Fig. 2. The flow chart of GEP.
head is an user-determined parameter, while the length of the tail is
estimated based on the following equation98,99: process is continued. The main process of the GEP algorithm is illu-
t = h (n − 1) + 1 (3) strated in Fig. 2.
As illustrated in Fig. 2, the conceptual basis of GEP is on basis of
where, t is the length of the tail, h is the length of the head, and n is the Darwin's theory of the ‘survival of the fittest’, and the main genetic
maximum arguments of the functions. In the GEP process, a certain operators in GEP consist of selection, mutation, transposition and re-
number (population size) of chromosomes will be generated in the in- combination, which are applied to the chromosomes. In the selection
itial step, and then the chromosomes are expressed as expression trees process, the individuals are selected based on the fitness and the roul-
and algebraic expressions, and their fitness is estimated based on the ette wheel sampling99 to produce the new generation. It should be
algebraic expressions. If the termination criterion is not satisfied, the mentioned that in the mutation operator, any symbol can be replaced
by another symbol. In addition, the mutation rate should be in the
range of 0.01–0.1, as was recommended by some researchers.100–103 In
the transposition, the gene in the chromosome can be moved to another
location, and the recommended transposition rate is between 0.01 and
0.1; additional explanation of transposition procedure can be found in
Ferreira's work.91,92 Another important part of GEP is recombination,
which is also called crossover, and the recombination part of GEP is the
same as that in a genetic algorithm.93

3.2. Simulated annealing-gene expression programming (SA-GEP)

Fig. 1. The expression tree. Nevertheless, GEP has two drawbacks: 1) the chromosome code in

4
M. Wang and W. Wan International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104094

GEP requires a large storage space; 2) the convergence rate is slow. The data set of UCS and Hr of rocks includes the results obtained for
Obviously, the storage of the chromosome code can hardly be changed a variety of rock types, including sandstone, limestone, marble, and
because it is determined by the GEP algorithm. Hence, the GEP algo- granite, and the sufficiently large size of the data set guarantees the
rithm can only be improved with regard to the convergence rate. effectiveness of the empirical formulas.
Adler104 suggested that changing the mutation rate in the mutation Combining the data set and SA-GEP theory, an objective empirical
process according to the simulated annealing algorithm105,106 can ac- formula will be determined. Prior to implementing gene expression
celerate the convergence of the genetic algorithm. Similarly, the GEP programming, the function set F should be determined, and in this
also contains a mutation process, hence, this improvement can also be paper, a total of eight functions were selected, as listed in Eq. (6).
applied to the GEP. According to Adler, the fitness of every individual
F = {sqrt,+,−,×,÷,exp,log,pow} (6)
can been changed as follows:
where ‘sqrt’ represents the square root function (one argument), ‘+’
e fi / T represents addition (two arguments), ‘-’ means subtraction (two argu-
fi − roulette wheel = M
∑i = 1 e fi / T (4) ments), ‘×’ represents multiplication (two arguments), ‘÷’ represents
division (two arguments), ‘exp’ represents the exponential function
where, fi is the fitness of the i th individual, fi − roulette wheel is the fitness of with the base e (one argument), ‘log’ represents the logarithmic func-
the i th individual in roulette wheel sampling99; M is the population tion (two arguments), and ‘pow’ represents the exponential function
size, and T is the temperature. The temperature T can be expressed as (two arguments).
follows. The terminal sets are as follows:
T = T0 (0.99 g − 1) (5) T = {P1, P2, P3, P4, P5, P6, P7, P8, P9, P10, P11, Hr } (7)
where T0 is the initial temperature, g is the number of evolutions, and T where ‘P1‘~ ’P11‘are the fitting parameters that must be determined when
is the temperature after g evolutions. the regression is conducted based on the algebraic expression and data
According to Adler, the mutations occur based on the simulated set, and ‘Hr’ is the Schmidt hammer rebound value. It is worth men-
annealing algorithm (SA). When the fitness of the individual after tioning that the terminal symbols in the chromosomes of GEP do not
mutation is greater than that before the mutation, the mutation will have any arguments.
occur, or, the mutation will occur at the probability of e (fi − before − fi − after )/ T . In this paper, the head length of chromosomes h is 20, and n
fi − before is the fitness of the i th individual before the mutation, and (maximum argument of all functions in head) is 40, and based on Eq.
fi − after is the fitness of the i th individual after the mutation. (3) the tail of the chromosomes is 781. Therefore, the length of the
For convenience of implementation of SA-GEP in the following, the chromosomes is 801.
flow chart of SA-GEP is illustrated in Fig. 3. Every chromosome can be decoded as an algebraic expression, then
the algebraic expression can be treated as f (Hr ) , and the UCS values
3.3. Empirical formula determined by SA-GEP can be expressed by the algebraic expression f (Hr ) .
UCS = f (Hr ) (8)
Before determining the empirical formulas, the data set of UCS and
Hr should be collected. In this paper, 1106 groups of data sets were The fitting coefficient R of Eq. (8) based on the data set (Fig. 4) was
2

