Results of Application of Artificial Neural Networks in Predicting Geo-Mechanical Properties of Stabilised Clays-A Review

Download as pdf or txt
Download as pdf or txt
You are on page 1of 25

Review

Results of Application of Artificial Neural Networks in Predicting


Geo-Mechanical Properties of Stabilised Clays—A Review
Jeremiah J. Jeremiah, Samuel J. Abbey *, Colin A. Booth and Anil Kashyap

Faculty of Environment and Technology, University of the West of England, Bristol BS16 1QY, UK;
[email protected] (J.J.J.); [email protected] (C.A.B.); [email protected] (A.K.)
* Correspondence: [email protected]

Abstract: This study presents a literature review on the use of artificial neural networks in the
prediction of geo-mechanical properties of stabilised clays. In this paper, the application of ANNs
in a geotechnical analysis of clay stabilised with cement, lime, geopolymers and by-product cemen-
titious materials has been evaluated. The chemical treatment of expansive clays will involve the
development of optimum binder mix proportions or the improvement of a specific soil property
using additives. These procedures often generate large data requiring regression analysis in order to
correlate experimental data and model the performance of the soil in the field. These analyses involve
large datasets and tedious mathematical procedures to correlate the variables and develop required
models using traditional regression analysis. The findings from this study show that ANNs are
becoming well known in dealing with the problem of mathematical modelling involving nonlinear
functions due to their robust data analysis and correlation capabilities and have been successfully
 applied to the stabilisation of clays with high performance. The study also shows that the supervised
 ANN model is well adapted to dealing with stabilisation of clays with high performance as indicated
Citation: Jeremiah, J.J.; Abbey, S.J.; by high R2 and low MAE, RMSE and MSE values. The Levenberg–Marquardt algorithm is effective
Booth, C.A.; Kashyap, A. Results of in shortening the convergence time during model training.
Application of Artificial Neural
Networks in Predicting Keywords: geo-mechanical properties; regression models; stabilisation of clays; predictive models;
Geo-Mechanical Properties of artificial neural networks
Stabilised Clays—A Review.
Geotechnics 2021, 1, 147–171. https://
doi.org/10.3390/geotechnics1010008

1. Introduction
Academic Editor: Wen-Chieh Cheng
Expansive clays at shallow depths and other poor ground conditions are common
problems faced in geotechnical engineering, and these problematic soils pose enormous
Received: 24 June 2021
Accepted: 11 August 2021
challenges to the proposed infrastructural development [1,2]. Oftentimes, the option of
Published: 17 August 2021
soil replacement is not economical, and there is a need to perform some form of ground
improvement or stabilisation prior to construction [3]. Soil stabilisation may be done chem-
Publisher’s Note: MDPI stays neutral
ically or mechanically, depending on the peculiarity of the problem at hand. However, the
with regard to jurisdictional claims in
treatment of most expansive clays will involve the use of chemical stabilisation. Irrespective
published maps and institutional affil- of the method used, it becomes imperative to investigate the performance of such stabilised
iations. soils using key parameters of the soil such as Unconfined Compressive Strength (UCS),
Maximum Dry Density (MDD), Optimum Moisture Content (OMC), California Bearing
Ratio (CBR), Plasticity index (PI), Liquid Limit (LL), Plastic Limit (PL), grain size, specific
gravity GS , etc. [4–8]. These parameters are most times influenced by several variables
Copyright: © 2021 by the authors.
leading to a multi-variable problem for which researchers have used several statistical
Licensee MDPI, Basel, Switzerland.
methods to analyse, investigate and model trends using laborious mathematical processes
This article is an open access article
involving multivariable regression analyses.
distributed under the terms and The Artificial Neural Network (ANN) is an emerging trend in dealing with ground
conditions of the Creative Commons improvement involving weak clays. ANN is a computing tool with its architectural concept
Attribution (CC BY) license (https:// drawn from that of the biological brain [9–11]. Like other artificial intelligence applications
creativecommons.org/licenses/by/ or machine-enabled activities, the application of ANN in geotechnical engineering with
4.0/). regards to soil stabilisation is to overcome the supposed complexities and shortcomings

Geotechnics 2021, 1, 147–171. https://doi.org/10.3390/geotechnics1010008 https://www.mdpi.com/journal/geotechnics


Geotechnics 2021, 1 148

of traditional methods in other to save time and resources while obtaining more reliable
results. This study explores the usefulness of ANN as a stand-alone machine learning tool
in the stabilisation of expansive clays and evaluates its capability in the development of
useful mathematical models for the performance evaluation of key properties of stabilised
clays. Previous experimental studies are reviewed with a view of assessing the state-of-art
in terms of model development, training and performance evaluation.

1.1. Artificial Neural Network


ANN as a branch of artificial intelligence is simply an automated optimisation sys-
tem capable of learning the relationship and inter-dependencies between multiple input
variables of a given system and modelling such relations (trends and patterns) in the form
of mathematical functions for easy prediction [12]. ANN has been successfully used in
the study of complex systems to identify patterns and model real-life problems relating to
complex behaviours involving nonlinear functional relations. The capability of ANN to
discover the mapping between several domains of data has drawn the interest of many
researchers in geotechnical engineering [13]. ANNs are classified based on numerous
criteria such as the learning condition (supervised and unsupervised networks), based on
model topography (feedforward or recurring networks), based on a number of hidden
layers (shallow or deep networks), based on training algorithm (Back-Propagation Net-
works, Hopfield Networks, Self-Organizing Map Networks [14]. This paper simplifies the
underlying concepts of back propagation ANN models and explores its applicability in
modelling the behaviour of stabilised clays viz a viz predicting the response of key soil
parameters in other to clear the wide-spread complexities and misconceptions associated
with the method and to encourage its use in soil stabilisation problems for more reliable
solutions.

1.2. Components of Artificial Neural Network


Neurons and Edges
ANN building blocks are a collection of neurons (nodes) and links mimicking the
biological neural network, as shown in Figure 1. The neurons are linked to other neu-
rons by edges and are connected to others so that results from preceding neurons might
automatically become inputs for succeeding ones, thereby creating the network. These
neurons are the data collection or processing points in the network. Here, signals (input)
are processed and transferred to other neurons through the connecting links with each
neuron generating a unique output that may become inputs to multiple neurons. In this
current subject area of application, these inputs would be laboratory results of key soil
parameters described as the dependent variables. The input value of a given neuron is
simply obtained by computing the weighted sum of the inputs from connected neurons
with the addition of a bias [15,16]. This output of the weighted summation then becomes
the input for the activation function- a linear or non-linear function [17,18].
Edges are the links or connections between neurons and convey signals with associated
weights depending on the influence of the input from such a link on the output of a given
neuron [19]. Inputs parameters with greater importance are assigned a higher weight than
those with lower importance [20]. For example, in a soil classification problem, the weights
will be dependent on the contribution of the features in determining the class of soil [21].
In a typical perceptron, as in Figure 2, the connection weights can be represented as Wj,
which describes the importance of the connection.
Geotechnics 2021, 1 149
Geotechnics 2021, 1, FOR PEER REVIEW 3

(a) (b)
Figure 1. (a,b) Biological Neural Network [10].

Edges are the links or connections between neurons and convey signals with associ-
ated weights depending on the influence of the input from such a link on the output of a
given neuron [19]. Inputs parameters with greater importance are assigned a higher
weight than those with lower importance [20]. For example, in a soil classification prob-
lem, the weights will(a)be dependent on the contribution of the features
(b)
in determining the
class
Figureof
1. soil
(a,b) [21]. In a Neural
Biological typicalNetwork
perceptron,
[10]. as in Figure 2, the connection weights can be
Figure 1. (a,b) Biological Neural Network [10].
represented as Wj, which describes the importance of the connection.
Edges are the links or connections between neurons and convey signals with associ-
ated weights depending on the influence of the input from such a link on the output of a
given neuron [19]. Inputs parameters with greater importance are assigned a higher
weight than those with lower importance [20]. For example, in a soil classification prob-
lem, the weights will be dependent on the contribution of the features in determining the
class of soil [21]. In a typical perceptron, as in Figure 2, the connection weights can be
represented as Wj, which describes the importance of the connection.

Figure2.2.Typical
Figure TypicalNeuron
Neuronshowing
showingAssociated
AssociatedEdge
EdgeWeights.
Weights.

2.2.Materials
Materialsand andMethods
Methods
The
Thepropagation
propagation function
function [20][20]computes
computes the the
weighted
weightedsummation
summation of inputs from each
of inputs from
preceding neuron with the addition of a bias term to create an
each preceding neuron with the addition of a bias term to create an activation that is uti- activation that is utilized
by
lizedtheby activation function
the activation in computing
function in computing the output signal.
the output ForFor
signal. a typical case
a typical casesuch
suchasas
represented
represented ininFigure
Figure 2, GG(I(Ij )j)isisthe
theactivation function where I is the propagation function
Figure 2. Typical Neuron2,showing activation
Associated Edgefunction
Weights.where jIj is the propagation function
defined
definedininEquation
Equation(1)(1) [22,23].
[22,23]. TheThe activation
activation function
functioncould be any
could of binary
be any stepstep
of binary function
func-
(Equation
tion (3)),
(Equation logistic
2. Materials and Methods(3)), function
logistic (Equation
function (4)),
(Equationhyperbolic
(4)), function
hyperbolic (Equation
function (5)), Gaussian
(Equation (5)),
function
Gaussian (6), etc. [12].
function (6), function
etc. [12]. [20] computes the weighted summation of inputs from
The propagation
Ij = Øj +∑ni =1 Wji × Xi (1)
each preceding neuron with the addition I Ø of+a ∑bias term
W toXcreate an activation that is uti-(1)
W ji is the
lized associated
by the activationweights
function of the connections.
in computing the output signal. For a typical case such as
XW is the associated
i are the input
represented data. weights
in Figure 2, G (Ij) isofthe theactivation
connections.function where Ij is the propagation function
ØX j isare
defined thethe
in input
threshold
Equation data.
or (1)
bias.[22,23]. The activation function could be any of binary step func-
nØ is is
the the threshold
number of or bias.
inputs.
tion (Equation (3)), logistic function (Equation (4)), hyperbolic function (Equation (5)),
The connection
Gaussian function (6), weights,
etc. [12]. Wji , are continuously adjusted for each training iteration step
using the expression below [24].
I Ø + ∑ W X (1)
m −1
W is the associated weights
m
Wji of=the + βε j Xi + ΩδWji m−1
Wjiconnections. (2)
X are the input data.
In Equation (2), δWji is the change in the weight of a given connection, β, ε j and Ω
Ø is the threshold or bias.
represent the learning rate, the error in a neuron’s output and the momentum, respectively,
Geotechnics 2021, 1 150

while m is the iteration number. The momentum is used in factoring the weight correction
and plays an important role in model training in the sense that too large values cause the
model to skip the optimum values of the weight, causing oscillations. On the other hand,
too small values will result in delayed convergence between predicted and experimental
values [25].
An activation function, such as the binary step function, expresses the output of a
given neuron as either of 0 or 1 depending on the input x and is given here as Equation (3).

