Nigerian Journal of Technology (NIJOTECH)

Vol. 35, No. 2, April 2016, pp. 260 – 269

Copyright© Faculty of Engineering, University of Nigeria, Nsukka,
Print ISSN: 0331-8443, Electronic ISSN: 2467-8821
www.nijotech.com
http://dx.doi.org/10.4314/njt.v35i2.5

ESTIMATION OF SHEAR STRENGTH PARAMETERS OF LATERITIC SOILS

USING ARTIFICIAL NEURAL NETWORK

S. D Iyeke1, *, E. O Eze2, J. O Ehiorobo, 3 and S. O Osuji4

1.2,3,4 DEPT. OF CIVIL ENGINEERING, FACULTY OF ENGINEERING, UNIVERSITY OF BENIN, BENIN CITY, EDO STATE, NIGERIA.

E-mail addresses: 1 [email protected] , 2 [email protected], 3 [email protected], 4 [email protected]

ABSTRACT
This research work seeks to develop models for predicting the shear strength parameters (cohesion and angle of
friction) of lateritic soils in central and southern areas of Delta State using artificial neural network modeling
technique. The application of these models will help reduce cost and time in acquiring geotechnical data needed for
both design and construction in the study area. A total of eighty-three (83) soil samples were collected from various
locations in Delta State of Nigeria. The geotechnical soil properties were determined in accordance with British
Standards. The range of the angle of internal friction and cohesion obtained from the tests are 2 to 43 degrees and
3 to 82 kN/m2 respectively. The optimum artificial neural network architecture network was found to be 3-9-1,
that is three inputs, nine hidden layer nodes, and one output node for cohesion. While, the angle of friction had an
optimal network geometry of 3-11-1, that is three inputs, eleven hidden layer nodes, and one output node. The
results of the coefficient of determination and root mean square showed that the artificial neural network method
outperforms some selected empirical formulae in the prediction of shear strength parameters.

Keywords: Artificial Neural Network, Lateritic Soil, Angle of Friction, Cohesion

1. INTRODUCTION the laboratory. The triaxial compression and direct

One of the most essential engineering properties of shear tests are the most common tests for
soil is its capacity to oppose sliding along inner determining the cohesion and angle of friction values
surfaces within a mass. The solidity of any structure in the laboratory. Measurements of shear strength
built on soil will depend upon the shearing resistance properties both at field and laboratory conditions are
presented by the soil along likely surfaces of slippage cumbersome, expensive, time-consuming and labour-
[1]. The understanding of the shear strength of a soil is intensive [7, 8, 12]. In order to cope with the difficulty
important in the assessment of bearing capacities of of experimental investigation, engineering design
foundations [2], slope stability [3], retaining models are needed. Correlations and empirical
structures, embankment dams [4], tunnel linings, relationships are principally useful in preliminary
pavement [5] and the resistance traction and tillage studies, or when due to time and/or financial
tools in agricultural applications [6].The shear constraints that a thorough geotechnical examination
strength of soils is generally represented by the Mohr– cannot to be conducted. This is most relevant in third
Coulomb theory. The theory, indicate that the shear world countries where up-to-date testing equipment
strength of soils varies linearly with the applied stress are lacking together with the trained manpower
through two shear strength factors; cohesion and needed to operate them. In the recent years, the
angle of shearing resistance [7, 8]. The tangent to the development of prediction models that use easier to
Mohr–Coulomb failure envelopes is defined by its determine secondary information to spatially extend
slope and intercept. The slope expressed in degrees is sparse and expensive soil measurements has been a
the angle of shearing resistance and the intercept is sharpening focus of research [13]. Empirical
the cohesion [9, 10, 11]. The soil shear strength correlations are widely used in geotechnical
parameters can be determined either in the field or in engineering practice as a tool to estimate the