collected from twenty-seven references, and the basic information selected as the fitness of the corresponding chromosome, and R2 can be
contained in data sets is listed in Table 3. calculated based on the following equation.
Because two types of the Schimidt hammer exist, to avoid the in- n
∑i = 1 (f (Hri ) − UCSi )2
fluence of the Schmidt hammer type, the collected data were divided R2 = 1 − n
∑i = 1 (UCSi − UCSmean )2 (9)
into two parts: the data determined by the L-type Schmidt hammer
(Fig. 4(a), 640 groups of data sets) and those determined using the N- where UCSi are the measured UCS values from the collected data set,
type Schmidt hammer (Fig. 4(b), 466 groups of data sets). UCSmean is the average UCS value in the data set, f (Hri ) are the pre-
dicted UCS values calculated based on the fitting of Eq. (8) and Hr in the
collected data set, and n is the size of the data set.
Furthermore, a series of parameters should be determined before
implementing the SA-GEP algorithm. In our study, the initial tem-
perature in the mutation process is set to 100, the transposition rate is
0.01, the recombination rate is 0.7, the population size is 500, these
parameters were determined based on the work of Ferreira,91,92 Du-
panloup105 and Kipkpatrick.106 The SA-GEP process is terminated when
now new best individuals were generated within the last 1000 evolu-
tion iterations.
To conveniently obtain the empirical formula based on the data set
by using SA-GEP, the process was determined as illustrated in Fig. 5.
The SA-GEP algorithm used to obtain the empirical equation was
implemented by the Python programming language that has many
merits; Python is elegant, easy to learn, and commonly used in the
artificial intelligence field.
After 4929 evolution iterations, the best results were obtained, and
the best chromosome (L-type Schmidt hammer test) in the population
that appeared in the evolution was as follows:
−/
 P1 P5 − P9 Hr Hr +P
5/
+/ ∗ − P5 + P8 HP1 
r P1……
P
11
head(length is 20) tail(length is 781) (10)
Eq. (10) can be decoded into an expression tree, which is depicted in
Fig. 3. Flow chart of SA-GEP. Fig. 6. Then the expression tree can be transformed into an algebraic

5
M. Wang and W. Wan International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104094

Table 3
The ba]sic information of the collected data sets.
Reference Rock type Schmidt hammer type Number of data sets

66
Kahraman (2001) dolomite, sandstone, limestone, marl, diabase N-type 46
Aydin and Basu (2005)67 Granite L-type 40 (L-type)
N-type 40 (N-type)
Fener et al. (2005)68 Basalt Granite, andesite, metagabro, granodiorite, quartzite, marble, limestone, travertine N-type 11
Karakus and Tutmez (2006)107 dacite, biosparitic limestone, sparitic crystallised limestone, Elazig marble, sandy sparitic L-type 9
limestone, Listwanite
Buyuksagis and Goktan (2007)108 granite, marble, limestone, Breccia L-type and N-type 27 (L-type)
27 (N-type)
53
Faisal et al. (2007) Dolomite, Dolomitic limestone L-type 35
Kilic and Teymen (2008)60 Diorite, quartzite, limestone, sandstone, granodiorite, basalt, marble, trachyte, andesite N-type 19
Cobanoglu and Celic (2008)54 Sandstone, limestone L-type 15
Torabi et al. (2010)22 Shale, argile, siltstone L-type 41
Kayabali and Selcuk (2010)109 Gypsum, sandstone, Andesite, Tuff, Ignimbrite, Marble, Limestone L-type and N-type 65 (L-type)
65 (N-type)
Yarali and Soyer (2013)110 Quartzite, limestone, Diabase, Granodiorite, Siltstone, Syenite, Dolomit, Basalt, Marl, N-type 22
Sandstone
83
Yurdakul and Akdas (2013) marble, limestone, travertine, lacustrine, Breccia N-type 10
Ataei et al. (2015)111 magnetite N-type 11
Momeni et al. (2015)84 limestone, granite L-type 66
Tandon and Gupta (2015)112 Himalayan quartzites, granite, geneisses, Himalayan metabasics, dolomite N-type 60
Karaman and Kesimal (2015)113 Basalt, Metabasalt, dacite, limesonte, sandy limestone L-type 46
Iok-Tong et al. (2015)114 granite L-type 115
Armaghani et al. (2016)74 Granite, metamorphic, sedimentary rocks L-type 71
Hebib et al. (2017)77 limestone, sandstone, Dolomite, Calcareous tuff L-type 19
Azimian (2017)76 limestone L-type 30
Akram et al. (2017)115 wackestone, mudstone, packstone L-type 37
Afolagboye et al. (2017)116 granite, migmatite gneiss, quartzite, charnokites N-type 50
Fattahi (2017)86 Basalt, Metabasalt L-type 10
Kong and Shang (2018)78 Magnesian limestone, Woodekirk sandstone type 9 (L-type)
N-type 9 (N-type)
Heidari et al. (2018)87 grainstone, wackestone-mudstone, boundstone, gypsum, silty marl N-type 115
Ghasemi et al. (2018)117 carbonate rocks L-type 1
Petrakis and Komnitsas (2018)118 quartz, marble, quartzite, metasandstone L-type 4

expression: 3.4. Empirical formula verification

P5/(P9 − Hr ) − P1 (11) To further verify the validity of the proposed empirical formula (Eq.
(13)), five kinds of rocks (granite, yellow rust granite, red sandstone,
Moreover, based on the calculation result, the value of fitting Maokou limestone and skarn) were selected, and the corresponding
parameters P5, P9 and P1 are 6222, 88.15 and 70.38, respectively. cylindrical specimens with dimensions of Φ50mm × 100mm were pre-
Hence, Eq. (11) can be simplified as: pared. Meanwhile, four specimens were prepared for each kind of rock,
for a total of 20 rock specimens.
6222
− 70.38 The HT-225B Schmidt hammer (L-type) was applied to obtain the
(88.15 − Hr ) (12)
Schmidt hammer rebound value, and the Schmidt hammer tests were
repeated fifty times for each specimen, with the average value used as
Based on the calculation results, the relation between the UCS va-
the result. Then, the uniaxial compressive tests were performed by
lues and Hr can be expressed as:
using a New SANS testing machine with a loading rate of 100 N/s. The
6222 experimental results obtained using the Schmidt hammer tests and the
UCS = − 70.38 uniaxial compressive tests are listed in Table 4.
(88.15 − Hr ) (13)
To compare the results obtained using the proposed empirical for-
The corresponding fitting coefficient (R2) is 0.6081. Meanwhile, the mula and the experimental data, the R2 value is calculated based on Eq.
calculation was also implemented using GEP, but it takes 10000 evo- (9), and the fitting coefficient R2 is found to be 0.95, indicating that the
lution iterations, hence, the use of SA-GEP is preferred. Similarly, the proposed empirical formula can reflect the experimental data well and
empirical formula for estimating UCS using the N-type Schmidt confirming the validity of the empirical formula Eq. (11). Fig. 7 illus-
hammer was established and is given by: trates the proposed empirical formula and the experimental results.