 0, when Ij < 0
g Ij = (3)
1, when Ij > 0

The binary function due to the gap in the output (only 0 and 1) is unsuitable for
training algorithms, which rely on the derivatives of the activation functions [26]. The
logistic function also limits the output of a neuron to 0–1 and is expressed by [27].
 1
g Ij =   (4)
1 + e− Ij

The hyperbolic tangent function, as shown in Equation (5), is also a commonly used
activation function in ANN [28], and it limits the output of a neuron to −1 and 1. The
Gaussian function is another commonly used activation function that defines the nor-
mal distribution of a given random input variable about its mean, [29] as presented in
Equation (6).
 e Ij − e− Ij
g Ij = I (5)
e j + e− Ij
  ( I j − µ )2
 1 −( )
g Ij = √ e 2σ2 (6)
σ 2π
where σ and µ are the standard deviation and mean, respectively.

2.1. ANN Architecture


Deciding the ANN topography is a critical part of the model development and in-
volves an iterative trial and error process (training) [17,30–32]. In most studies, an initial
model topography is assumed and trained while monitoring the performance of the
model using predefined statistical measures such as coefficient of determination (R2), root
mean square error (RMSE), mean average error (MAE) and mean square error (MSE).
The hyperparameters are continuously modified, and the model retrained until an op-
timum model architecture is obtained with the lowest error and highest R2 [11,33–35].
This training, in simple terms, is “showing the network an example” of the problem us-
ing experimental input and output data. Many training algorithms exist; quasi-newton
backpropagation, Bayesian regularization backpropagation algorithm, gradient descent,
Levenberg–Marquardt optimization, etc., but the process is similar and begins with feeding
the model with a quality dataset and allowing the system to process this data in order
to learn the relationship between the variables and hence generate weighted associations
between the data within the network and predict the result. The predicted result is then
compared with the experimental result to evaluate the error, which is then used to modify
the weights of the connections by a reverse error minimization process using a chosen
cost function. The process is repeated until there is an insignificant change in the output
of the cost function [36,37]. Based on the performance of various models with different
hyperparameters, the best model topography is then selected.

2.2. Feedforward and Recurring Networks


The ANN topography is such that the neurons are grouped into layers, namely input
layers and output layers, and in some cases, there is a need to include hidden layers
Geotechnics 2021, 1 151

between the input and output layer, making a multilayered perceptron neural network
model in order to create sufficient degrees of freedom to avoid overfitting. The hidden layer
could be made one layer (shallow networks), as shown in Figure 3, or multiple layers (deep
neural network, DNN) [38]. In addition, the connections between layers could be in such a
way that a neuron in one layer could be connected to all neurons in the succeeding layer
and is said to be fully connected, resulting in a larger number of neurons in the succeeding.
Additionally, the models could be organized such that multiple neurons in a layer are
connected to a single neuron of the succeeding layer. The latter condition is said to be a
pooled connection and is synonymous with a lesser number of neurons in the succeeding
layer [39], even though one may be tempted to believe that a larger number of neurons
will always result in a better prediction using ANNs. However, the optimum number of
neurons will depend on several factors such as the amount of data and the complexity of
the relationship. In certain types of NNs, such as the Deep Neural Network, the number
of neurons has a lesser effect on the overall performance of the network than the number
of layers [12], and DNNs with more hidden layers have been shown to yield more results
than shallow networks (network architecture with a single hidden layer). However, the
number of hidden layers to be used in each network will depend on the complexity of
the mapping between input and output domain, the quality and the amount of data [40].
Additionally, even though it is expected that a greater number of experimental datasets
used in training the model will improve the performance of the model, recent studies have
shown that it might be advisable to use fewer experimental datasets of high quality than
a large amount of experimental data which may be prone to errors [12]. Moreover, the
quality of the output is dependent on how the input database is utilized in the training.
In terms of the way data is transferred from one layer of the network to another, one
can generalize that there are two broad categories of ANN architecture—the recurring
network and feedforward network. In the recurring network, there is a connection between
neurons of a given layer and that of preceding and/or succeeding layers, forming a loop
and allowing an input to be processed many times by the same neuron. Conversely, in the
Geotechnics 2021, 1, FOR PEER REVIEW
feedforward network, neurons in each layer are only connected to neurons in other layers 6
as presented in Figure 4.

Figure
Figure 3.3.AATypical
Typical ANN
ANN Architecture
Architecture Showing
Showing Neurons
Neurons and
and Layers.
Layers.
Geotechnics 2021, 1 152

Figure 3. A Typical ANN Architecture Showing Neurons and Layers.

(a) (b)
Figure 4. (a) Recurring Network; (b) Feed Forward Network.
Figure 4. (a) Recurring Network; (b) Feed Forward Network.

Data Preprocessing
The input
input dataset
datasetusedusedtotodevelopdevelop ANN ANN models
models forfor geotechnical
geotechnical engineering
engineering prob-
problems
lemscomprise
will will comprise of various
of various rangesranges of input of variables
input variables
because because
the ANN the neurons
ANN neurons
connection con-
weights are representatives
nection weights are representatives of the importance of the variables.
of the importance The weights
of the variables. of the connec-
The weights of the
tions are influenced
connections by the by
are influenced Euclidean
the Euclidean distance whichwhich
distance for any forgiven points,
any given y1 (x11y,1 x(x2111,, xx31,
points, 21,

x. 31, , xn1
. . …, xn1 andyy22 (x12
) )and 12,,xx
2222 x32,
, x, 32, …,. x. .n2,) xinn2a) data
in a space
data space
can becan be expressed
expressed as Equation
as Equation (7). There- (7).
Therefore,
fore, in order in to
order
ensureto ensure
that proper that significance
proper significance
is attributedis attributed to theitfeatures,
to the features, is imperative it is
imperative
to to scale the
scale the features. features.
Feature scaling Feature
can bescaling
achieved can be achieved
using methods using
such asmethods such as
standardization
standardization
(Z-score (Z-score
normalization) ornormalization)
the max-min normalization or the max-min method.normalization method. For
For standardization, thestan-fea-
ture 𝑋 is expressed in its istandardised form as in Equations (7) and (8) below.
dardization, the feature X is expressed in its standardised form as in Equations (7) and (8)
below.
(𝑥 ) (𝑥 ) (𝑥 𝑥 2)
de𝑑= = ( x12 − − x11𝑥)2 + + ( x22 − − x21𝑥)2 + +. . .⋯++( xn2 −−xn1
q
)
X𝑋i −−
µ 𝜇
(7)
(7)
X
𝑋 = σ
s =
𝜎
Xi − Xmin
Xnorm = 𝑋 − 𝑋 (8)
𝑋 =Xmax − Xmin (8)
𝑋 − 𝑋
In Equations (7) and (8), Xs is the standardised value of Xi , µ is the mean and σ is the
standard deviation. The input variables can also be scaled using the max-min normalisation
to achieve the same purpose using the expression, where Xnorm is the normalised value
and Xmin and Xmax are the minimum and maximum values, respectively.

3. Selecting Design Parameters for ANN in Soil Stabilisation


The training of the neural network is actually the process of selecting or ‘designing’ the
best network model parameters (hyperparameters). However, there is a start point where a
first architecture is proposed. Selecting the right number of neurons in the hidden layer is
critical as it influences the performance of the network. Too few a number of neurons in
the hidden layer can lead to underfitting. In this case, the training data presents a more
complex problem than the network is modelled to handle. For some problems, increasing
the number of neurons by introducing additional features can make learning easier for the
model and resolve the problem. In some other cases, for example, in some geotechnical
applications, the input parameters of interest may have been predetermined and measured,
and this option may not be feasible. Alternatively, increasing the number of hidden layers
and neurons may be helpful, but again, if the number of neurons becomes excess, this
leads to overfitting. In this case, the model is training on less complex data than it is
designed to analyse. The effect is that the model is unable to properly generalize on new
data set outside the training data as the weights are not optimally adjusted. Additionally,
an excessively high number of neurons in the hidden layer can extend training time and
lead to poor training even with a sizable database for training. The goal, therefore, is to find
a balance. However, this is not a straightforward process. A good step would be to make a
Geotechnics 2021, 1 153

good initialization of the network parameters, and there are several ideas concerning how
these parameters can be initialized. Amongst the huge suggestions that are available and,
of course, effective under different conditions, one simple empirical rule for selecting the
number of neurons in the hidden layer is to use the mean of the number of neurons in both
input and output layers. Some other idea involves taken about 60–70% of the total input
and output neurons. In general, the idea is to provide a reasonable start so that during the
training, the model can be optimised or pruned, and redundant neurons can be removed
based on the assigned weights while keeping track of the performance. However, for most
regression analyses relating to stabilisation, one hidden layer has been found very effective.
In rare cases, two hidden layers have been used, but there are seldom cases where over
two layers have been needed in developing reliable predictive models. As shown in the
succeeding section, almost all the applications in soil stabilisation have utilised one hidden
layer, with one or two utilizing more than one hidden layer.

3.1. Training, Validation, and Testing


As mentioned in the earlier section, the training of most neural networks applied to
modelling geomechanical properties of stabilised soils is done under supervised conditions.
In this type of training, the supplied data are partitioned, and a part (training dataset) is
utilised in learning the relationship between the variables, thereby providing the initial
weights. This sample is continuously fed to the network with a view to understanding
the data rather than recognizing it. If the network learns progressively, it converges with
reduced error after each iteration until a pre-defined error range is attained. The quality
of the training dataset influences the convergence of the model [41]. A dataset may lack
the necessary independent variables required for the model to understand the data and
hence can lead to non-convergence. A very small sample space can lead to the network
memorizing rather than learning. Hence, it is important that part of the data is separated
to be used in evaluating the training. This is the validation dataset. The major aspect of
developing a suitable model would then be to continuously monitor and tweak the number
of neurons in the hidden layer, or the number of hidden layers, modify the activation
function or even the training algorithm [42]. The model validation utilizes the successive
trial of the trained model on the validation dataset [43]. This is an unbiased evaluation of
how well the model understands the training data. The final test of the model’s predictive
ability is carried out on the test dataset, which was never seen by the model. In some
cases, the data are continuously partitioned into two as in cross-validation. The dataset is
switched and utilised for training and validation in a crossed pattern.
The performance of models is usually evaluated by using statistical measures such as
the coefficient of determination (R2 ), the mean absolute error (MAE), the root mean square
error (RMSE), the mean square error (MSE) and others. The R2 , MAE, RMSE and MSE
expressions are defined in Equations (9), (10), (11) and (12), respectively.
    2
∑1n yexp (i) − yexp y pre(i) − y pre
R2 =  2 2
 (9)
∑1n (yexp (i) − yexp ) ∑1n (ypre(i) − y pre )

I 
MAE = yexp (i) − ypre(i) (10)
n
s
n  2
1
RMSE = x ∑ yexp (i) − y pre(i) (11)
n 1
n 2
1 
MSE =
n
x ∑ yexp (i) − y pre(i) (12)
1

For which yexp (i) and ypre(i) are experimental and predicted values of a given depen-
dent variable, while y pre and yexp are the mean values of the predicted and experimental
values.
Geotechnics 2021, 1 154

3.2. Estimating the Amount of Training Data


Determining the sample size required for successful model training is a vital step
in successful model development. A common start-off point is the “rule of ten”, which
proposes that the training sample size is taken as not less than ten times the number of
network parameters. The number of parameters may be estimated as the number of edges
or connections, including biased neurons. It is expected that the performance of the model
would improve with increasing sample size following a power function up to a point
is reached where there is no significant increase in performance. Usually, in practical
situations, the data set is split into the ratio of 70%:30% or 80%:20%, where the higher
percentage is that of the training sample space. Although this split is only important with a
relatively low sample size. The underlying idea is to make available sufficient example
data with which the network is trained and evaluated. Too little training data will result in
a higher variance of the network parameters. Additionally, two few testing data will create
higher variance during the evaluation of the performance of the model.

