Research Article

Prediction of the depth of local scouring at a bridge pier using a gene

expression programming method
Waqed H. Hassan1 · Halah K. Jalal1

Received: 28 July 2020 / Accepted: 30 December 2020 / Published online: 19 January 2021
© The Author(s) 2021  OPEN

Abstract
Local scouring around the piers of a bridge is the one of the major reasons for bridge failure, potentially resulting in heavy
losses in terms of both the economy and human life. Prediction of accurate depth of local scouring is a difficult task due
to the many factors that contribute to this process, however. The main aim of this study is thus to offer a new formula for
the prediction the local depth of scouring around the pier of a bridge using a modern fine computing modelling tech-
nique known as gene expression programming (GEP), with data obtained from numerical simulations used to compare
GEP performance with that of a standard non-linear regression (NLR) model. The best technique for prediction of the
local scouring depth is then determined based on three statistical parameters: the determination coefficient (R2), mean
absolute error (MAE), and root mean squared error (RMSE). A total data set of 243 measurements, obtained by numerical
simulation in Flow-3D, for intensity of flow, ratio of pier width, ratio of flow depth, pier Froude number, and pier shape
factor is divided into training and validation (testing) datasets to achieve this. The results suggest that the formula from
the GEP model provides better performance for predicting the local depth of scouring as compared with conventional
regression with the NLR model, with ­R2 = 0.901, MAE = 0.111, and RMSE = 0.142. The sensitivity analysis results further
suggest that the ratio of the depth of flow has the greatest impact on the prediction of local scour depth as compared
to the other input parameters. The formula obtained from the GEP model gives the best predictor of depth of scouring,
and, in addition, GEP offers the special feature of providing both explicit and compressed arithmetical terms to allow
calculation of such depth of scouring.

Keywords Local scouring · Scour depth prediction · Gene expression programming · Non-linear regression

1 Introduction researchers have studied the scouring problem under vari-

ous conditions, determining that the main cause of con-
Scouring is a term used to describe the severe localised cern about the stability of bridge foundations is indeed
wear of bed material around the pier of a bridge that scouring around the bridge piers [5]. Deformation of the
occurs when the corrosive strength of the water exceeds river bed is thus a major concern among hydraulic and
the capacity of the bed material [1, 2]. Scouring at the pier infrastructure engineers [6, 7], as the presence of various
of a bridge considered to be the main reason for bridge hydraulic structures such as bridges that block flow tend to
failure and, as such, may lead to huge losses of life and cause contractions of flow and scouring at the abutments
economic impacts [3]. General scouring, contraction scour- and piers [8, 9].
ing and local scouring are among the main types of scour- It is important to anticipate the depth of local scouring
ing patterns identified at bridge sites [4]. Across the world, around bridge piers for both safety and economic reasons,

* Waqed H. Hassan, [email protected]; [email protected] | 1Civil Engineering Department, College of Engineering,

University of Kerbala, Kerbala, Iraq.

SN Applied Sciences (2021) 3:159 | https://doi.org/10.1007/s42452-020-04124-9

Vol.:(0123456789)
Research Article SN Applied Sciences (2021) 3:159 | https://doi.org/10.1007/s42452-020-04124-9