0.6549 4. Discussion
UCS = ⎛2.0798 +

⎞ × (Hr +
⎟ Hr ) − 28.02
⎝ − 4.0339Hr + 328.5 ⎠ (14)
The UCS of a rock is a critically important parameter for rock mass
The corresponding fitting coefficient is 0.4301. By comparing to the engineering stability and rock mass design, particularly when the rocks
N-type Schmidt hammer type, the empirical formula for estimating the are subjected to compressive stresses with low confining pressure.119
UCS using the L-type Schmidt hammer can clearly provide good pre- Therefore, determining the UCS accurately and simply is of critical
diction results, and the empirical formula for estimating UCS using the importance for rock mass engineering. Generally, two types of methods
L-type Schmidt hammer is much simpler than that using the N-type are used for the determination of the UCS: (a) direct laboratory tests on
Schmidt hammer. Hence, the L-type Schmidt hammer was strongly rock samples; (b) indirect estimations based on some correlated para-
recommended for predicting the UCS, which is consistent with the meters that can be obtained much more easily than the UCS itself. The
study of ISRM study.5,6 direct laboratory tests require very strict conditions for the preparation

6
M. Wang and W. Wan International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104094

Fig. 4. The data set collected from 27 references.

of the rock specimens, which are difficult and sometimes even im- the linear, exponential, power, and logarithmic functions, and then, the
possible to realize for cracked rocks. Additionally, obtaining core best empirical formula types were be determined based on the fitting
samples is costly and time-consuming because it requires a highly coefficient. Clearly, this process is not scientific, and as a result, the
skilled operator.4,8,98,120 Therefore, the indirect estimation methods of reliability and applicability of the previous empirical formulas are
the UCS of rocks have been widely used due to simplicity and non- questionable.
destructive nature. Among these techniques, the Schmidt hammer test Meanwhile, with the development of soft computation techniques,
to estimate the UCS of rock is more practical and economical. The some artificial algorithms were applied to UCS prediction. An analysis
Schmidt hammer is portable and can be used in both the laboratory and of the soft computation techniques found that a large data set is needed
the field. in the prediction models, and the computational process is time-con-
Hence, in this paper, the references related to UCS prediction using suming, while soft computation techniques are strongly dependent on
the Schmidt hammer tests were reviewed in detail, and it was found the software and computational equipment.
that the UCS estimated using the Schmidt hammer test can be divided Combining the merits of the soft computation techniques and the
into two kinds: those using the empirical formulas and those using soft empirical formulas, a soft computation technique, namely, SA-GEP, was
computation techniques. Previously, the empirical formulas were used to determine an empirical formula for predicting the UCS values.
commonly utilized to estimate the UCS based on Hr. To obtain more In this paper, 1106 groups of data set were collected from twenty-seven
accurate UCS, considerable efforts have been devoted to the develop- references, and the SA-GEP calculations were applied to obtain the
ment of empirical formulas to predict the UCS for various rock types by empirical formulas. Consequently, two empirical formulas were ob-
linear regression analysis,87,104,121–123 the multiple regression80,120 tained for UCS evaluation by using the L-type or N-type Schmidt
and nonlinear regression models.23,124–127 For empirical formulas, hammer. According to the computation results, the use of the L-type
there exist two problems that should be addressed properly: the ob- Schmidt hammer was strongly suggested, as was also found by previous
jective empirical formula type and the limited rock types. Con- studies.5,6,44,67 Consequently, the validity of the proposed empirical
ventionally, the empirical formula types were often determined based formula was verified by the Schmidt hammer tests and the uniaxial
on researchers’ experience. In the process of determining the empirical compressive tests, and the proposed empirical formula can be applied to
formula, some typical formula types were frequently utilized, such as engineering practice.

7
M. Wang and W. Wan International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104094

Table 4
The experimental results of Schmidt hammer tests and uniaxial compressive
tests.
Specimen Hr UCS(MPa)

Granite-1 67 189
Granite-2 64 178
Grainte-3 65 185
Granite-4 66 199
Yellow rust granite-1 55 124
Yellow rust granite-2 57 138
Yellow rust granite-3 62 150
Yellow rust granite-4 57 141
Red sandstone-1 20 25
Red sandstone-2 39 53
Red sandstone-3 29 40
Red sandstone-4 20 23
Markou limestone-1 50 92
Markou limestone-2 45 68
Markou limestone-3 49 87
Markou limestone-4 51 97
Skarn-1 52 99
Skarn-2 53 101
Skarn-3 46 84
Skarn-4 54 110

Fig. 5. The flow chart of SA-GEP for obtaining the empirical formula.

Fig. 7. The proposed empirical formula and the experimental results.