4. Application of ANN in Predicting the Properties of Stabilised Clays


The variability of soils following uncontrollable and imprecise chemical and mechan-
ical processes of their formation makes it very complex to predict their behaviour even
in the natural state. Several mathematical and graphical models have been proposed for
use in modelling the post-stabilisation behaviour of treated clays through the prediction of
various engineering parameters for the purpose of design and construction. These methods
rely on laboratory results of a few samples with which in-situ post-stabilisation behaviour is
to be predicted. However, in many cases, the post-stabilisation behaviour of any soil will be
dependent on multiple variables such as curing time, curing duration, soil type, binder con-
tent, curing temperature, moisture content, compaction method and effort, plasticity, etc.,
which are key and influence the dependent variables [44,45]. The large variability of input
parameters, the extensive laboratory experiments required and the unknown relationship
between the variables put together makes it even more complex to predict the behaviour
of the soils [46,47]. In other to simplify the problem, many studies tend to concentrate on
a set of few parameters and employ simple mathematical models to map the domains of
input to our variables. An example of such simplified mathematical models is the multiple
linear regression model, which, for a given set of features x1n , x2n , x3n , . . . , xmn , can be
related to its dependent variable as stated in Equation (13) below [23].

y= β 0 + β 1 x1n + β 2 x2n + β 3 x3n + . . . β 1 xmn + ε (13)

where β 0 , β 1 , β 3 , . . . β n are the coefficients, and ε is the error.


ANN application has shown good results in terms of the prediction of engineering
parameters of stabilised soils for various purposes. ANN’s ability to learn the relations
between a wider and more complex set of experimental variables and map these variables
to the target output domain using well-adjusted weights makes it a more reliable tool in the
prediction of the in-situ post-stabilisation behaviour of treated clay soils. In addition, the
ability of ANNs to simultaneously handle multiple dependent and independent variables
using the same experimental dataset and its adaptability in finding correlations for highly
non-linear data, which are characteristic of many civil engineering problems, makes it more
advantageous over traditional regression analysis [48,49]. A review of studies modelling
various soil properties is presented in the subsequent sections. The review is in a bid to
explore the input parameters considered in the study, the number of data utilised in model
development, the training algorithm utilised, the hyperparameters of the model, the model
performance and, finally, the results or predictive models developed.

4.1. Unconfined Compressive Strength


ANN has been employed in tracking and modelling the UCS of geopolymer stabilised
clays [50]. In the study, ground granulated blast furnace slag (GGBS) and pulverised fly ash
Geotechnics 2021, 1 155

(PFA) were considered as binders for the improvement of the compressive strength of three
clays with varying properties. The soil characteristics such as liquid limit (LL), plastic limits
(PL), etc., were evaluated according to the Indian standards and, in combination with key
experimental variables, were utilised as input data for the ANN model development. Eight
input variables namely, LL, plasticity index (PI), GGBS content, PFA content, the molarity
of alkaline activator used (M), the ratio of the activator to the binder, the ratio of sodium to
aluminium in the activator–binder mixture (Na/Al), the ratio of silicon to aluminium in
the activator–binder mixture (Si/Al), and 28 days UCS, were considered. A multi-layered
perceptron ANN model was chosen with one hidden layer to study the experimental data.
The optimum architecture was selected by varying the number of neurons in the hidden
layer while evaluating the performance of the model and was obtained as one hidden layer
Geotechnics 2021, 1, FOR PEER REVIEW
with nine neurons. The ANN showed a better ability to learn the relationship of the data10
set, as shown in Figure 5.

Figure5.5.Experimental
Figure Experimentalversus
versusANN-Predicted
ANN-PredictedUCS
UCSvalues
values[50].
[50].

FromFigure
From Figure5,5,ititisisobvious
obviousthat
thatalmost
almostall alldata
datapoints
pointsare
arewithin
withinthethe99%
99%confidence
confidence
limit.The
limit. Theline
line
of of
bestbest fitalmost
fit is is almost aligned
aligned withwith theofline
the line of equality
equality at whichat which all experi-
all experimental
observations and predicted
mental observations values are values
and predicted the same.areTable 1. compares
the same. Table 1.thecompares
performance
the of the
perfor-
ANN
mancemodel with
of the ANN thatmodel
of a multivariable
with that ofregression analysis
a multivariable (MVR) onanalysis
regression the same data set.
(MVR) It
on the
issame
obvious
datathat
set.ANN performed
It is obvious better
that ANN at performed
correlating better
the independent
at correlatingvariables and their
the independent
influence
variablesonandthetheir
UCSinfluence
of the stabilised soils.of the stabilised soils.
on the UCS

Table Table 1. Statistical


1. Statistical Evaluation
Evaluation of the of the Performance
Performance of ANN
of ANN and MVR
and MVR models
models [50]. [50].

StatisticalStatistical
Parameter Parameter
Model DatasetModel Dataset
2
R2 R MSE MSE MAE
MAE (%)
(%)
Training data 0.992
Training data 0.992 0.34 0.34 3.65
3.65
ANN ANN
Testing data 0.964
Testing data 0.964 1.50 1.50 8.348.34
Training data 0.828
Training data 0.828 7.24 7.24 19.20
19.20
MVR MVR
Testing data 0.808
Testing data 0.808 8.04 8.04 19.26
19.26

Expansive soils are known for their vulnerability to significant volume changes with
moisture due to their high plasticity properties. Table 2 shows the classification of expan-
sive clays according to the Building Research Establishment (BRE) based on a modified
plasticity index.
Geotechnics 2021, 1 156

Expansive soils are known for their vulnerability to significant volume changes with
moisture due to their high plasticity properties. Table 2 shows the classification of expansive
clays according to the Building Research Establishment (BRE) based on a modified plasticity
index.

Table 2. Classification of shrink-swell clays [51].


0
Modified Plasticity Index IP Volume Change Potential (VCP)
>60 Very high
40–60 High
20–40 Medium
<20 Low

Geotechnics 2021, 1, FOR PEER REVIEW In another study, ref. [52] employed a feedforward ANN model in studying the UCS 11

of expansive clays stabilised with cement-kiln dust (CKD). The liquid limit, plastic limit
and plasticity index of the clay were determined to be 48.2%, 27.2% and 21%, respectively.
Theexpansive
The expansiveclayclayisischaracterised
characterisedbybya afree freeswell
swellofof80%.
80%.TheThenatural
naturalandandstabilised
stabilisedclay
clay
samples were subjected to compaction and UCS test in accordance
samples were subjected to compaction and UCS test in accordance with [53,54]. Three with [53,54]. Three
ANNmodels
ANN modelswere
weredeveloped
developedwith witha atotal
totalofofeight
eight input
input variables,
variables, namely
namely specificgravity
specific gravity
(Gs),linear
(Gs), linear shrinkage
shrinkage (LS(L ), coefficient
), Scoefficient of uniformity
of uniformity (CU),(Ccoefficient
U ), coefficient of gradation
of gradation (CC), (C
LL,C ),
LL,optimum
PL, PL, optimummoisturemoisture
content content
(OMC), (OMC), and maximum
and maximum dry density
dry density (MDD)(MDD) forUCS
for three three
UCS outputs
outputs (7, 14
(7, 14 and 28 and
days).28 A days). A total
total of 72 dataof 72 data data
sample sample data
were were utilised
utilised in modelindevel-
model
development. The ANN model topography consisted of one input
opment. The ANN model topography consisted of one input layer with eight neurons and layer with eight neurons
and
one one output
output neuronneuron (the value).
(the UCS UCS value).
However,However, the number
the number of hidden
of hidden layers
layers was was varied
varied in
in order to determine the optimum number of neurons in the hidden
order to determine the optimum number of neurons in the hidden layer. Figure 6 shows layer. Figure 6 shows
theperformance
the performanceofofthe thetrial
trialmodels.
models.

0.012
7 Days UCS 14 Days UCS 28 Days UCS
0.01

0.008
MSE

0.006

0.004

0.002

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Number of Hidden Layers

Figure 6. 6.
Figure The selection
The ofof
selection optimum number
optimum ofof
number neurons ininhidden
neurons hiddenlayer
layer[52].
[52].

Theoptimum
The optimummodel
modelwaswasfound
foundtotobebeofofnine
nineneurons
neuronsininthethehidden
hiddenlayer
layerbased
basedonon
lowestMSE.
lowest MSE.The Theauthors
authorsofof[52]
[52]suggest
suggestthat
thatthe
theMSE
MSEisisa amore
morereliable
reliableparameter
parameterfor for
networkselection
network selectionwhen
whenRRvalues
values alone
alone become insufficient
insufficientfor
foroptimum
optimumnetwork
networkselection.
selec-
TheThe
tion. results of the
results of analysis showed
the analysis that that
showed the ANN
the ANNmodel was able
model was to predict
able the variation
to predict the vari-of
UCSofasUCS
ation a function of the predictor
as a function variables
of the predictor with awith
variables highacorrelation coefficient,
high correlation as seenasin
coefficient,
Figure 7.
seen in Figure 7.
The optimum model was found to be of nine neurons in the hidden layer based on
lowest MSE. The authors of [52] suggest that the MSE is a more reliable parameter for
network selection when R values alone become insufficient for optimum network selec-
Geotechnics 2021, 1
tion. The results of the analysis showed that the ANN model was able to predict the vari-157
ation of UCS as a function of the predictor variables with a high correlation coefficient, as
seen in Figure 7.

Geotechnics 2021, 1, FOR PEER REVIEW 12

Figure7.7.Observed
Figure ObservedVersus
VersusPredicted
PredictedUCS
UCS[52].
[52].