*Corresponding author, tel: +234-803-916-7870

 ESTIMATION OF SHEAR STRENGTH PARAMETERS OF LATERITIC SOILS USING ARTIFICIAL NEURAL NETWORK S. D Iyeke et al

engineering properties of soils. Useful correlations

exist between the index properties obtained from 2. MATERIALS AND METHODS
simple routine testing and the strength properties of 2.1 Sampling and Sampling Locations
soils among others [14]. Soils samples were collected at random points from
Several empirical procedures have been developed central and southern areas of Delta State. The
over the years to estimate the shear strength sampling locations lie within longitude 50 30’ E and 60
parameters for soils. Among the various models 20’E and latitude 50 10’ N and 60 20’ N. Delta State is
developed are; Masada [15] for clay and silt one of the states in the Niger Delta Region, it has
embankments, Mofiz and Rahman [16] for Barind twenty-five local government areas. The location of
soils, Cola and Cortellazo [17] for peaty soils and Delta State and soil samples locations are shown in
Hajarwish and Shakor [18] for mudrock. Also, some Figures 1 and 2 respectively. Disturbed soil samples
models have been developed based on sustained were obtained at different depths (1m – 25m) from
hypothesis imposed by the researchers before road cuttings and borrow pits at various locations
estimation of the model parameters according to their using hand auger. Also additional soil data were
assumptions concerning how the model parameters obtained from construction and consultancy firms. A
the dependent and explanatory variables are related, total of eighty – three soil data were collected.
[7, 8, 18-21]. Hence these various models may not
have been the appropriate ones. The advent of soft 2.2 Geotechnical Analysis of the Soils
computing methods like artificial neural The classification tests as well as tests to determine
network(ANN) in developing robust models, where the moisture-density relationship and shear strength
the data trend are allowed to evolve a appropriate were carried out in accordance to BS 1377 [22]. The
models is becoming widely accepted. There are little shear tests were conducted in compliance to BS 1377
or no documentation with regards to the use of these [22]. The geotechnical tests are; grain size
modeling tools for the prediction of shear strength distribution, plastic limit, liquid limit, specific gravity,
parameters for lateritic soils. This research work compaction, shear box tests and triaxial compression
seeks to develop a relationship between index soils tests.
properties and shear strength parameters for lateritic
soils.

Figure 1. Map of Nigeria Showing Delta State

Nigerian Journal of Technology Vol. 35, No. 2, April 2016 261

 ESTIMATION OF SHEAR STRENGTH PARAMETERS OF LATERITIC SOILS USING ARTIFICIAL NEURAL NETWORK S. D Iyeke et al

Figure 2. Map of Delta State Showing Samples Locations and Study Area

2.3 ANN Modeling Procedure

The input factors considered for shear strength
parameters included; plasticity index, percentage of
particles passing sieve No.200, specific gravity, liquid 3
limit, plastic limit [7,20,21,23,24]. The inputs selected
is based on the fact that soils classification is based on
these parameters and the specific gravity is a [ ]
reflection of the inert properties of the soil. Cohesion The nonlinear function can be written as:
and angle of friction were the single output variables ( )
in the various models. Correlation matrix for input The output of the second layer can then be written as
variables was determined using the Pearson
( ) 5
correlation. Pearson correlation is one of the variable
ranking criteria used in selecting inputs for ANN [25, The two-layer perceptron shown in Figure 3 can be
26]. also written in nested form as
The proposed multi-layer perception for the ( ( )) 6
prediction of angle of internal friction and cohesion in There are no general guidelines in determining the
the soil problem is shown in Figure 3. number of data for training the artificial neural
The input to the multi-layer perceptron is a vector of network model to perform effectively. However
M attribute values Lawrence and Fredeickson [27] suggested the
[ ] 1 following rule of thumb;
Where j = 1,2,3,4,5 are the input variables. 2 10
The output of the weighting and summation in the Where: i is the number of input neurons; h is the
first layer the “hidden” layer can be written as number of hidden neuron; o is the number of output
2 neurons; n is the number of data.
The database was randomly divided into three sets:
1 training, testing, and validation. In total, 80% of the
data were used for training and 20% were used for
validation. The training data were further subdivided
Where , into 70% for the training set and 30% for the testing
set [28].
The back propagation learning algorithm is the most
[ ] popular and extensively used neural network
[ ]
and algorithm [29-32].