Schmidt hammer rebound orientation, weathering state of the rock, and


Fig. 6. The expression tree of Eq. (10). other factors,34,128,129 which can influence the UCS estimation, but
these factors are not discussed and considered in this paper, the con-
It should be noted that the empirical formulas obtained by SA-GEP sideration of these factors will be our next task. Nevertheless, the em-
are inherently very different from the conventional regression method. pirical formula was determined objectively using a large number of
Strictly speaking, the regression is only part of the procedure for de- data set, and the proposed empirical formula was verified by the
termining an empirical formula using the SA-GEP algorithm, and the Schmidt hammer tests and uniaxial compressive tests so that the em-
process of selecting the best empirical formulas according to the data pirical formula can be applied in engineering practice.
set uses artificial intelligence so that the SA-GEP method can be used to
determine the empirical formula more objectively. 5. Conclusions
However, it should be mentioned that the empirical formulas pro-
posed in this paper still exhibit some drawbacks that are caused by the The UCS of rock is regarded as the most widely used design para-
limitation of the Schmidt hammer test and the data set. In practice, the meter in the general field of rock engineering. As a nondestructive,
application of the Schmidt hammer test on very soft and extremely hard portable and cost-effective device for a hardness testing device, the
rocks is not recommended,63 hence, the empirical formula cannot ac- Schmidt hammer is often used to obtain an indirect estimation of UCS.
curately estimate the UCS of soft rocks and extremely hard rocks. Ad- To predict the UCS based on the Schmidt hammer rebound value ac-
ditionally, the Hr value in the data set for determining the empirical curately and simply, in this paper, references related to UCS estimation
formula are in the range of 10–70, and the Hr values in the experimental using the Schmidt hammer test were reviewed in detail. Afterwards,
data for verifying the proposed empirical formula were also between 10 1106 data sets in total were collected from twenty-seven references,
and 70. Hence, the Hr values in Eq. (13) should be between 10 and 70. and the SA-GEP algorithm was utilized. The corresponding empirical
Furthermore, the Schmidt hammer rebound values are influenced by formula was determined. In addition, the proposed empirical formula
the specimens’ size, rock type, density, anisotropy, moisture content, was verified by the Schmidt hammer tests and the uniaxial compressive