The
Thestudy
studyreveals
reveals aa high correlation
correlationbetween
betweenexperimental
experimental values
values andandANN ANN pre-
predicted
dicted
values. values. The high
The high correlation
correlation coefficient
coefficient with with low RMSE
low RMSE confirms
confirms the performance
the performance of the
ofmodel.
the model.
Moreover,
Moreover, the authors
the authorsof of
[55]
[55]conducted
conducted anan
experimental
experimental investigation
investigationon onthethe
suita-
suit-
bility andand
ability performance
performance of kaolin
of kaolin clay stabilised
clay with
stabilised withflyfly
ashash(PFA), rice
(PFA), husk
rice husk ash (RHA)
ash (RHA)
and
and cement.
cement. The
The improvements
improvements inin the
thesoil
soilwere
wereascertained
ascertainedbybyconsidering
consideringthe theincrease
increase
inin
thethe UCS.
UCS. The
The LL,LL, PL,
PL, and
and PIPIofof the
the unstabilised
unstabilised kaolin
kaolin clay
clay were
were determined
determined asas 43.3%,
43.3%,
19.5%and
19.5% and23.8%.
23.8%.The The input
input variables
variables considered
considered forfor
thethe correlation
correlation were
were claycontent,
clay content,
RHA
RHA content,
content, cement,
cement, PFAcontent
PFA contentand andcuring
curingduration.
duration.The The number
number ofofneurons
neuronsininthe the
hidden
hidden layer
layer waswas selected
selected manually
manually and
and variedwhile
varied whileevaluating
evaluatingthe theperformance
performanceofofthe the
model using the MSE. Figure 8 shows the variation of MSE and the number of neurons inin
model using the MSE. Figure 8 shows the variation of MSE and the number of neurons
the
the hidden
hidden layer.
layer.

200

150
MSE

100

50

0
0 2 4 6 8 10 12
Number of neurons in hidden layer

Training Testing

Figure
Figure 8. 8.
TheThe selection
selection of of
thethe optimum
optimum number
number ofof neurons
neurons inin hidden
hidden layer
layer [55].
[55].

As seen in Figure 8, the optimum model performance was found at 10 neurons in


As seen in Figure 8, the optimum model performance was found at 10 neurons in the
the hidden layer based on Levenberg–Marquardt algorithm. The results of the analysis
hidden layer based on Levenberg–Marquardt algorithm. The results of the analysis using
using the optimized model showed a good correlation between the datasets. The predictive
the optimized model showed a good correlation between the datasets. The predictive
model developed from ANN is given in Equation (14).
model developed from ANN is given in Equation (14).
. . .
= −1.003 + 1.568 + 2.477 + 2.287 + 6.731 + 0.017𝐶 (14)

where 𝑈𝐶𝑆 is the dependent variable to be estimated and represents the UCS for a given
combination of independent variables. A comparison of ANN performance with multi-
variable regression analysis further proves ANN’s advantage, as shown in Table 3. ANN
Geotechnics 2021, 1 158

Clay 2 PFA 0.5 RH A 0.5 Cement 0.5


       
UCSm
= −1.003 + 1.568 + 2.477 + 2.287 + 6.731 + 0.017Cd (14)
UCSC7 100 100 100 100

where UCSm is the dependent variable to be estimated and represents the UCS for a
given combination of independent variables. A comparison of ANN performance with
multivariable regression analysis further proves ANN’s advantage, as shown in Table 3.
ANN has been described as a black box since it is difficult to predict the way the predictive
model will be selected. However, sensitivity analysis can be carried out in determining the
influence of the input variables on the target variable.

Table 3. Statistical evaluation of the performance of ANN and MVR models [55].

Statistical Parameter
Model Dataset
R2 MSE RSME
Training data 0.9813 0.0395 0.1987
ANN
Testing data 0.9714 51.34 7.1651
MVR 0.8870 68.7603 8.2921

In a separate study, [56] investigated the performance of stabilised soils with a view
to predict the effects of dynamic disturbance on already stabilised soils. The study was to
understand the effects of transportation and compaction on strength recovery of already
setting stabilised soft clay. Hence, stabilised soil samples were agitated and compacted
after curing for some days to simulate the disturbed sample. The UCS was utilised as
a performance evaluation parameter. The soils were treated with varying quantities of
cement. A kaolin clay with liquid limit, plastic limit and plasticity index of 77.5%, 30.3%
and 47.2%, respectively. Ordinary Portland cement was used as a binder for different
ratios by mass of dry soil. The UCS test was conducted according to ASTM D2 166-06. It
was required to estimate the influence of disturbance on the UCS in terms of an increase
factor. ANN was employed in correlating the predictive variables and the dependent
variable. A total of 80 experimental data points were generated from combinations of input
variables, namely curing duration and compaction energy. The trial and error method
was utilised in determining the network topography. However, the performance of one
hidden layer was unsatisfactory due to high errors and low R2 . The optimum network
model was taken as two hidden layers with eight neurons each. The result of the analysis
showed a good understanding of the relationship and modelling of the increase factor
with a high coefficient of determination and low error. As seen in Figure 9, there is a close
approximation of the UCS by the ANN model. The prediction and experimental values
are clustered around the line of equality for all three conditions of training, validation
and testing.
The UCS of cement stabilised clays was investigated by [57] in order to find a rela-
tionship between the several key variables. Different types of cement were utilised to
study the effects of cement type on the UCS. The soil under investigation was collected
at various depths below the ground surface to also study the effect of confining pressure
on the strength of the stabilised soils. Three machine learning algorithms were utilised.
ANN was used in correlating the predictive variables. A total of 216 experimental data
points were generated from various combinations of fourteen input variables, namely soil
type, moisture content (MC), bulk density (We), the mass of cement (CM), sample diameter
(DI), the length of the sample (L), the cross-sectional area of the sample (CA), the volume
of the sample (SV), the depth of sample collection (D), the mass of the sample (MS), sample
density (DS), curing condition (CD), curing time and cement type (CT). ANN training was
done using the Quasi-Newton method, Stochastic Gradient Descent and Adam in order to
select the optimal hyperparameters for the ANN topography. The dataset was divided in
the ratio of 80% and 20% for training and testing. Performance evaluation of the developed
47.2%, respectively. Ordinary Portland cement was used as a binder for different ratios by
mass of dry soil. The UCS test was conducted according to ASTM D2 166-06. It was re-
quired to estimate the influence of disturbance on the UCS in terms of an increase factor.
ANN was employed in correlating the predictive variables and the dependent variable. A
Geotechnics 2021, 1 159
total of 80 experimental data points were generated from combinations of input variables,
namely curing duration and compaction energy. The trial and error method was utilised
in determining the network topography. However, the performance of one hidden layer
was unsatisfactory
ANN model showeddue to high
a good errors and
correlation. low R
ANN
2. The optimum network model was taken
performed better than the other machine
as two hidden
learning layers
algorithms. with
The eight
high neurons
number each. The
of neurons in result of thelayer
the hidden analysis
mayshowed
be due toa good
the
understanding
multiple of theofrelationship
dimensions and modelling
the input vector of thedataset.
in the sample increaseThefactor withshowed
study a high coeffi-
that
cientcould
ANN of determination and low error.
be utilised effectively As seen in Figure
for stabilisation 9, there
problems is a close
to track approximation
and model a wide
of theofUCS
range by the ANN
independent model. The prediction and experimental values are clustered
variables.
around the line of equality for all three conditions of training, validation and testing.

Figure9.9.Predicted
Figure Predictedand
andexperimental
experimentalobservations
observations[56].
[56].

ToThe UCS ofthe


alleviate cement stabilised
difficulties clays was
associated withinvestigated
the need for by [57] in orderexperimental
continuous to find a rela-
tionship between
determination of UCS,the ref.
several key variables.
[58] employed Different
machine types
learning in of cement and
analysing weremodelling
utilised to
study
the the effects of
performance of stabilised
cement type on the UCS.
dredged The soil
sediments. Aunder
total ofinvestigation was collected
51 experimental datasetsat
various
were depths
collated from below the ground
existing literaturesurface
for thetodevelopment
also study theofeffect of confining
the ANN model. pressure
The inputon
the strength
predictive of the stabilised
variables soils. Three
were percentage machine
moisture learning
content (MC),algorithms
cement were utilised.
content, ANN
air foam
was used
content, andinwaste
correlating
fishing the
net predictive
content. variables.
The ANNAmodel total of
to 216 experimental
topography data points
was initialized
towere
two generated
hidden layers fromwhile
various combinations
varying the number of fourteen
of neurons input variables,
in order namely
to find soil type,
the optimum
moisture
model. Thiscontent
approach (MC),has bulk density
been taken in (We), the stabilisation
most soil mass of cement (CM), sample
applications. diameter
The optimum
ANN
(DI), architecture
the length ofwas made up
the sample ofthe
(L), twocross-sectional
hidden layersarea withof12 and
the 10 neurons
sample based
(CA), the on
volume
training using the
of the sample (SV),Levenberg–Marquardt algorithm.
the depth of sample collection (D),From the results
the mass of the analysis,
of the sample the
(MS), sample
model
densityhas been
(DS), able condition
curing to find the trend
(CD), within
curing timetheand
available
cement training
type (CT). dataset. However,
ANN training was
for certain values of the water content, the error was high. For
done using the Quasi-Newton method, Stochastic Gradient Descent and Adam in order the model to generalize
better, a wider
to select training
the optimal dataset may still
hyperparameters forbethe
needed. Previous studies
ANN topography. have suggested
The dataset was divideda
training dataset
in the ratio of 80%at least
and 20%ten times the network
for training parameters.
and testing. PerformanceIn a lot of original
evaluation research
of the devel-
on stabilisation, there is usually limited experimental data. A cross-validation approach
may be combined for a relatively lower sample size. The statistical evaluation of the
performance of the model in training and testing is presented at the end of Section 4.

4.2. California Bearing Ratio


The California bearing ratio (CBR) of soils is of interest in pavement design. Several
geotechnical investigations are concerned with measuring the CBR of stabilised soils. The
aim of some soil stabilisation is for improvements in CBR to meet the desired outcome.
However, the determination of CBR is expensive. It is desirable to be able to correlate
and predict CBR from known soil characteristics. A study by [59] involved modelling the
CBR of stabilised expansive soils. The expansive clays were treated with lime and quarry
dust as a binder. The quarry dust is a by-product of the crushing of rock for aggregates.
The characteristics of the expansive clay, such as Atterberg properties, compaction and
CBR, were determined following the Indian standards. The stabilisation of the soil was
Geotechnics 2021, 1 160

carried out by the addition of lime and quarry dust of different percentages to the soil
and tested of CBR at 0, 7 and 28 days. A total of 49 data samples were generated for
correlation from the combination of four input parameters, namely LL, PI, OMC and MDD.
To develop the required regression models, three ANN models were utilised based on
different training algorithms using Differential evolution (DE), Bayesian regularization
(BR) and Levenberg–Marquardt (LM) algorithm. The training data was taken as 70% of the
49 data samples, while 30% was utilised for testing. Again, the limitations in the availability
of extensive experimental data are seen and may be due to cost implications. However, the
Levenberg–Marquardt-trained ANN performed better based on the limited data sample.
Additionally, the models showed signs of overfitting, which may be due to insufficient
training data. Detailed summary of model performances for the 28-day CBR model is
shown at the end of Section 4.
The experimental study by [60] employed ANN in modelling soaked CBR of expansive
stabilised clays. The clay was stabilised with varying proportions of coal ash, bagasse ash
and groundnut shell ash (GSA) and geogrid layers. The soil classification was conducted in
line with Indian Standards. The input variables used were the type of ash, mix proportion,
LL, PL, MDD, OMC and the number of geogrid layers. The soil was stabilised with various
combinations of the different features to generate 210 data samples, which were utilised
in training and testing the ANN model. The optimum model architecture was found to
be made of one hidden layer with seven neurons. The model was trained with different
algorithms in order to select the best algorithm based on R2 and MSE. The performance of
the models based on the training algorithm is given in Table 4.