as over-prediction or under-prediction of the depth of its calculations are tedious. Advances in mathematical
scouring may lead to extra expense during the construc- computation have since been achieved by applying arti-
tion of the bridge or to failure in use [10]. There is, how- ficial intelligence (AI) techniques, allowing modelling to
ever, a significant amount disagreement and uncertainty be done easily, accurately and with little effort [23, 24].
relating to the prediction of scour depth in the field, and The AI-based inductive modelling techniques revealed
most bridge failures arise from failures in oversight with in the more recent literature are now widely applied to
regard to the scour problem [11]. create complex response functions, including analysis of
In recent years, thanks to the ongoing development of the scouring mechanisms that occur at bridge piers; these
computer science and numerical modelling of hydraulic offer robust and nonlinear modelling structures and thus
structures, computational fluid dynamics (CFD) has been have the ability to define cause and effect relationships
used widely n the field of engineering to simulate the for such complicated operations. The relevant AI technol-
behaviours of fluid flow and the resulting depth of scour- ogies include artificial neural networks (ANN), adaptive
ing at the piers of bridges [12]. A number of recent studies neural fuzzy inference system (ANFIS), genetic program-
have used Flow-3D as a numerical modelling technique ming (GP), genetic algorithms (GA), and gene expression
to simulate the depth of scouring around bridge piers programming (GEP) [18, 25], as well as alternative synthe-
[13–15], as local scouring around the pier of a bridge may sis processors that incorporate artificial neural networks
be precisely simulated using the correct Flow-3D model. with an adaptive neurotransmitter system [26]. The latter
This Flow-3D model offers a powerful capacity to examine was recently adopted because it gives effective estima-
the manner in which gases and liquids move, allowing the tions of local scouring depth, while ANNs have been used
solution of transient problems, free surface modelling, and frequently due to their reasonable solutions to various
the assessment of sediment transport [16]. The main aim in problems of hydraulic engineering that arise due to the
using a numerical model such as Flow-3D is that, instead extremely complicated non-linear relationships between
of having to design a huge model and expensive appur- input and output parameters for the corresponding data
tenances to allow measurement of specific factors, the [27, 28]. Another key tool is the GEP Soft Computing Tech-
behaviours of fundamental fluid movement, incorporating nologies tool, which has recently replaced several other
distribution velocity, bed shear, turbulent kinetic energy, tools due to its ease of coding, rapid calculations, and
and pressures, can be acquired through the application simple modelling. Several researchers in different fields of
of Flow-3D [17]. engineering have thus illustrated that the use of AI tech-
Various field conditions effect on the depth of scouring niques is more exact and functional than the application
at the piers of a bridge, including pier shape, intensity of of other, older, technologies [29, 30].
flow, flow depth, width of pier, and angle of alignment. A review of existing studies suggested, however, that
Field condition variations make determining the mecha- the application of GEP for the prediction of scour depth
nisms of scouring at the pier of a bridge highly complex, around bridge piers had not been implemented on a large
making it hard to create a universal equation for predict- scale, creating an immediate need for action in this regard.
ing local depth of scouring at a given bridge pier. Several In the current study, therefore, the main objective is to
different methods have nevertheless been developed to develop a new scour depth formula based on the use of
predict the depth of scouring that may occur at a specific parameters such as flow depth, flow intensity, pier Froud
site under the described conditions. These methods were number, pier width, and pier shape by using GEP with data
developed by specifying certain conditions and using from numerical simulation, to allow GEP performance to
techniques to dictate the depth of the scouring that must be compared with that of NLR models.
then take place. Within the previous literature, several pre- GEP and NLR models were thus used for formulaic
dictive equations have thus been proposed to estimate prediction of the local depth of scouring at the pier of a
the local depth of scouring at bridge piers using conven- bridge based on data obtained from numerical simula-
tional regression-based mechanisms with both field and tions in Flow-3D, which were divided into a training and a
experimental data [18]. Recently, [19] indicated that Colo- validation data set. The dimensionless parameters of flow
rado State University work in [20] and [21] provided rea- intensity, pier width ratio, flow depth ratio, pier shape and
sonable estimates, while [22] also predicted the depth of pier Froude number were selected as the effective param-
excessive scouring around a pier successfully, based on a eters for predicting local depth of scour, and these dimen-
comparison of several bridge pier scouring formulas utilis- sionless parameters were utilised as output and input vari-
ing laboratory and field data. Equations to calculate depth ables in both GEP and NLR models. The best technique for
of scouring were developed by these researchers by apply- predicting local scouring depth around a bridge pier was
ing dimensional analysis followed by nonlinear regression assessed using three statistical parameters: RMSE, ­R2, and
analysis; however, this mechanism is low-precision, and MAE. After the best technique was identified, sensitivity