8
M. Wang and W. Wan International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104094

tests. The main conclusions of this paper are summarized below. 13. Monjezi M, Khoshalan HA, Razifard M. A neuro-genetic network for predicting
Through an analysis of the references, the development of the UCS uniaxial compressive strength of rock. Geotech Geol Eng. 2012;30:1053–1063.
14. Yagiz S, Sezer EA, Gokceoglu C. Artificial neural networks and nonlinear regression
estimation using Hr values can be classified into three stages. The merits techniques to assess the influence of slake durability cycles on the prediction of
and drawbacks of conventional empirical formulas and soft computa- uniaxial compressive strength and modulus of elasticity for carbonate rocks. Int J
tion techniques were analyzed. It was found that previous empirical Numer Anal Methods Geomech. 2012;36:1636–1650.
15. Deer DU, Miller R. Engineering Classification and Index Properties for Intact Rock.
formulas were determined subjectively. Deformation Curve AFNL-TR; 1966:65–116.
Conventionally, the empirical formula type was determined ac- 16. Aufmuth ER. A systematic determination of engineering criteria for rocks. Bull Assoc
cording to the researchers' experience, which is not scientific. To ad- Eng Geol. 1973;11:235–245.
17. Dearman WR, Irfan TY. Assessment of the degree of weathering in granite using
dress this problem, the SA-GEP method was applied to determine the petrographic and physical index tests. Proc. Int. Symp. On Deterioration and Protection
empirical formulas based on the data set. The empirical formulas were of Stone Monuments. Paris: Unesco; 1978:1–35.
established. According the computation results, the L-type Schmidt 18. Aggistalis G, Alivizatos A. Correlating uniaxial compressive strength with Schmidt
hammer rebound number, point load index, Young's modulus, and mineralogy of
hammer was recommended for practical use, and the empirical formula
6222 gabbros and basalts. Bull Eng Geol. 1996;54:3–11.
for the L-type Schmidt hammer is UCS = (88.15 − H ) − 70.38 19. Tugrul A. The effect of weathering on pore geometry and compressive strength of
r
(10 ≤ Hr ≤ 70) . However, the application of the Schmidt hammer test selected rock types from Turkey. Eng Geol. 2004;75:215–227.
20. Sabatakakis N, Koukis G, Tsiambaos G, Papanakli S. Index properties and strength
on very soft and extremely hard rocks is not recommended, hence, the variation controlled by microstructure for sedimentary rocks. Eng Geol. 2008;97(1-
empirical formula cannot accurately estimate the UCS of soft rocks and 2):80–90.
extremely hard rocks. Furthermore, the Schmidt hammer rebound va- 21. Andrade PS, Saraiva AA. Physical and mechanical characterization of phyllites and
metagreywackes in central Portugal. Bull Eng Geol Environ. 2010;69:207–214.
lues are influenced by the specimens’ size, rock type, density, aniso- 22. Torabi SR, Ataei M, Javanshir M. Application of Schmidt rebound number for es-
tropy, moisture content, Schmidt hammer rebound orientation, timating rock strength under specific geological conditions. J Min Environ.
weathering state of the rock, and other factors, which can influence the 2010;1(2):1–8.
23. Moomivand H. Development of a new method for estimating the indirect uniaxial
UCS estimation, but these factors are not discussed and considered in
compressive strength of rock using Schmidt hammer. BHM. 2011;156(4):142–146.
this paper, the proposed empirical formula can be applied in en- 24. Mishara DA, Basu A. Estimation of uniaxial compressive strength of rock materials
gineering practice according to specific condition. by index tests using regression analysis and fuzzy inference system. Eng Geol.
2013;160:54–68.
25. Fereidooni D. Determination of the geotechnical characteristices of Hornfelsic rocks
Acknowledgments with a particular emphasis on the correlation between physical and mechanical
properties. Rock Mech Rock Eng. 2016;49(7):2595–2608.
This study was funded by the National Natural Science Foundation 26. Uehigashi K, Tokairin Y, Ishikawa K, Kikuchi T. Possibility of rock excavation by
boom-type tunneling machines. 6th Australian Tunneling Conference. 1987;
of China (51774131, 51774132 & 51804110). 1987:253–259 Melbourne.
27. Day M, Goudie A. Field assessment of rock hardness using the Schmidt test hammer.
Appendix A. Supplementary data Geomorph Res Group Technol Bull. 1977;18:19–29.
28. Day MJ. Rock hardness: field assessment and geomorphic importance. Prof Geogr.
1980;32(1):72–81.
Supplementary data to this article can be found online at https:// 29. Mattews JA, Shakesby RA. The status of the ‘Little Ice Age’ in southern Norway:
doi.org/10.1016/j.ijrmms.2019.104094. relative-age dating of Neoglacial moraines with Schmidt hammer and lichenometry.
Boreas. 1984;13(3):333–346.
30. Stahl T, Winkler S, Quigley M, Bebbington M, Duffy B, Duke D. Schmidt hammer
6. Compliance with ethical standards exposure-age dating (SHD) of late Quaternary fluvial terraces in New Zealand. Earth
Surf Process Landforms. 2013;38(15):1838–1850.
31. Schmidt E. Anon-destructive concrete tester. Concrete. 1951;59(8):34–35.
Conflict of interest: The authors declare that they have no conflict of
32. Goudie AS. The Schmidt hammer in geomorphological research. Porgr Phys Geogr.
interest. 2006;30(6):703–718.
33. Hoseinie SH, Ataei M, Mikaiel R. Comparison of some rock hardness scales applied
References in drillability studies. Arabian J Sci Eng. 2012;37(5):1451–1458.
34. Demirdag S, Yavuz H, Altingdag R. The effect of sample size on Schmidt rebound
hardness value of rocks. Int J Rock Mech Min Sci. 2009;46(4):725–730.
1. Bieniawski ZT. Point load test in geotechnical practice. Eng Geol. 1975;9(1):1–11. 35. Yagiz S. Predicting uniaxial compressive strength, modulus of elasticity and index
2. ISRM. The complete ISRM suggested methods for rock characterization, testing and properties of rocks using the Schmidt hammer. Bull Eng Geol Environ.
monitoring. In: Ulusay R, Hudson JA, eds. Suggested Methods Prepared by the 2009;68(1):55–63.
Commission on Testing Methods, International Society for Rock Mechanics. Ankara, 36. Basu A, Aydin A. A method for normalization of Schmidt hammer rebound values.
Turkey: RSRM Turkish National Group; 2007:1974–2006. Int J Rock Mech Min Sci. 2004;41:1211–1214.
3. ASTM. Standard Test Method for Unconfined Compressive Strength of Intact Rock Core 37. Shakesby RA, Matthews JA, Owen G. The Schmidt hammer as a relative-age dating
Specimens. Soil and Rock, Building Stones: Annual Book of ASTM Standards 4.08. tool and its potential for calibrated-age dating in Holocene glaciated environments.
Philadelphia, Pennsylvania: ASTM; 1984. Quat Sci Rev. 2016;25(21):2846–2867.
4. Gokceoglu C, Zorlu K. A fuzzy model to predict the unconfined compressive strength 38. Howarth D, Adamson W, Berndt J. Correlation of model tunnel boring and drilling
and modulus of elasticity of probluematic rock. Eng Appl Artif Intell. 2004;17:61–72. machine performances with rock properties. Int J Rock Mech Min Sci.
5. ASTM. Standard Practice for Preparing Rock Core Specimens and Determining 1986;23(2):171–175.
Dimensional and Shape Tolerances. D4543 Philadelphia Pa: ASTM; 2001. 39. Kahraman S. Rotary and percussive drilling prediction using regression analysis. Int
6. Suggested ISRM. Methods for determining the uniaxial compressive strength and J Rock Mech Min Sci. 1999;36(7):981–989.
deformability of rock materials. Int J Rock Mech Min Sci Geomech Abstr. 40. Kahraman S, Balc C, Yazc S, Bilgin N. Prediction of the penetration rate of rotary
1979;16:135–140. blast hole drills using a new drillability index. Int J Rock Mech Min Sci.
7. Meulenkamp F, Grima MA. Application of neural networks for the unconfined 2000;37(5):729–743.
compressive strength (UCS) from Equotip hardness. Int J Rock Mech Min Sci. 41. Li X, Rupert G, Summers DA, Santi P, Liu D. Analysis of impact hammer rebound to
1999;36(1):29–39. estimate rock drillability. Rock Mech Rock Eng. 2000;33(1):1–13.
8. Diamantis K, Gartzos E, Migiros G. Study on uniaxial compressive strength point 42. Kahraman S, Bilgin N, Feridunoglu C. Dominant rock properties affecting the pe-
load strength index dynamic and physical properties of serpentinites from Central netration rate of percussive drills. Int J Rock Mech Min Sci. 2003;40(5):711–723.
Greece: test results and empirical relations. Eng Geol. 2009;108:199–207. 43. Poole R, Farmer I. Consistency and repeatability of Schmidt hammer rebound data
9. Gokceoglu C, Sonmez H, Zorlu K. Estimating the uniaxial compressive strength of during field testing. Int J Rock Mech Min Sci. 1980;17(3):167–171.
some clay-bearing rocks selected from Turkey by nonlinear multivariable regression 44. ISRM. Suggested Methods for Determining Hardness and Abrasiveness of Rocks. Part 3
and rule-based fuzzy models. Expert Syst. 2009;26(2):176–190. 1981; 1981:101–102.
10. Gokceoglu C, Zorlu K, Ceryanc S, Nefeslioglu HA. A comparative study on indirect 45. Kidybinski A. Rebound number and the quality of mine roof strata. Int J Rock Mech
determination of degree of weathering of granites from some physical and strength Min Sci. 1968;5(4):283–292.
parameters by two soft computing techniques. Mater Char. 2009;60:1317–1327. 46. Karpuz C. A classification system for excavation of surface coal measures. Min Sci
11. Yilmaz I, Yuksek G. Prediction of the strength and elastic modulus of gypsum using Technol. 1990;11(2):157–163.
multiple regression ANN and ANFIS models. Int J Rock Mech Min Sci. 47. Sachpazis C. Correlating Schmidt hardness with compressive strength and Young's
2009;46:803–810. modulus of carbonate rocks. Bull. Int. Assoc. Eng. Geol.-Bull. de 1’ Assoc.
12. Del Potro R, Hurlimann M. A comparison of different indirect techniques to evaluate Internationale de Geologie de l’Ingnieur. 1990;42(1):75–83.
volcanic intact rock strength. Rock Mech Rock Eng. 2009;42:931–938. 48. Kazi A, Al-Mansour A. Empirical relationship between LosAngeles abrasion and