Table 4. ANN performance using various training algorithms [60].

Training Algorithm R MSE


Conscience bias learning function 0.7693 7.08
Gradient descent weight and bias learning function 0.8991 3.98
Gradient descent with momentum weight 0.9163 6.66
Levenberg–Marquardt function 0.94317 0.49
Hebb weight learning rule 0.8761 2.4

The Levenberg–Marquardt training algorithm was selected as the best performing


based on R and MSE values. The model performed well with good predictive ability. The
results showed that the ANN model was capable of modelling the soaked CBR of the
stabilised soils. The performance of black cotton soil stabilised with fly ash and geotextile
was studied by [61] with a view to develop a regression model for predicting the CBR
of stabilised black cotton soil. The soil used in the study was characterised according to
Indian standards. The LL and PL of the soil were reported as 47% and 17%, respectively.
Additionally, the compaction characteristics were determined using standard proctor
compaction tests with MDD and OMC, resulting in 16.26 kN/m3 and 15%, respectively.
The ANN models were developed using LL, PL, fly-ash content, OMC, MDD and the
number of geotextile layers as input variables, while soaked CBR was the target variable in
the output neuron. Different training algorithms were employed while varying the number
of neurons in the hidden layer to select the best model architecture. The performance of
the models is presented here, as shown in Table 5.
Geotechnics 2021, 1 161

Table 5. ANN performance using various training algorithms [61].

Training Algorithm R MSE


Quasi-Newton back propagation 0.88712 1.083 × 10−4
Bayesian regularisation back propagation 0.85190 4.983 × 10−5
Conjugate gradient back propagation with Powell–Beale restarts 0.94122 3.776 × 10−7
Conjugate gradient back propagation with Fletcher–Reeves updates 0.81167 7.339 × 10−6
Conjugate gradient back propagation with Polak–Ribiére updates 0.85819 2.964 × 10−9
Gradient descent back propagation 0.94862 9.985 × 10−9
Levenberg–Marquardt back propagation 0.98695 8.0242 × 10−11
One-step secant back propagation 0.92335 1.388 × 1010
Scaled conjugate gradient back propagation 0.96904 1.946 × 10−6

The Levenberg–Marquardt function was adopted based on minimum error and the
highest R2 . The models performed well with minimum error in relation to a higher
coefficient of determination. A regression analysis of the CBR of soils stabilised with
lime and rice husk ash (RHA) was carried out by [62]. The LL and PL of the soil were
reported as 34% and 20%, with clay and silt content of 20% and 71%. Additionally, the
MDD and OMC of the soil were reported as 17.7 kN/m3 and 14.7%, respectively. The CBR
test was conducted in line with ASTM D1883 -07. Soaked CBR tests were conducted on 0,
7 and 28 days cured samples. The predictive variables considered in the study were the
percentage of RHA, percentage of lime, curing duration, OMC and MDD. The topography
of the ANN model was determined by trial and error by varying the number of neurons
in the hidden layer. Optimum model topography was found to be one hidden layer and
twelve neurons from the R value and MAE. The model performance as per the training
data showed higher R and low MAE. The number of data utilised for training is deemed
small. The performance of the model in generalizing new data may be affected due to
overfitting. This again is a common factor confronting the application of ANN in soil
stabilisation studies.
Furthermore, [63] investigated a predictive model for the correlation of soaked and
unsoaked CBR of stabilised black clay. The black clay is an expansive clay found in semi-
arid regions of tropical and temperate climates. The stabilisation of the clay was done with
cement kiln dust as a binder. The characteristics of the soil and cement kiln dust were
determined according to the British Standards [53,54]. Eight input parameters, namely PL,
LL, specific gravity (GS ), linear shrinkage (LS ), coefficient of uniformity (CU ), coefficient of
gradation (CC ), OMC and MDD, were utilised in developing 72 data samples for two ANN
models; soaked and unsoaked CBR. The model architecture was determined by varying
the number of neurons in the hidden layer from 1 to 20 while evaluating the MSE and R
values. Figure 10 shows the performance of various model architectures.
The optimum was determined using the lowest MSE value. As seen in Figure 10,
for soaked CBR, a model architecture with 8 neurons in the hidden layer was chosen,
while 17 neurons in the hidden was optimum for unsoaked CBR. The models performed
effectively in correlating and predicting the CBR of the stabilised soils with high R and
low MSE.
cement kiln dust as a binder. The characteristics of the soil and cement kiln dust were
determined according to the British Standards [53,54]. Eight input parameters, namely PL,
LL, specific gravity (GS), linear shrinkage (LS), coefficient of uniformity (CU), coefficient of
gradation (CC), OMC and MDD, were utilised in developing 72 data samples for two ANN
Geotechnics 2021, 1 models; soaked and unsoaked CBR. The model architecture was determined by varying 162

the number of neurons in the hidden layer from 1 to 20 while evaluating the MSE and R
values. Figure 10 shows the performance of various model architectures.

0.008
Soaked CBR Unsoaked CBR
0.007
0.006
0.005

MSE 0.004
0.003
0.002
0.001
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Number of neurons in hidden layer

Figure 10.10.
Figure Mean square
Mean error
square versus
error number
versus ofof
number neurons inin
neurons the
thehidden
hiddenlayer
layer[63].
[63].

4.3.The
Permeability
optimumand wasResilient Modulus
determined using the lowest MSE value. As seen in Figure 10, for
soakedTheCBR,permeability of soil is an
a model architecture important
with 8 neurons property of soil layer
in the hidden that describes
was chosen,the while
ease with
17
which in
neurons water flows through
the hidden was optimuma soilformass. In certain
unsoaked geotechnical
CBR. The applications
models performed such as
effectively
inearth embankment
correlating and dam
and predicting thedesigns,
CBR of theit isstabilised
desirablesoils
to be able
with to model
high andMSE.
R and low predict the
permeability of the soil. The study by [64] investigated the effects of lime and pozzolan
onPermeability
4.3. the permeability of dispersive
and Resilient Modulus clays. Dispersive clays are highly susceptible to erosion
dueThe
to their high sodium ion content,
permeability of soil is an important which often leads
property to structural
of soil disintegration
that describes the ease withinto
finer particles easily washed away during seepage. The LL and PL
which water flows through a soil mass. In certain geotechnical applications such as earth of the unstabilised
clay were determined as 34% and 3 and
embankment and dam designs, it is 17%, while
desirable tothe
be MDD
able toand OMC
model andwere 17.85the
predict kN/m
permea-
16.5%,
bility respectively.
of the The clays
soil. The study by [64]soil was stabilised
investigated with various
the effects combinations
of lime and pozzolan on of the
limeper-
and
pozzolan. Permeability tests were conducted to develop a dataset of 69
meability of dispersive clays. Dispersive clays are highly susceptible to erosion due to theirsamples for the
regression analysis. A total of six independent variables were used in the development
high sodium ion content, which often leads to structural disintegration into finer particles
of the ANN model, including percentage passing 0.005 mm size sieve, PI, MDD, lime
easily washed away during seepage. The LL and PL of the unstabilised clay were deter-
percentage, pozzolan percentage and curing time (Cd ). To develop 3the predictive model,
mined as 34% and 17%, while the MDD and OMC were 17.85 kN/m and 16.5%, respec-
ANNs with varying numbers of neurons in the hidden layer were tried. An optimum
tively. The clays soil was stabilised with various combinations of lime and pozzolan. Per-
model with nine neurons in the hidden layer was selected based on low MSE and high R
meability tests were conducted to develop a dataset of 69 samples for the regression analy-
value as with other studies. In line with other studies, 70% of the data was used for training,
sis. A total of six independent variables were used in the development of the ANN model,
15% for testing, and 15% for validation. The performance of the model was satisfactory in
including percentage passing 0.005 mm size sieve, PI, MDD, lime percentage, pozzolan per-
terms of low MSE and high R values.
centage and curing time (Cd). To develop the predictive model, ANNs with varying num-
The resilient modulus of soil is also a useful parameter in pavement design. Few
bers of neurons in the hidden layer were tried. An optimum model with nine neurons in the
studies are available on the resilient modulus of soils. Hence, predictive models for resilient
hidden layer was selected based on low MSE and high R value as with other studies. In line
modulus of stabilised clays are desirable. ANN has also been utilised in correlating and
with other studies,
modelling 70%
resilient of the data
modulus (Mr)wasof used for training,
stabilised 15% The
clays [65]. for testing,
LL, PL,and 15% for
particle vali-
size and
dation. The performance
compaction of the
characteristics model
of four was satisfactory
unstabilised in terms
soil samples of low
were MSE andaccording
determined high R
values.
to ASTM D4318, ASTM D4318, ASTM D422 and ASTM D698. The resilient moduli were
determined by subjecting treated samples to cyclic load test according to AASHTO T307
standard method (AASHTO 2003). A total of 125 data samples were generated for ANN
modelling. Eight inputs parameters, namely cement content, lime content, PI, silt content,
PFA, OMC, MC and clay content, were considered. The trial and error method was used to
determine the optimum model as one hidden layer and nine neurons. The performance
evaluation was based on R values. The model could predict the resilient modulus with a
high correlation. The predicted and experimental values were very close, which was in-
dicative of the accuracy of the developed predictive model. Figure 11 shows the correlation
between predicted and experimental data.
ard method (AASHTO 2003). A total of 125 data samples were generated for ANN model-
ling. Eight inputs parameters, namely cement content, lime content, PI, silt content, PFA,
OMC, MC and clay content, were considered. The trial and error method was used to de-
termine the optimum model as one hidden layer and nine neurons. The performance eval-
uation was based on R values. The model could predict the resilient modulus with a high
Geotechnics 2021, 1 163
correlation. The predicted and experimental values were very close, which was indicative
of the accuracy of the developed predictive model. Figure 11 shows the correlation between
predicted and experimental data.

Figure11.
Figure 11.The
Theexperimental
experimental and
and predicted
predicted resilient
resilient modulus
modulus [65].
[65].