Nigerian Journal of Technology Vol. 35, No. 2, April 2016 262

 ESTIMATION OF SHEAR STRENGTH PARAMETERS OF LATERITIC SOILS USING ARTIFICIAL NEURAL NETWORK S. D Iyeke et al

Figure3. Proposed Model structure Architecture

Figure 4. Scatter Plot of Cohesion and Other Soil Figure5. Scatter Plot of Angle of Friction and Other Soil
Properties Properties

The back-propagation neural network has been In equation (9) there is no superscript on the weights
applied with great success to model many phenomena since we will use this equation to compute the values
in the field of geotechnical and geo-environmental for both sets of weights. The difference in the weight
engineering [28, 33, 34]. Logarithmic sigmoid transfer values can be evaluated using the gradient descent
function was used as the activation function for method [35], in which we differentiate the error term
hidden and output layers. The input – output data of with respect to the weights, giving
each ANN model were pre-processed to lie between 0
10
and 1 by using Eq. (8);
where η is a scaling value between 0 and 1. The
weights are updated iteratively by the equation
Where is the normalized value, X is the actual w (u+1) = w(u Δw(u) (11)
value, is the maximum value and is the where w (u+1) is the new weight; w(u) is the old
minimum value. weight and Δw(u) is the variation between the values
The weights were first initialized to small arbitrary of new and old weights.
values. It is transmitted to obtain the solution using Once the training phase of the model has been
these initial weights. Once the output value has been effectively completed, the presentation of the trained
calculated, we decrease the squared error, which can model is evaluated using the validation data, which
be written as a function of the weights as; have not been used as part of the model building
1 process.. The coefficient of determination (r2), the
( ) root-mean-square error (RMSE), and the mean
2
Nigerian Journal of Technology Vol. 35, No. 2, April 2016 263
 ESTIMATION OF SHEAR STRENGTH PARAMETERS OF LATERITIC SOILS USING ARTIFICIAL NEURAL NETWORK S. D Iyeke et al

absolute error (MAE) [36], are the main criteria that The scatter diagrams plots for between cohesion and
are used to assess the performance of the ANN models liquid limit, plasticity index, passing sieve No. 200 is
obtained in this work. shown in Figure 4, while that of angle of friction and
liquid limit, plasticity index, passing sieve No. 200 is
∑
√ 12 shown in Figure 5.
As can be seen from the Figures 4 and 5, there are
∑ | | extremely nonlinear relationships among the selected
13
parameters, and several uphill and downhill points
which exist on the graphs. In all the graphs, the points
∑ ̌ ̌
1 are not very useful at explaining the relationship with
√∑ ̌ ̌ reference to the dependent variables. A correlation
matrix was carried out on the soil parameters using
the Pearson’s correlation. The correlation matrix is
3. RESULTS AND DISCUSSIONS given in Table 2. The matrix indicated a high
3.1 Statistical Analysis of Data correlation between liquid limit, plastic limit and
Statistical descriptions of examined soils parameters plasticity index.
are given in Table 1. It can be seen from the Table, that The liquid limit and plastic limit parameters were
the distribution of the medium and average values of expunged from the models to avoid multicollinearity.
the soil properties are close together. This shows that Also the correlation of cohesion and angle of internal
soil experimental data are approximately normally with the other soil parameters did not give a high
distributed. This is further collaborated by the values relationship, hence using regression analysis is likely
of the skewness (-0.925 to 1.529) and kurtosis (-0.673 to produce inappropriate model. The soil data were
to 2.873). These values are close to zero indicating divided into three sets; Training set, Testing set and
slight skewness and asymmetry degree with reference Validation sets. The statistics for each set is given in
to normal distribution. Table 3

Table 1. Basic Descriptive Statistics for the soil Data

Passing Sieve Plasticity Liquid Limit Specific Cohesion Angle of
Statistics
No.200 (%) Index (%) (%) Gravity (kN/m2) Friction(0)
Minimum 11.20 0.0 12 2.2 3 2
Maximum 52.80 40.0 55 2.69 82 43
Average 29.86 15.5 31.25 2.54 21.24 16.71
Medium 30.1 16.0 31 2.56 16.0 16
Standard deviation 10.36 10.4 8.741 0.10 15.27 9.52
Skewness 0.108 -0.085 0.310 -0.956 1.529 0.425
Kurtosis -0.672 -0.673 -0.162 0.841 2.873 -0.231
n 83 83 83 83 83 83

Table 2.Correlation Matrix of Soil Parameters with Shear Strength Parameters.