9
M. Wang and W. Wan International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104094

Schmidt hammer strength tests with application to aggregates around Jeddah. Q J compressive strength and modulus of elasticity for Travertine samples using re-
Eng Geol. 1980;13(1):45–52. gression and artificial neural networks. Min Sci Technol. 2010;20:41–46.
49. Cargill JS, Shakoor A. Evaluation of empirical methods for measuring the uniaxial 82. Rezaei M, Majdi A, Monjezi M. An intelligent approach to predict unconfined
compressive strength of rock. Int J Rock Mech Sci. 1990;27(6):495–503. compressive strength of rock surrounding access tunnels in longwall coal mining.
50. Yilmaz I, Sendr H. Correlation of Schmidt hardness with unconfined compressive Neural Comput Appl. 2012;24(1):233–241 2012.
strength and Young's modulus in gypsum from Sivas (Turkey). Eng Geol. 83. Yurdakul M, Akdas H. Modeling uniaxial compressive strength of building stones
2002;66(3):211–219. using non-destructive test results as neutral networks input parameters. Constr Build
51. Yasar E, Erdogan Y. Estimation of rock physicomechanical properties using hard- Mater. 2013;47:1010–1019.
ness methods. Eng Geol. 2004;71(3):281–288. 84. Momeni E, Armaghani DJ, Hajihassani M, Amin MFM. Prediction of uniaxial
52. Dincer I, Ascar A, Cobanoglu I, Uras Y. Correlation between Schmidt hardness, compressive strength of rock samples using hybrid particle swarm optimization-
uniaxial compressive strength and Young's modulus for andesites, basalts and tuffs. based arificial neural netwoks. Measurement. 2015;60:50–63.
Bull Eng Geol Environ. 2004;63(2):141–148. 85. Armaghani DJ, Amin MFM, Yagiz S, Faradonbeh RS, Abdullah RA. Prediction of the
53. Shalabi FI, Cording EJ, Al-Hattamleh OH. Estimation of rock engineering properties uniaxial compressive strength of sandstone using various modeling techniques. Int J
using hardness tests. Eng Geol. 2007;90(3):138–147. Rock Mech Min Sci. 2016;85:174–186.
54. Cobanoglu I, Celik SB. Estimation of uniaxial compressive strength from point load 86. Fattahi H. Applying soft computing methods to predict the uniaxial compressive
strength, Schmidt hardness and P-wave velocity. Bull Eng Geol Environ. strength of rock from schmidt hammer rebound values. Comput Geosci.
2008;67(4):491–498. 2017;21(1):1–17.
55. Singh R, Hassani F, Elkington P. The application of strength and deformation index 87. Heidari M, Mohseni H, Jalali SH. Prediction of uniaxial compressive strength of
testing to the stability assessment of coal measures excavations. The 24 Th US some sedimentary rocks by fuzzy and regression models. Geotech Geol Eng.
Symposium on Rock Mechanics (USRMS). American Rock Mechanics Association; 2018;36:401–412.
1983. 88. Yagiz S, Gokceoglu C, Sezer E, Iplikci S. Application of two non-linear prediction
56. Sheorey P, Barat D, Das M, Mukherjee K, Sigh B. Schmidt hammer rebound data for tools to the estimation of tunnel boring machine performance. Eng Appl Artif Intell.
estimation of large scale in situ coal strength. Int J Rock Mech Min Sci Geomech Abstr. 2009;22(4):808–814.
1984;1:39–42. 89. Sonmez H, Gokceoglu C, Nefeslioglu HA, Kayabasi A. Estimation of rock modulus:
57. Haramy K, DeMarco M. Use of the Schmidt hammer for rock and coal testing. The 26 for intact rocks with an artificial neural network and for rock masses with a new
Th US Symposium on Rock Mechanics (USRMS). American Rock Mechanics empirical equation. Int J Rock Mech Min Sci. 2006;43:224–235.
Association; 1985. 90. Sonmez H, Gokceoglu C. Discussion on the paper by H. Gullu and E. Ercelebi A
58. Ghose A. Empirical strength indices of Indian coals-an investigation. The 27 Th US neural network approach for attenuation relationships: an application using strong
Symposium on Rock Mechanics (USRMS). American Rock Mechanics Association; ground motion data from Turkey. Eng Geol. 2008;97:91–93.
1986. 91. Ferreira C. Gene expression programming: a new adaptive algorithm for solving
59. O'Rourke J. Rock index properties for geoengineering in underground development. problems. Complex Syst. 2001;13(2):87–129.
Min Eng. 1989;41:106–110. 92. Ferreira C. Gene expression programming in problem solving. In: Roy R, Koppen M,
60. Kilic A, Teymen A. Determination of mechanical properties of rock using simple Ovaska S, Furuhashi T, Hoffmann F, eds. Soft Computing and Industry-Recent
methods. Bull Eng Geol Environ. 2008;67(2):237–244. Applications. Springer-Verlag; 2002:635–654.
61. Beverly BE, Schoenwolf DA, Brierly GS. Correlations of Rock Index Values with 93. Faradonbeh RS, Armaghani DJ, Monjezi M, Mohamad ET. Genetic programming
Engineering Properties and the Classification of Intact Rock. Washington, DC: FHWA; and gen expression programming for flyrock assessment due to mine blasting. Int J
1979. Rock Mech Min Sci. 2016;88:254–264 2016.
62. Kidybinski A. Bursting liability indices of coal. Int J Rock Mech Min Sci Geomech 94. Steeb WH. The Nonlinear Workbook: Chaos, Fractals, Cellular Automata, Neural
Abstr. 1980;17:167–171. Networks, Genetic Algorithms, Gene Expression Programming, Support Vector Machine,
63. Xu S, Grasso P, Mahtab A. Use of Schmidt hammer for estimating mechanical Wavelets, Hidden Markov Models, Fuzzy Logic with C++, Java and Symbolic C++
properties of weak rock. 6 Th International IAEG Congress. Rotterdam: Balkema; Programs. third ed. Singapore: World Scientific; 2011.
1990:511–519. 95. Brownlee J. Clever Algorithms: Nature-Inspired Programming Recipes. Jason Brownlee;
64. Tugrul A, Zarif IH. Correlation of mineralogical and textural characteristics with 2011.
engineering properties of selected granitic rocks from Turkey. Eng Geol. 96. Karakus M. Function identification for the intrinsic strength and elastic properties of