4.4.
4.4.Plasticity
PlasticityIndex
Indexand
and Compaction
Compaction Characteristics
ANN
ANNwas wasalso
alsoutilised
utilisedbyby[66][66]ininregression
regression analysis
analysis toto
develop
develop predictive
predictive models
modelsfor
MDD and OMC
for MDD and OMC usingusing
eighteight
inputinput variables LL, PI,
variables LL,LS,
PI, clay-silt ratio,ratio,
LS, clay-silt sandsandcontent, lime
content,
content, cement
lime content, content
cement and asphalt
content and asphalt content in percentage.
content in percentage. A total of 192
A total of datasets were
192 datasets
obtained from reported
were obtained laboratory
from reported experiments
laboratory and utilised
experiments in training
and utilised based on
in training gradient
based on
descent
gradient momentum algorithmalgorithm
descent momentum for the development of the regression
for the development model formodel
of the regression MDD for and
OMC
MDDpredictions. The optimum
and OMC predictions. Themodeloptimum wasmodel
obtained
wasby trial and
obtained byerror
trial using 52%using
and error of the
data
52%for training,
of the data for24% for testing
training, 24% for andtesting
24% for andvalidation. Performance
24% for validation. evaluation
Performance of the
evalua-
tion ofshows
models the models
a goodshows a good
correlation as correlation
with similarasstudies.
with similar studies.[67]
Furthermore, Furthermore,
investigated [67]
the
investigated
MDD and OMC theofMDD
cement andstabilised
OMC of black cement stabilised
cotton clay. Theblackclaycotton clay. The
was treated clay
with was
various
treated withofvarious
percentages cementpercentages
kiln dust (CKD). of cementThe kiln dust
input (CKD). The
parameters input
used were parameters used
LL, PL, specific
were LL,
gravity (GSPL, specific
), linear gravity (G
shrinkage (LSS),), linear shrinkage
free swell, grain(L S), free
sizes (D10,swell,
D30grain sizes (D10,
and D60), D30
coefficient
ofand D60), coefficient
uniformity (CU ), andof uniformity
coefficient of (CUgradation
), and coefficient
(CC ). Twoof gradation
separate (C C). Two
ANN separate
models were
ANN models
developed were developed
for MDD and OMCfor usingMDD fiveand
andOMC
sevenusing five and
neurons, seven neurons,
respectively, in therespec-
hidden
tively, The
layers. in the hidden
models layers.
were The models
reported to have were reportedeffectively
performed to have performed effectively
and predicted and
the MDD
predicted
and OMC with the MDD and OMC with
high correlation and low higherrors.
correlation
Table 6and low detailed
shows errors. Table
results 6 shows de-
of different
tailedmodels
ANN results and
of different ANN models and their performances.
their performances.
Geotechnics 2021, 1 164

Table 6. Details of ANN Models and Performance.

ANN Data Network Architecture Model Performance References


No. of No. of No. of No. of No. of Training
Soil Parameter No. of Data
ANN Input Variables Training Data Testing Data Validation Hidden Neurons in Algorithm Training Validation Testing
Investigated Used
Used Used Data Layers Hidden Layer Utilised
UCS 8 Inputs. LL, PI, GGBS 283 70% 15% 15% 1 9 Bayesian R = 0.996, R = 0.982, [50]
content, PFA content, regularization MSE = 0.34, MSE = 1.50,
Molarity, activator/ MAPE (%) = 3.65 MAPE (%) = 8.34
binder ratio, Na/Al
ratio, Si/Al ratio
UCS (7 days) Gs, LS , CU , CC , LL, PL, 72 70% 15% 15% 1 11 - R = 0.9782 R = 0.996 R = 0.9711 [52]
OMC, MDD
UCS (14 days) Gs, LS , CU , CC , LL, PL, 72 70% 15% 15% 1 11 - R = 0.9824 R = 0.9843 R = 0.9615 [52]
OMC, MDD
UCS (28 days) Gs, LS , CU , CC , LL, PL, 72 70% 15% 15% 1 11 - R = 0.9946 R = 0.9992 R = 0.9929 [52]
OMC, MDD
UCS Percentage of Clay, 129 70% 15% 15% 1 10 Levenberg– R2 = 0.9813, R2 = 0.9714, [55]
percentage of RHA, Marquardt MSE = 0.0395, MSE = 51.34,
percentage of cement, RMSE = 0.1987 RMSE = 7.1651
percentage of PFA,
curing duration(days)
UCS Curing duration, 80 70% 15% 15% 2 8 in each layer R2 = 0.924 R2 = 0.908 R2 = 0.889 [56]
compaction energy.
UCS ST , MC, We , CM , DI , L, 216 80% 20% 1 50 Adam R = 0.980 R = 0.925 [57]
CA , SV , D, MS , DS , CD , (Average), (Average),
curing time (days), CT MAE = 115.29, MAE = 292.2,
RMSE = 231.2 RMSE = 419.82
UCS Moisture content, 51 70% 30% 2 12 and 10 Levenberg– R = 0.95, R = 0.94, [58]
cement (percentage), air neurons Marquardt MAE = 3.9001, MAE = 8.6535,
foam(percentage) and RMSE = 9.1948 RMSE = 10.3390
waste fishing net
content (percentage)
CBR (28 days) LL, PL, OMC and MDD 49 70% 30% - - - Differential R = 0.98 R= 0.86 [59]
evolution
CBR (28 days) LL, PL, OMC and MDD 49 70% 30% - - - Levenberg– R = 0.96 R= 0.93 [59]
Marquardt
CBR (28 days) LL, PL, OMC and MDD 49 70% 30% - - - Bayesian R = 0.96 R= 0.93 [59]
regularization
CBR Type of ash, mix 210 - - - 1 7 Levenberg– R = 0.94472 R = 0.93327 R = 0.94685 [60]
proportion (percentage), Marquardt
LL, PL, MDD, OMC and
number of geogrid
layers
Geotechnics 2021, 1 165

Table 6. Cont.

ANN Data Network Architecture Model Performance References


No. of No. of No. of No. of No. of Training
Soil Parameter No. of Data
ANN Input Variables Training Data Testing Data Validation Hidden Neurons in Algorithm Training Validation Testing
Investigated Used
Used Used Data Layers Hidden Layer Utilised
CBR (28 days LL, PI, percentage of - - - - 1 - Levenberg– R = 0.99846 R = 0.98508 R = 0.92149 [61]
Soaked) PFA, OMC, MDD and Marquardt
no. of geotextile layers
CBR (28 days Percentage of RHA, 48 70% 15% 15% 1 12 - R = 0.9948 R = 0.98909 R = 0.98895 [62]
Soaked) percentage of lime,
curing time (days),
OMC and MDD
CBR (Soaked) PL, LL, GS , LS , CU , CC , 72 70% 15% 15% 1 8 - R = 0.9988 R = 0.9996 R = 0.9976 [63]
OMC and MDD
CBR PL, LL, GS , LS , CU , CC , 72 70% 15% 15% 1 17 - R = 0.9912 R = 0.9993 R = 0.9806 [63]
(Unsoaked) OMC and MDD
Coefficient of percentage passing 69 70% 15% 15% 1 9 R = 0.9968 R = 0.98883 R = 0.99405 [64]
permeability 0.005 mm, PI, MDD,
(K) lime percentage,
pozzolan percentage, Cd
Resilient Percentage of cement, 125 - - - 1 9 - R = 0.9517 - R = 0.9467 [65]
Modulus (Mr) percentage of lime, PI,
percentage of silt,
percentage of PFA,
OMC, MC and clay
MDD LL, PI, LS, clay-silt ratio, 192 52% 24% 24% 1 18 Gradient R2 = 0.9183, R2 = 0.9101, [66]
sand content, lime descent MSE = 0.28%, MSE = 0.26%,
content, cement content, momentum MAE = 4.44% MAE = 4.24%
asphalt content in
percentage
OMC LL, PI, LS, clay-silt ratio, 192 52% 24% 24% 1 15 Gradient R2 = 0.9025, R2 = 0.8916, [66]
sand content, lime descent MSE = 88.21%, MSE = 89.57%,
content, cement content, momentum MAE = 118.37% MAE = 113.03%
asphalt content in
percentage
MDD GS , LS , free swell, D10 , 90 70% 15% 15% 1 7 R = 0.9946 R = 0.9715 R = 0.9754 [67]
D30 , D60, CU , CC , LL, PL
OMC GS , LS , free swell, D10 , 90 70% 15% 15% 1 5 - R = 0.9977 R = 0.9779 R = 0.8855 [67]
D30 , D60, CU , CC , LL, PL

- Not reported.
Geotechnics 2021, 1 166

5. Discussions
Problematic clays are commonly encountered in geotechnical engineering practice
and will demand extensive ground improvement programs under practical conditions.
Undoubtedly, chemical stabilisation involving the use of calcium and non-calcium-based
binders have been successfully used to stabilise and improve the engineering properties
of clays, making them fit for a specified purpose. The stabilisation process will often
require series of detailed laboratory procedures, including sample preparation, curing and
testing. This is followed by the analysis of the experimental data, observation of trends,
the establishment of the relationship between variables and development of predictive
models [68–74]. In the case of a large collection of data involving multiple variables, it is
very tedious to learn the functional relations and inter-variable dependencies without the
use of a more sophisticated tool or technique, such as the concept of artificial intelligence.
Reference [68] has observed that ANN’s application in soil stabilisation is a relatively
new development. However, the present study shows that it is fast becoming a more
reliable tool in the development of predictive models for the prediction of geotechnical
properties of stabilised clays. The ability of ANN to learn the functional relations and
inter-variable dependencies within large collections of independent variables gives it a
major advantage over traditional regression analysis involving cumbersome mathematical
procedures. Some studies have utilised a relatively large database of soil properties, as
seen in Table 3, and will require a longer analysis time using regression analysis. These
traditional procedures become even more cumbersome if a change is made in any variable
in the dataset as the analysis must be repeated. However, in using ANN, such changes
can easily be accommodated without any repetition, and this gives ANN the advantage
of fitting a closer curve on the experimental data for the parameter of interest as a larger
dataset encompasses a wider variety of soil responses. Another advantage of ANN is the
possibility of modelling multiple dependent or target variables simultaneously. ANN can
search the relationship between a given dataset and two or more dependent variables while
assigning required coefficients for the various targets. As seen in Table 3, some studies have
simultaneously modelled more than one dependent variable and obtained good predictive
models for the dependent variables. This, again, is a time-saving advantage.
From Table 3, it can be seen that the performance of the model is not directly pro-
portional to the number of neurons in the hidden layer. However, a sufficient number
of neurons based on model training must be allowed for optimum performance and is
majorly dependent on several parameters depending on the complexity of the problem
investigated. ANN has been successfully utilized in performance evaluation and post-
stabilisation behaviour of treated clays with a high degree of reliability. From some of the
studies presented in Table 3, statistical evaluation of ANN’s performance show superior
modelling capabilities than the usual regression models. The high R2 and low MSE, MAE
and RMSE of ANN regression analysis are pointers to its advantage. From the study, it
was discovered that most of the ANN models utilized were backpropagation feed-forward
networks with one hidden layer. Only two studies [56,58] used two hidden layers for
the analysis. Other studies have also pointed out that most regression analyses in soil
stabilisation problems are easily solved using a simple architecture of one hidden layer.
In addition, the quality and size of the database used in training contributes to the
robustness of the model and enhances its performance in terms of generalisation. As seen
in Table 7, many of the studies have used a relatively small dataset for the analysis, and
this raises concerns if the models would be reliable under practical conditions.
Geotechnics 2021, 1 167

Table 7. The number of datasets utilised for model development.