Passing Angle of
Liquid Plastic Sieve Internal
Limit Plasticity Limit Specific No. 200 Cohesion Friction
Soil Parameters (%) Index (%) (%) Gravity (%) (KN/m2) ( 0)
Liquid Limit (%) 1.0000
Plasticity Index (%) 0.9120 1.0000
Plastic Limit (%) 0.6626 0.6807 1.0000
Specific Gravity -0.2970 -0.3866 -0.1695 1.0000
Passing Sieve No. 200 (%) 0.6197 0.6647 0.6074 -0.5002 1.0000
Cohesion (KN/m3) 0.4898 0.5291 0.3689 -0.0346 0.3841 1.0000
Angle of Internal Friction (0) 0.0044 -0.0466 -0.0929 0.2627 -0.1756 -0.0935 1.0000

Nigerian Journal of Technology Vol. 35, No. 2, April 2016 264

 ESTIMATION OF SHEAR STRENGTH PARAMETERS OF LATERITIC SOILS USING ARTIFICIAL NEURAL NETWORK S. D Iyeke et al

Table 3. Artificial Neural Network Input and Output Statistics

Statistical Parameters
Model variables and data sets Standard
Mean Minimum Maximum Range
Deviation
Plasticity Index (%)
Training Set 16.0 10.7 0.0 38.0 38.0
Testing Set 13.5 11.0 0.0 32.0 32.0
Validation Set 16.8 9.2 0.0 40.0 40.0
Specific Gravity
Training Set 2.5 0.1 2.2 2.7 0.5
Testing Set 2.6 0.1 2.5 2.7 0.2
Validation Set 2.6 0.1 2.4 2.6 0.2
Passing Sieve No.200 (%)
Training Set 30.0 10.9 11.2 50.8 39.6
Testing Set 28.9 11.5 12.5 52.8 40.3
Validation Set 30.6 7.1 16.9 43.3 26.4
Cohesion (KN/m2)
Training Set 20.0 12.1 6.0 52.0 46.0
Testing Set 20.0 15.8 5.0 70.0 65.0
Validation Set 26.1 21.4 3.0 82.0 79.0
Angle of Friction (0)
Training Set 14.7 8.5 2 31 29
Testing Set 19.6 9.6 5 41 36
Validation Set 18.6 11.1 2 43.0 41

Despite trying numerous random combinations of

training, testing, and validation sets, there are still
some slight inconsistencies in the statistical
parameters for the training, testing, and validation
sets that are most closely matched. This can be
attributed to the fact that the data contain singular,
rare events, that cannot be replicated in all three data
sets. However, on the whole, the statistics are in good
agreement and all three data sets may be considered
to represent the same population.

3.2 Visual Basic Programme for ANN

A program was written and run in Visual Basic for the Figure 6. Visual Basic Interface Showing the
proposed ANN model. The interface for the Processing Functions
programme is shown in Figure 6. The multilayer
perception (MLP) can have more than one hidden The optimal network architecture obtained for the
layer; however, several works have shown that a cohesion model was 3-9-1 i.e., three inputs, nine
single hidden layer is sufficient for an ANN to hidden layer nodes, and one output node, the
approximate any complex nonlinear function [36, 37]. optimum learning rate was found to be 0.2 and
Therefore, in this study, a one-hidden-layer MLP was results of the prediction is shown in Figures 7,8 and 9.
used. Logarithmic sigmoid transfer function was used Also, the optimal network architecture obtained for
as the activation function for hidden and output the angle of friction model was 3-11-1 i.e., three
layers. The numbers of hidden layer neurons were inputs, eleven hidden layer nodes, and one output
found using simple trial-and-error method adjudged node, the optimum learning rate was found to be 0.4
by the root mean square errors (RMSE). and the results of the prediction is shown in Figures
9,10 and 11.
Nigerian Journal of Technology Vol. 35, No. 2, April 2016 265
 ESTIMATION OF SHEAR STRENGTH PARAMETERS OF LATERITIC SOILS USING ARTIFICIAL NEURAL NETWORK S. D Iyeke et al

Figure 7(a).Predicted cohesion values versus Figure 7(b). Predicted cohesion values versus
Experimental Cohesion Values for Training Set Experimental Cohesion Values for Testing Set

Figure 8. Predicted cohesion values versus Experimental

Figure 9. Predicted Angle of Friction values versus
Cohesion Values for Validation Set
Experimental Angle of Friction Values for Training Set

Figure 10. Predicted Angle of Friction values versus

Figure 11.Predicted Angle of Friction values versus
Experimental Angle of Friction Values for Testing Set
Experimental Angle of Friction Values for Validation Set.