1999;51:303–317. granitic rock via genetic programming (GP). Comput Geosci. 2011;37(9):1318–1323.
65. Katz O, Reches Z, Roegiers JC. Evaluation of mechanical rock properties using a 97. Koza JR. Genetic Programming: On the Programming of Computers by Means of Natural
Schmidt hammer. Int J Rock Mech Min Sci. 2000;37(4):723–728. Selection. Cambridge, MA: MIT press; 1992.
66. Kahraman K. Evaluation of simple methods for assessing the uniaxial compressive 98. Baykasoglu A, Gullu H, Canakci H, Ozbakir L. Prediction of compressive and tensile
strength of rock. Int J Rock Mech Min Sci. 2001;38:981–994. strength of limestone via genetic programming. Expert Syst Appl. 2008;35:111–123.
67. Aydin A, Basu A. The Schmidt hammer in rock material characterization. Eng Geol. 99. Silva S, Almeida J. Dynamic maximum tree depth—a simple technique for avoiding
2005;81:1–14. bloat in tree-based GP. In: Cantu-Paz E, Foster JA, Deb K, eds. Proceedings of the
68. Fener M, Kahraman S, Bilgil A, Gunaydin O. A comparative evaluation of indirect GECCO 2003. Berlin: Springer; 2003:1776–1787.
methods to estimate the compressive strength of rocks. Rock Mech Rock Eng. 100. Teodorescu L, Sherwood D. High energy physics event selection with gene ex-
2005;38(4):329–343. pression programming. 2008 Comput Phys Commun. 2008;178:409–419.
69. Gupta V. Non-destructive testing of some higher Himalayan rocks in the Satluj 101. Yang Y, Li X, Gao L, Shao X. A new approach for predicting and collaborative
Valley. Bull Eng Geol Environ. 2009;68:409–416. evaluating the cutting force in face milling based on gene expression programming.
70. Minaeian B, Ahangari K. Estimation of uniaxial compressive strength based on P- J Netw Comput Appl. 2013;36(6):1540–1550 2013.
wave and Schmidt hammer rebound using statistical method. Arab J Geosci. 102. Kayadelen C. Soil liquefaction modeling by genetic expression programming and
2013;6:1925–1931. neuro-fuzzy. Expert Syst Appl. 2011;38:4080–4087.
71. Bruno G, Vessia G, Bobbo L. Statistical method for assessing the uniaxial com- 103. Mollahasani A, Alavi AH, Gandomi AH. Empirical modeling of plate load test
pressive strength of carbonate rock by Schmidt hammer tests performed on core moduli of soil via gene expression programming. Comput Geotech. 2011;38:281–286
samples. Rock Mech Rock Eng. 2013;46(1):199–206. 2011.
72. Saptono S, Kramadibratab S, Sulistiantob B. Using the Schmidt hammer on rock 104. Adler D. Genetic algorithms and simulated annealing: a marriage proposal.
mass characteristic in sedimentary rock at Tutupan Coal Mine. Procedia Earth Planet International Symposium on Neural Networks. 1993; 1993:1104–1109.
Sci. 2013;6:390–395. 105. Dupanloup I, Schneider S, Excoffier L. A simulated annealing approach to define the
73. Karaman K, Kesimal A. A comparative study of Schmidt hammer test methods for genetic structure of populations. Mol Ecol. 2002;11(12):2571–2581.
estimating the uniaxial compressive strength of rocks. Bull Eng Geol Environ. 106. Kipkpatrick S, Gelatt CD, Vecchi MP. Optimization by simulated annealing. Science.
2015;74:507–520. 1983;220(4598):671–680.
74. Armaghani DJ, Mohamad ET, Momeni E, Monjezi M, Narayanasamy MS. Prediction 107. Karakus M, Tutmez B. Fuzzy and multiple regression modelling for evaluation of
of the strength and elasticity modulus of granite through an expert artificial neural intact rock strength based on point load, Schmidt hammer and sonic velocity. Rock
network. Arab J Geosci. 2016;9:48. Mech Rock Eng. 2006;39(1):45–57.
75. Liang M, Mohamad ET, Faradonbeh RS, Armaghani DJ, Ghoraba S. Rock strength 108. Buyuksagis IS, Goktan RM. The effect of Schmidt hammer type on uniaxial com-
assessment based on regression tree technique. Eng Comput. 2016;32:343–354. pressive strength prediction of rock. Int J Rock Mech Min Sci. 2007;44:299–307.
76. Azimian A. Application of statistical methods for predicting uniaxial compressive 109. Kayabali K, Selcuk L. Nail penetration test for determining the uniaxial compressive
strength of limestone rocks using nondestructive tests. Acta Geotechnica. strength of rock. Int J Rock Mech Min Sci. 2010;47:265–271.
2017;12(2):1–13. 110. Yarali O, Soyer E. Assessment of relationships between drilling rate index and
77. Hebib R, Belhai D, Alloul B. Estimation of uniaxial compressive strength of North mechanical properties of rock. Tunn Undergr Space Technol. 2013;33:46–53.
Algeria sedimentary rocks using density, porosity, and Schmidt hardness. Arab J 111. Ataei M, Kakaie R, Ghavidel M, Saeidi S. Drilling rate prediction of an open pit mine
Geosci. 2017;10:383. using the rock mass drillability index. Int J Rock Mech Min Sci. 2015;73:130–138.
78. Kong K, Shang J. A validation study for the estimation of uniaxial compressive 112. Tandon RS, Gupta V. Estimation of strength characteristics of different Himalayan
strength based on index tests. Rock Mech Rock Eng. 2018;51:2289–2297. rocks from Schmidt hammer rebound, point load index, and compressional wave
79. Garret JH. Where and why artificial neural networks are applicable in civil en- velocity. Bull Eng Geol Environ. 2015;74:521–533.
gineering. J Comput Civ Eng. 1994;8:129–130. 113. Karaman K, Kesimal A. Correlation of Schmidt hammer rebound hardness with
80. Yilmaz I, Yuksek AG. An example of artificial neural network (ANN) application for compressive strength and P-wave velocity of rock materials. Arabian J Sci Eng.
indirect estimation of rock parameters. Rock Mech Rock Eng. 2008;41(5):781–795. 2015;40:1897–1906.
81. Dehghan S, Sattari GH, Chehreh Chelgani S, Aliabadi MA. Prediction of uniaxial 114. Ng IT, Yuen KV, Lau CH. Predictive model for uniaxial compressive strength for