Network Architecture ANN Data


No of No of Neurons No. of No. of References
ANN Input No. of Data
Hidden in Hidden Network Training Remarks
Variables Used
Layers Layer Parameters Data Used
1 9 8 91 283 198 Sufficient [50]
1 11 8 111 72 50 May overfit [52]
1 11 8 111 72 50 May overfit [52]
1 11 8 111 72 50 May overfit [52]
1 10 5 71 129 90 Sufficient [55]
2 8 in each layer 2 105 80 56 May overfit [56]
1 50 14 801 216 172 May overfit [57]
2 12 and 10 4 201 51 36 May overfit [58]
- - 4 - 49 34 No Remark [59]
- - 4 - 49 34 No Remark [59]
- - 4 - 49 34 No Remark [59]
1 7 7 - 210 - No Remark [60]
1 - 6 - - - No Remark [61]
1 12 5 85 48 34 May overfit [62]
1 8 8 81 72 50 May overfit [63]
1 17 8 171 72 50 May overfit [63]
1 9 6 73 69 48 May overfit [64]
1 9 8 - 125 - No Remark [65]
1 18 8 181 192 100 May overfit [66]
1 15 8 151 192 100 May overfit [66]
1 7 10 85 90 63 May overfit [67]
1 5 10 61 90 63 Sufficient [67]
- Not reported.

One major factor may be due to the high cost of conducting geotechnical experimenta-
tion. This factor may be more significant in other areas of geotechnical engineering other
than soil stabilisation. However, as shown in Table 3, satisfactory performance has also
been achieved with a relatively small dataset within the available training dataset.
The capabilities of ANN in soil stabilisation are wide-ranged, including the determi-
nation of the optimum mix design for a given additive, the optimisation of soil parameters
by adjusting dependent soil properties, predicting the mineralogical composition of sta-
bilised soils, classifying soils based on established criteria and modelling and predicting
the performance of stabilised soils. In summary, it is obvious that ANN can be utilised in
regression analysis for the development of useful models for modelling the geo-mechanical
properties of stabilised clays and indeed in soil stabilisation as a whole.

6. Conclusions
• The advantages of the artificial neural over traditional regression analysis as applied
to stabilisation have been highlighted in the foregoing sections. In a typical field sta-
bilisation project, in order to improve the properties of expansive clays, experimental
data are usually generated from several field and laboratory tests to monitor and
ascertain the progress made in terms of improvement. These procedures are expensive
Geotechnics 2021, 1 168

and time-consuming and may be reduced to a minimum using ANN to predict the
field response of the soils. In summary, the following conclusions are made.
• An artificial neural network is reliable and can be employed in modelling various
properties of stabilised clays for easy prediction of soil response while eliminating the
need for extensive experimental procedures.
• Backpropagation feedforward networks are the most used models in dealing with the
problem of regression analysis for stabilisation of clays.
• An artificial neural network should be developed with a relatively substantial dataset
to regression models with good correlation. Many of the studies in regression analysis
of stabilised clays have used relatively small data sets, although the models have
performed well. The ability of the models to generalize can be improved with a larger
dataset which fields a wide range of possible soil behaviour for proper training of
the model.
• Shallow networks made up of one hidden layer are the most used ANN architecture
in developing predictive models for the prediction of geotechnical characteristics
of stabilised clays and in modelling the response of stabilised expansive clays. The
Levenberg–Marquardt training algorithm has been reported to be the most used
among the studies reviewed.

Author Contributions: Conceptualization, S.J.A. and J.J.J.; methodology, J.J.J. and S.J.A.; software,
A.K.; validation, J.J.J. and S.J.A.; formal analysis, J.J.J. and S.J.A.; investigation, J.J.J. and S.J.A.;
resources, S.J.A., A.K. and C.A.B.; data curation, J.J.J. and S.J.A.; writing—original draft preparation,
J.J.J. and S.J.A.; writing—review and editing, J.J.J., S.J.A. and C.A.B.; visualization, S.J.A., C.A.B. and
A.K.; supervision, S.J.A., C.A.B. and A.K.; project administration, C.A.B., A.K. and S.J.A. All authors
have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Some or all data that support the findings of this study are available
from the corresponding author.
Acknowledgments: The lead author is thankful to the Ph.D. supervisory team for their enormous
support during this study.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Gillott, J.E. Some clay-related problems in engineering geology in North America. Clay Miner. 1986, 21, 261–278. [CrossRef]
2. Das, A.; Roy, S. Effect of Expansive Soil on Foundation and Its Remedies. Int. J. Innov. Res. Sci. Eng. Technol. 2014, 3, 92–95.
3. Balasingam, M.; Farid, S. Interpretation of Geotechnical Properties of Cement Treated Soil; Washington State Department of Transporta-
tion (WSDOF): Olympia, WA, USA, 2008; pp. 77–122.
4. Chew, S.H.; Kamruzzaman, A.H.M.; Lee, F.H. Physicochemical and Engineering Behavior of Cement Treated Clays. J. Geotech.
Geoenvironmental Eng. 2004, 130, 696–706. [CrossRef]
5. Abbey, S.J.; Eyo, E.U.; Oti, J.; Amakye, S.Y.; Ngambi, S. Mechanical Properties and Microstructure of Fibre-Reinforced Clay
Blended with By-Product Cementitious Materials. Geosciences 2020, 10, 241. [CrossRef]
6. Abbey, S.J.; Ngambi, S.; Olubanwo, A.O.; Tetteh, F.K. Strength and Hydraulic Conductivity of Cement and By–Product Cementi-
tious Materials Improved Soil. Int. J. Appl. Eng. Res. 2018, 13, 8684–8694, ISSN 0973-4562.
7. Eyo, E.U.; Ng’Ambi, S.; Abbey, S.J. Incorporation of a nanotechnology-based additive in cementitious products for clay stabilisa-
tion. J. Rock Mech. Geotech. Eng. 2020, 12, 1056–1069. [CrossRef]
8. Eyo, E.U.; Ng’Ambi, S.; Abbey, S.J. Performance of clay stabilized by cementitious materials and inclusion of zeolite/alkaline
metals-based additive. Transp. Geotech. 2020, 23, 100330. [CrossRef]
9. Dongare, A.D.; Kharde, R.R.; Kachare, A.D. Introduction to Artificial Neural Network. Int. J. Eng. Innov. Technol. 2012, 2, 189–193,
ISSN 2277-3754.
10. Sharma, V.; Rai, S.; Anurag Dev, A. A Comprehensive Study of Artificial Neural Networks. Int. J. Adv. Res. Comput. Sci. Softw.
Eng. 2012, 2, 278–284, ISSN 2277 128X.
Geotechnics 2021, 1 169

11. Vastrad, C. Performance Analysis of Neural Network Models for Oxazolines and Oxazoles Derivatives Descriptor Dataset. Int. J.
Inf. Sci. Tech. 2013, 3, 1–15. [CrossRef]
12. Stepniewska-Dziubinska, M.M.; Zielenkiewicz, P.; Siedlecki, P. Development and evaluation of a deep learning model for
protein–ligand binding affinity prediction. Bioinformatics 2018, 34, 3666–3674. [CrossRef]
13. Lal, B.; Tripathy, S.S. Prediction of dust concentration in open cast coal mine using artificial neural network. Atmos. Pollut. Res.
2012, 3, 211–218. [CrossRef]
14. Chao, Z.; Ma, G.; Zhang, Y.; Zhu, Y.; Hu, H. The application of artificial neural network in geotechnical engineering. International
Conference on Civil and Hydraulic Engineering (IConCHE). IOP Conf. Ser. Earth Environ. Sci. 2018, 189, 22054. [CrossRef]
15. Nihat, S.I. Estimation of swell index of fine-grained soils using regression equations and artificial neural net-works. Sci. Res.
Essay 2009, 4, 1047–1056.
16. Rani, C.S.; Vaddi, P.K.; Togati, N.V.V.K. Artificial Neural Networks (ANN) for Prediction of Engineering Properties of Soils. Int. J.
Innov. Technol. Explor. Eng. 2013, 3, 123–130, ISSN 2278-3075.
17. Kalantary, F. and Kordnaeij, A. Prediction of compression index using artificial neural network. Sci. Res. Essays 2012, 7, 2835–2848.
[CrossRef]
18. Singh, V.; Bano, S.; Yadav, A.K.; Ahmad, S. Feasibility of Artificial Neural Network in Civil Engineering. Int. J. Trend Sci. Res. Dev.
2019, 3, 724–728. [CrossRef]
19. Winston, P.H. Artificial Intelligence, 3rd ed.; Addison-Wesley Pub. Co.: Boston, MA, USA, 1992; ISBN 0-201-53377-4.
20. Zell, A. Simulation of Neural Networks, 1st ed.; Addison-Wesley Pub. Co.: Boston, MA, USA, 2003; ISBN 978-3-89319-554-1. OCLC
249017987.
21. Cal, Y. Soil classification by neural network. Adv. Eng. Softw. 1995, 22, 95–97. [CrossRef]
22. Shahin, M.A.; Jaksa, M.B.; Maier, H.R. Artificial Neural Networks-Based Settlement Prediction Formula for Shallow Foundations
on Granular Soils. Aust. Geomech. 2002, 37, 45–52.
23. Gunaydin, O.; Gokoglu, A.; Fener, M. Prediction of artificial soil’s unconfined compression strength test using statistical analyses
and artificial neural networks. Adv. Eng. Softw. 2010, 41, 1115–1123. [CrossRef]
24. Kerh, T.; Hu, Y.G.; Wu, C.H. Estimation of consolidation settlement caused by groundwater drawdown using artificial neural
networks. Adv. Eng. Softw. 2003, 34, 559–568. [CrossRef]
25. Reuter, U.; Sultan, A.; Reischl, D.S. A comparative study of machine learning approaches for modeling concrete failure surfaces.
Adv. Eng. Softw. 2018, 116, 67–79. [CrossRef]
26. Snyman, J. Practical Mathematical Optimization: An Introduction to Basic Optimisation Theory and Classical and New Gradient-Based
Algorithms; Springer Science & Business Media: New York, NY, USA, 2005; ISBN 978-0-387-24348-1.
27. Hinkelmann, K. Neural Networks. University of Applied Sciences Northwestern Switzerland School of Business. Available
online: http://didattica.cs.unicam.it/lib/exe/fetch.php?media=didattica:magistrale:kebi:ay_1718:ke-11_neural_networks.pdf
(accessed on 26 April 2021).
28. Weisstein, E.W. Hyperbolic Tangent. MathWorld—A Wolfram Web Resource. Available online: https://mathworld.wolfram.
com/HyperbolicTangent.html (accessed on 26 April 2020).
29. Squires, G.L. Practical Physics; Cambridge University Press (CUP): New York, NY, USA, 2001; ISBN 978-0-521-77940-1. [CrossRef]
30. Mote, H.; Kumar, S.R.S. Use of Artificial Neural Network for Initial design of steel structures. IOP Conf. Ser. Mater. Sci. Eng. 2019,
660, 12064. [CrossRef]
31. Sildir, H.; Aydin, E.; Kavzoglu, T. Design of Feedforward Neural Networks in the Classification of Hyperspectral Imagery Using
Superstructural Optimization. Remote Sens. 2020, 12, 956. [CrossRef]
32. Naidu, K.; Ali, M.S.; Abu Bakar, A.H.; Tan, C.K.; Arof, H.; Mokhlis, H. Optimized artificial neural network to improve the
accuracy of estimated fault impedances and distances for underground distribution system. PLoS ONE 2020, 15, e0227494.
[CrossRef] [PubMed]
33. Renotte, C.; Wouwer, A.V.; Bogaertst, P.; Remy, M. Neural Network Applications in Non-linear Modelling of (Bio) chemical
Processes. Meas. Control. 2001, 34, 197–201. [CrossRef]
34. Zhang, T.; Fell, F.; Liu, Z.; Preusker, R.; Fischer, J.; He, M. Evaluating the performance of artificial neural network techniques for
pigment retrieval from ocean color in Case I waters. J. Geophys. Res. Space Phys. 2003, 108, 3286. [CrossRef]
35. Eyo, E.U.; Abbey, S.J. Machine learning regression and classification algorithms utilized for strength prediction of OPC/by-
product materials improved soils. Constr. Build. Mater. 2021, 284. [CrossRef]
36. Schmidhuber, J. Deep learning in neural networks: An overview. Neural Netw. 2015, 61, 85–117. [CrossRef]
37. Wald, A. Statistical Decision Functions; John Wiley & Sons: New York, NY, USA, 1950.
38. Ikizler, S.B.; Aytekin, M.; Vekli, M.; Kocabas, F. Prediction of swelling pressures of expansive soils using artificial neural networks.
Adv. Eng. Softw. 2009, 41, 647–655. [CrossRef]
39. Ciresan, D.; Meier, U.; Masci, J.; Gambardella, L.M.; Schmidhuber, J. Flexible, High Performance Convolutional Neural Networks
for Image Classification. Twenty-Second Int. Jt. Conf. Artif. Intell. 2011, 2, 1237–1242.
40. Sushama, K.; Bindhu, L. ANN Based Prediction of Shear Strength of Soils from their Index Properties. Int. J. Earth Sci. Eng. 2015,
8, 2195–2202, ISSN 0974-5904.
Geotechnics 2021, 1 170