Table 4. Comparison of Artificial Neural Network and Other Empirical Method for Cohesion Prediction
S/N Factors Considered ANN Ersoy et al.([8] Roy and Dass([7] Adunoye[20]
(a) 0.685*fines+2.2
- -0.525+0.241*specific (b) 5.287e0.028*fines
1 Cohesion equation -
204.5(PI/LL)+56.3(PI/LL)+31 gravity (c ) 27.21 In (fines)-65.28
(d) -0.004(fines)2 + 1.118(fines)
-7.383
(a) 0.219
Coefficient of Correlation (b) 0.242
2 0.861 0.14 0.04
(R2) (c ) 0.19
(d) 0.21
(a) 19.16
Root Mean Square Error (b) 23.77
3 8.33 81.51 33.31 (c ) 18.77
(RMSE)
(d) 18.97
(a) 14.74
Mean Absolute Error (b) 16.34
4 6.08 75.54 26.45
(MAE) (c ) 15.82
(d) 14.60

Nigerian Journal of Technology Vol. 35, No. 2, April 2016 266

 ESTIMATION OF SHEAR STRENGTH PARAMETERS OF LATERITIC SOILS USING ARTIFICIAL NEURAL NETWORK S. D Iyeke et al

Table 5. Comparison of Artificial Neural Network and Other Empirical Method for Angle of Friction Prediction
Factors
S/N ANN Ersoy et al.[8] Roy and Dass [7] Adunoye[21]
Considered

(a)0.404*fines+38.06
Angle of
(b)-16.3 In
1 Friction - -204.5(PI/LL)+56.3(PI/LL)+31 -29.604+34.220*density
(fines)+78.07
equation
(c) 0.001(fines)2 -
0.571(fines)+41.4
Coefficient of (a) 0.0098
2 Correlation 0.805 0.011 0.034 (b) 0.0191
(R2) (c ) 0.0092
Root Mean (a) 33.76
3 Square Error 4.77 21.50 19.35 (b) 11.71
(RMSE) (c ) 451.22
Mean (a) 31.76
4 Absolute 4.34 18.62 16.55 (b) 8.73
Error (MAE) (c ) 451.08

Many empirical methods for shear strength learning rate of 0.2. While the angle of friction had an
parameters prediction of soils are presented in optimal ANN geometry of 3-11-1 i.e., three inputs,
literature. Among these, three have been chosen for eleven hidden layer nodes and a learning rate of 0.4.
the purpose of assessing the relative performance of The results between the predicted and measured
the ANN model. These include the methods proposed shear strength parameters obtained by utilizing ANNs
by Ersoy et al. [8], Roy and Dass [7] and Adunoye were compared with three traditional methods. The
[20,21]. These methods are chosen as the database results obtained demonstrated that the ANN method
used in this work contains most parameters required outperforms the empirical methods considered.
to calculate shear strength parameters by these
methods. The performance of the empirical methods 5. REFERENCES
and the ANN model for the validation set are given in [1] Jain, R., Jain, P. K. and Bhadauria, S. S.,. Computational
Tables 4 and 5 for cohesion and angle of friction approach to predict, soil shear strength.
respectively International Journal of Engineering Science and
Technology, 2(8), 2010, pp. 3874-3885.
4. CONCLUSIONS [2] Too, V. K., 3D Modelling for geotechnical
A total of 83 set of soil data were obtained for the investigation, M.S. Thesis, Department of
study area. The soil test data obtained were scattered Environmental and Biosystems Engineering, School
of Engineering, 2012, University of Nairobi.
over central and southern areas of Delta State.
Generally, the ranges of the angle of internal friction [3] Liu R., Yun S. and Zhang, L.,. Vane shear strength
based stability analysis of slopes in unconsolidated
and cohesion used were 2° - 43 ° and 3 kN/m2--82
soft clay. Transactions of Tianjin University, Vol. 10,
kN/m2, respectively. The multilayer feed forward No. 3, 2004, pp.195-200.
network was used to demonstrate the feasibility of
ANNs to predict the shear strength parameters for
lateritic soils in some areas of Delta State. A Pearson [4] Ajdari, M., Habibagahi,G., Nowamooz, H.,Masrouri, F.,
and Ghahramani, A.,. Shear strength behavior and
correlation analysis was carried out to study the
soil water retention curve of a dual porosity silt-
relative relationship of the factors that affect shear bentonite mixture. Transaction A: Civil Engineering,
strength parameters. The correlation analysis Sharif University of Technology. Vol. 17, No. 6, 2010,
indicated a high level of relationship between plastic pp. 430-440.
limit, liquid limit and plasticity index. Hence only the [5] Tekinsoy, M.A., Kayadelen, C., Keskin , M.S., and
plasticity index was used in the modelling exercise. Soylemez, A.,. An equation for predicting shear
The optimum architecture for the ANN network for strength envelope with respect to matric suction .
Computers and Geotechnics 31: 2004, pp. 589–593.
cohesion was found to be 3-9-1 i.e., three inputs, nine
hidden layer nodes, and one output node with a