10
M. Wang and W. Wan International Journal of Rock Mechanics and Mining Sciences 123 (2019) 104094

grade III granitic rocks from Macao. Eng Geol. 2015;199:28–37. 122. Ulusay R, Tureli K, Ider MH. Prediction of engineering properties of a selected li-
115. Akram MS, Farooq S, Naeem M, Ghazi S. Prediction of mechanical behaviour from tharenite sandstone from its petrographic characteristics using correlation and
mineralogical composition of Sakesar limestone, central Salt range, Pakistan. Bull multivariate statistical techniques. Eng Geol. 1994;38:135–157.
Eng Geol Environ. 2017;76(2):601–615. 123. Li D, Wong LNY. Point load test on meta-sedimentary rocks and correlation to UCS
116. Afolagboye LO, Talabi AO, Oyelami CA. The use of index tests to determine the and BTS. Rock Mech Rock Eng. 2013;46(4):889–896.
mechanical properties of crushed aggregates from precambrian basedment complex 124. Grasso P, Xu S, Mahtab A. Problems and promises of index testing of rocks. 33 Rd US
rocks, Ado-Ekiti, SW Nigeria. J Afr Earth Sci. 2017;129:659–667. Symposium on Rock Mechanics. Sante Fe: Balkema; 1992:879–888 Rotterdam.
117. Ghasemi E, Kalhori H, Bagherpour R, Yagiz S. Model tree approach for predicting 125. Palchik V, Hatzor YH. The influence of porosity on tensile and compressive strength
uniaxial compressive strength and Young's modulus of carbonate rocks. Bull Eng of porous chalk. Rock Mech Rock Eng. 2004;37(4):331–341.
Geol Environ. 2018;77:331–343. 126. Quane SL, Russel JK. Rock strength as a metric of welding intensity in pyroclastic
118. Petrakis E, Komnitsas K. Correlation between Material properties and breakage rate doposits. Eur J Mineral. 2003;15:855–864.
parameters determined from grinding tests. Appl Sci. 2018;8(2):220. 127. Tsiambaos G, Sabatakakis N. Considerations on strength of intack sedimentary
119. Yilmaz I. A new testing method for indirect determination of the unconfined rocks. Eng Geol. 2004;72(3-4):261–273.
compressive strength of rocks. Int J Rock Mech Min Sci. 2009;46(8):1349–1357. 128. Hucka V. A rapid method of determining the strength of rocks in situ. Int J Rock
120. Ceryan N, Okkan U, Kesimal A. Prediction of unconfined compressive strength of Mech Min Sci. 1965;2(2):127–134.
carbonate rocks using artificial neural networks. Environ Earth Sci. 129. Goktan R, Ayday C. A suggested improvement to the Schmidt rebound hardness
2012;68:807–819. ISRM suggested method with particular reference to rock machineability. Int J Rock
121. Gunsallus KL, Kulhawy FH. A comparative evaluation of rock strength measures. Int Mech Min Sci. 1993;30(3):321–322.
J Rock Mech Min Sci. 1984;21:233–248.

11

You might also like