41. Murata, N.; Yoshizawa, S.; Amari, S. Learning curves, model selection and complexity of neural networks. In Advances in Neural
Information Processing Systems 5; Jose Hanson, S., Cowan, J.D., Lee Giles, C., Eds.; Morgan Kaufmann: San Mateo, CA, USA, 1993;
pp. 607–614.
42. Maind, S.B.; Wankar, P. Research Paper on Basic of Artificial Neural Network. Int. J. Recent Innov. Trends Comput. Commun. 2014,
2, 96–100.
43. Gareth, J. An Introduction to Statistical Learning: With Applications in R; Springer: New York, NY, USA, 2013; p. 176. ISBN
978-1461471370.
44. Pathivada, S. Effects of Water-Cement Ratio on Deep Mixing Treated Expansive Clay Characteristics; The University of Texas at
Arlington: Arlington, TX, USA, 2005.
45. Lorenzo, G.A.; Bergado, D.T. Fundamental Parameters of Cement-Admixed Clay—New Approach. J. Geotech. Geoenvironmental
Eng. 2004, 130, 1042–1050. [CrossRef]
46. Turk, G.; Logar, J.; Majes, B. Modelling soil behaviour in uniaxial strain conditions by neural networks. Adv. Eng. Softw. 2001, 32,
805–812. [CrossRef]
47. Fan, J.; Wang, D.; Qian, D. Soil-cement mixture properties and design considerations for reinforced excavation. J. Rock Mech.
Geotech. Eng. 2018, 10, 791–797. [CrossRef]
48. Shahin, M.A.; Jaksa, M.B.; Maier, H.R. Artificial Neural Networks Application in Geotechnical Engineering. Aust. Geomech. 2001,
36, 49–62.
49. Gardner, M.W.; Dorling, S.R. Artificial neural networks (The multilayer perceptron)—A review of applications in the atmospheric
sciences. Atmos. Environ. 1998, 32, 2627–2636. [CrossRef]
50. Mozumder, R.A.; Laskar, A.I. Prediction of unconfined compressive strength of geopolymer-stabilised clayey soils using artificial
neural network. Comput. Geotech. 2015, 69, 291–300. [CrossRef]
51. BRE. Low-Rise Buildings on Shrinkable Clay Soils: BRE Digest; CRC: London, UK, 1993; Volumes 240–242.
52. Salahudeen, A.B.; Sadeeq, J.A.; Badamasi, A.; Onyelowe, K.C. Prediction of Unconfined Compressive Strength of Treated
Expansive Clay Using Back-Propagation Artificial Neural Networks. Niger. J. Eng. Fac. Eng. Ahmadu Bello Univ. 2020, 27, 45–58,
ISSN 0794–4756.
53. BSI. BS 1377: 1990—Methods of Test. for Soils for Civil. Engineering Purposes; British Standards Institute: Milton Keynes, UK, 1990.
54. BSI. BS 1924: 1990—Stabilised Materials for Civil. Engineering Purposes; British Standards Institute: Milton Keynes, UK, 1990.
55. Priyadarshee, A.; Chandra, S.; Gupta, D.; Kumar, V. Neural Models for Unconfined Compressive Strength of Kaolin Clay Mixed
with Pond Ash, Rice Husk Ash and Cement. J. Soft Comput. Civ. Eng. 2020, 4, 85–120. [CrossRef]
56. Ayeldeen, M.; Yuki, H.; Masaki, K.; Abdelazim, N. Unconfined Compressive Strength of Compacted Disturbed Cement-Stabilised
Soft Clay. Int. J. Geosynth. Ground Eng. 2016, 2, 1–10. [CrossRef]
57. Ngo, H.T.T.; Pham, T.A.; Vu, H.L.T.; Giap, L.V. Application of Artificial Intelligence to Determined Unconfined Compressive
Strength of Cement-Stabilized Soil in Vietnam. Appl. Sci. 2021, 11, 1949. [CrossRef]
58. Tran, V.Q. Compressive Strength Prediction of Stabilized Dredged Sediments Using Artificial Neural Network. Adv. Civ. Eng.
2021, 2021. [CrossRef]
59. Sabat, A.K. Prediction of California Bearing Ratio of a Stabilised Expansive Soil using Artificial Neural Network and Support
Vector Machine. Electron. J. Geotech. Eng. 2015, 20, 981–991.
60. Rajakumar, C.; Babu, G.R. Experimental study and neural network modelling of expansive sub grade stabilized with industrial
waste by-products and geogrid. Mater. Today Proc. 2021, 46, 131–137. [CrossRef]
61. Vinodhkumar, S.; Balaji, S. Artificial Neural Network Modelling and Economic Analysis of Black Cotton Soil Subgrade Stabilised
with Flyash and Geotextile. Int. J. Earth Sci. Eng. 2016, 9, 81–86, ISSN 0974-5904.
62. Belal, A.; Rahman, M.A.; Rafizul, I.M. Prediction of California Bearing Ratio of Stabilised Soil Using Artificial Neural Network.
In Proceedings of the 3rd International Conference on Civil Engineering for Sustainable Development (ICCESD 2016), KUET,
Khulna, Bangladesh, 12–14 February 2016; pp. 977–987, ISBN 978-984-34-0265-3.
63. Salahudeen, A.B.; Sadeeq, J.A. California Bearing Ratio Prediction of Modified Black Clay Using Artificial Neural Networks. In
Proceedings of the West Africa Built Environment Research (WABER) Conference 10th Anniversary Conference, Accra, Ghana,
5–7 August 2019; Laryea, S., Essah, E., Eds.; WABER: Johannesburg, South Africa, 2019. ISBN 978–9988–2–6010–1. [CrossRef]
64. Vakili, A.H.; Davoodi, S.; Arab, A.; Razip, M.; Bin Selamat, M. Use of Artificial Neural Network in Predicting Permeability of
Dispersive Clay Treated with Lime and Pozzolan. Int. J. Sci. Res. Environ. Sci. 2015, 3, 23–37. [CrossRef]
65. Hanandeha, S.; Ardahb, A.; Abu-Farsakhc, M. Using artificial neural network and genetics algorithm to estimate the resilient
modulus for stabilised subgrade and propose new empirical Formula. Transp. Geotech. 2020, 24, 100358. [CrossRef]
66. Alavi, A.H.; Gandomi, A.H.; Mollahassani, A.; Heshmati, A.A.; Rashed, A. Modeling of maximum dry density and optimum
moisture content of stabilised soil using artificial neural networks. J. Plant. Nutr. Soil Sci. 2010, 173, 368–379. [CrossRef]
67. Salahudeen, A.B.; Ijimdiya, T.S.; Eberemu, A.O.; Osinubi, K.J. Artificial Neural Networks Prediction of Compaction Characteristics
of Black Cotton Soil Stabilised with Cement Kiln Dust. J. Soft Comput. Civ. Eng. 2018, 2, 50–71.
68. Bahmed, I.T.; Harichane, K.; Ghrici, M.; Boukhatem, B.; Rebouh, R.; Gadouri, H. Prediction of geotechnical properties of clayey
soils stabilised with lime using artificial neural networks (ANNs). Int. J. Geotech. Eng. 2017, 13, 191–203. [CrossRef]
69. Kamruzzaman, A.H.; Chew, S.H.; Lee, F.H. Structuration and destructuration behavior of cement-treated Singapore marine clay
cement-treated Singapore marine clay. J. Geotech. Geoenviron. Eng. 2009, 135, 573–589. [CrossRef]
Geotechnics 2021, 1 171

70. Taskiran, T. Prediction of California bearing ratio (CBR) of fine grained soils by AI methods. Adv. Eng. Softw. 2010, 41, 886–892.
[CrossRef]
71. Radhakrishnan, G.; Kumar, M.A.; Raju, G.V.R.P. Influence of Chloride Compounds on the Swelling and Strength Properties of
Expansive Subgrade Soil. J. Mech. Civ. Eng. IOSR-JMCE 2016, 13, 57–60.
72. Ikeagwuani, C.C.; Nwonu, C.D. Emerging trends in expansive soil stabilisation: A review. J. Rock Mech. Geotech. Eng. 2019, 11,
423–440. [CrossRef]
73. Eyo, E.; Ng’Ambi, S.; Abbey, S. Investigative Study of Behaviour of Treated Expansive Soil Using Empirical Correlations. In
IFCEE 2018; American Society of Civil Engineers (ASCE): Orlando, FL, USA, 2018; pp. 373–384.
74. Eyo, E.U.; Abbey, S.J.; Ngambi, S.; Ganjian, E.; Coakley, E. Incorporation of a nanotechnology-based product in cementitious
binders for sustainable mitigation of sulphate-induced heaving of stabilised soils. Eng. Sci. Technol. Int. J. 2021, 24, 436–448.
[CrossRef]

You might also like