Nigerian Journal of Technology Vol. 35, No. 2, April 2016 267

 ESTIMATION OF SHEAR STRENGTH PARAMETERS OF LATERITIC SOILS USING ARTIFICIAL NEURAL NETWORK S. D Iyeke et al

[6] Fredlund, D.G. and Vanapalli, S.K.,. Shear strength of [19] Bravo, E, L.,Suarez, M. H.,Cueto, O. G., Tijskens, E.
unsaturated soils. Agronomy Soil Testing Manual, and Ramon, H.,. Determination of basics mechanical
Agronomy Society of America, 2002, pp. 329-361. properties in a tropical clay soil as a function of dry
bulk density and moisture. Revista Ciencias Técnicas
Agropecuarias, Vol. 21, No. 3: 2012, pp 5-11.
[7] Roy, S. and Dass, G.,. Statistical models for the
[20] Adunoye, G. O.,. Study of relationship between fines
prediction of shear strength parameters at Sirsa,
content and cohesion of soil.British Journal of
India. International Journal Of Civil and Structural
Applied Science and Technology 4(4), 2014: 682-
Engineering, Vol. 4, No. 4, 2014, pp. 483-498
692.
[8] Ersoyl, H., Karsli, M. B., Cellek, S., Kul, B., Baykan, I.,
[21] Adunoye, G. O.,. Fines content and angle of internal
and Parsons, R.,. Estimation of the soil strength
friction of a lateritic soil: An experimental study.
parameters in tertiary volcanic regolith (NE Turkey)
American Journal of Engineering Research. 3(3),
using analytical hierarchy process. J. Earth Syst. Sci.
2014, pp-16-21.
122, No. 6, 2013, pp. 1545–1555.
[22] B. S. 1377. Method of Testing for Civil Engineering
[9] Murthy, V. N. S., Geotechnical Engineering: Principles
Purpose. British Standard Institution, London, 1990,
and Practices of Soil Mechanics, 2nd edn, Taylor and
England.
Francis, 2008, CRC Press, UK.
[23] Goktepe, B., A., SelimAltun, GokhanAltintas and
[10] Alsaleh, M. I., Numerical modeling for strain
Ozcan Tan, Shear strength estimation of plastic clays
localization in granular materials using cosserat
with statistical and neural approaches. Journal of
theory enhanced with microfabric properties. Ph.D.
Building and Environment, Vol. 43,(5), 2008, pp.
Thesis, Dept. of Civil and Environmental Engineering,
849-860.
Louisiana State University, 2004.
[24] Khanlari, G. R., Heidari, M., Momeni, A. A. and
[11] Mollahasani, A., Alavi, A. H., Gandomi, A. H. and
Abdilor, Y. Prediction of shear strength parameters
Rashed, A., , Nonlinear neural-based modeling of soil
of soils using artificial neural networks and
cohesion intercept, KSCE Journal of Civil
multivariate regression methods. Engineering
Engineering, 15(5), 2011, pp 831–840.
Geology, 131–132, 2012, pp.11–18.
[12] Arvidsson J. and Keller, T.,. Comparing
[25] Guyon, I. and Elisseeff, A., An Introduction to
penetrometer and shear vane measurements with
variable and feature selection. Journal of Machine
measured and predicted moldboard plow draft in a
learning research, 3, 2003, 1157-1182.
range of Swedish soils. Soil and Tillage Research,
2011, 111:219-223. [26] Wilby, R. L., Abrahart, R. and Dawson, C. W..
Detection of conceptual model rainfall-runoff
[13] Bishop, T. F. A. and McBratney, A.B., A comparison of
processes inside an artificial neural network.
prediction methods for the creation of field-extent
Hydrological Sciences, 48 (2), 2003, 163-181.
soil property maps.Geoderma, 103: 2001, pp. 149–
160. [27] Lawrence, J. and Fredrickson, J. BrainMaker User's
Guide and Reference Manual, 7th Ed., Nevada City,
[14] Sorensen, K. K. and Okkels, N., , Correlation between
CA: California Scientific Software. 1998.
drained shear strength and plasticity index of
undisturbed over consolidated clays. Proceedings of [28] Shahin, M. A., Jaksa, M. B. and Maier, H. R.,
the 18th International Conference on Soil Mechanics .Predicting the settlement of shallow foundations on
and Geotechnical Engineering, Paris, 2013, pp. 1-6. cohesionless soils using back-propagation neural
networks. Journal of Geotechnical and
[15] Masada, T.,. Shear Strength of Clay and Silt
Geoenvironmental Engineering, Vol. 128, No. 9,
Embankments. Ohio Research Institute for
2002, pp.785–793.
Transportation and the Environment. Ohio
University, Ohio, 2009. [29] Haque, M. E. and Sudhakar, K. V.. ANN back-
propagation prediction model for fracture toughness
[16] Mofiz, S.A. and Rahman M. M.. Shear strength
in microalloy steel. International Journal of Fatigue,
behavior of barind soil on triaxial extension stress
24, 2002, 1003–1010.
path tests.: 11th IAEG Congress, Auckland, New
Zealand 2010, pp 2249-2256. [30] Kim, H., Rauch, A. F. and Haas, C. T. Automated
quality assessment of stone aggregates based on
[17] Cola, S. and Cortellazzo, G., The shear strength
laser imaging and a neural network. Journal of
behavior of two peaty soils. Geotechnical and
Computing in Civil Engineering 18 (1), 2004, 58–64.
Geological Engineering, 2005, 23:679-695.
[31] Singh, T. N., Gupta, A.R. and Sain, R. A comparative
[18] Hajdarwish, A. and Shakoor, A.. Predicting the shear
analysis of cognitive systems for the prediction of
strength parameters of mudrocks.The Geological
drill ability of rocks and wear factor, Geotechnical
Society of London. 2006, IAEG Paper number 607.
and Geological Engineering 24 2006 299–312.

Nigerian Journal of Technology Vol. 35, No. 2, April 2016 268

 ESTIMATION OF SHEAR STRENGTH PARAMETERS OF LATERITIC SOILS USING ARTIFICIAL NEURAL NETWORK S. D Iyeke et al

[32] Das, S.K. and Basudhar, P.K., Prediction of coefficient reservoir parameters using seismic attributes. PhD
of lateral earth pressure using artificial neural Thesis, Department of Geology and Geophysics,
networks, Electronic Journal of Geotechnical Calgary, Alberta.
Engineering,10—Bundle 2005, A, paper 0506.
[36] Erzin, Y., Hanumantha Rao, B., Patel, A., Gumaste,
[33] Shahin, M. A., Jaksa, M. B. and Maier, H. R.,. Artificial S.D. and Singh, D.N.,. Artificial neural network models
neural network applications in geotechnical for predicting electrical resistivity of soils from their
engineering, Australian Geomechanics 36 (1), 2001, thermal resistivity. International Journal of Thermal
49–62. Sciences , 49, 2010, 118–130.
[34] Amegashie, F., Shang, J. Q., Yanful, E. K., Ding, W. and [37] Gunaydin, O., Gokoglu, A. and Fener, M.,. Prediction
Al-Martini, S.,. Using complex permittivity and of artificial soil’s unconfined compression strength
artificial neural networks to identify and classify test using statistical analyses and artificial neural
copper, zinc, and lead contamination in soil, networks. Advances in Engineering Software 41,
Canadian Geotechnical Journal 43, 2006, 100–109. 2010, 1115–1123.
[35] Rusell, B. H.,. The application of multivariate
statistics and neural networks to the prediction of

Nigerian Journal of Technology Vol. 35, No. 2, April 2016 269