Vol:.(1234567890)
SN Applied Sciences (2021) 3:159 | https://doi.org/10.1007/s42452-020-04124-9 Research Article

analysis was also conducted to identify the most sensi- their dimensions. By using Buckingham π-theorem and
tive parameter for the prediction of scour depth, for the selecting ρ, V, and b as the repeated variables while con-
purposes of focusing future studies. sidering ρ,d50 , K θ , t as constant parameters so that their
effects may be ignored, the final dimensionless function
that describes the influence of the variables on local depth
2 Dimensional analysis of local scouring of scouring around the pier of a bridge was extracted as
around a bridge pier illustrated in Eq. (2).
� �
The first step towards determining the functional rela- ds V B y V
= f2 ,K , , , √ (2)
tionships affecting scour depth was the selection of the b Vc s b b gb
parameters controlling the depth of scouring, that is, the
ds
characteristics of bed heights upstream and downstream where b
represents the maximum scour depth ratio, VV is
c
of the pier. The parameters that impact on the depth of y
the intensity of flow, Bb is the ratio of pier width, b is the
scouring around a bridge pier under clear water conditions ratio of flow depth, and √V represents the pier Froude
are illustrated in Fig. 1. In the current study, the important gb
number (Frp ). These dimensionless parameters were then
parameters affecting local scour depth were defined as
used to develop new scour depth formulae utilising GEP
the velocity of approaching flow (V), the critical velocity of
and NLR models.
approaching mean flow (Vc), flow depth (y), gravitational
acceleration (g), fluid density (ρ), median sediment size
(d50 ), channel width (B), pier width (b), the pier shape fac-
tor (Ks ), the correlation factor of flow alignment (K θ), and
flow time (t). Nine shapes of pier were used in this study:
3 Numerical simulation data sets
circular, square, rectangular, elliptic, oblong, ogival, len-
Localised scouring problems in bridge piers with differ-
ticular, hexagonal, and octagonal. All pier models were
ent shapes of piers and various conditions of flow have
aligned to the flow at a zero angle of alignment, causing
been studied numerically by applying Flow-3D model
the t correlation factor of flow alignment, K θ, to be equal
previously. In particular, flow-3D models have been used
to one in all cases. The time of flow was similarly set as
to state the maximum depth of scouring occurring at a
equal to 30 min. for all cases in this study. Studying the
bridge pier to define the critical parameters in bridge
remaining effective factors influencing the depth of local
design. To validate the efficacy of the Flow-3D model
scouring (ds) and its mechanisms thus allowed the func-
in terms of simulating the local scour depth at a bridge
tional relationships of general dimension analysis to be
pier, the results from the Melville [8] laboratory model
written as [17]
were compared with the results obtained from the rele-
vant Flow-3D numerical simulation model. The error rate
( )
ds = f1 V , y, g, d50 , Vc , B, b, Ks , K 𝜃, 𝜌, t (1)
obtained from this comparison of results was equal to
10%, indicating good validation between the numerical
The main reason for the use of dimensional analysis is to simulation model and the Melville [8] laboratory model,
formulate a problem in way that describes the relationship and suggesting that Flow-3D is an effective method
between the various quantities based on the selection of for simulating the depth of local scouring at the pier of
a bridge. The total number of data sets obtained from
numerical simulation was 243, each of which represented
the maximum depth of local scouring occurring at the
pier of a bridge with certain parameter values, based on
variations in intensity of flow, flow depth ratio, pier shape
factor, pier width ratio, and pier Froude number. The range
of these parameters is summarised in Table 1. These data
were then divided into an 80% training group and a 20%
validation group before being modelled in GEP, ANN, and
NLR to develop the required scour depth equations. Scour
depth ratio (ds/b) was taken as the dependent parameter
in the prediction formulae, while the other parameters
were deemed to be independent. Three statistical param-
Fig. 1  Flow and local scouring and flow features around a circular eters, ­R2, RMSE, and MAE, were then used to identify the
pier of the bridge [32]

Vol.:(0123456789)
Research Article SN Applied Sciences (2021) 3:159 | https://doi.org/10.1007/s42452-020-04124-9

Table 1  The limitations of the variables used in the training and GEP method is to create a simple arithmetical model, and
validating of the GEP and NLR models such arithmetical functions can be prepared by provid-
Variables Data Limitations ing a data set to the GEP model. The mathematical equa-
tion implemented in the GEP model during the current
Minimum Maximum
study was a symbolic regression on most GA genotypes.
V/Vc 0.55 1.00 The procedure of creating such a mathematical model
b/B 0.11 0.20 begins with the generation of a random chromosome
y/b 0.20 2.95 within a specific primary population; each chromosome
Ks 0.71 1.26 from the primary population was thus evaluated using
Frp 0.12 0.47 the fitness function against a group of fitness conditions.
ds/b 0.00 1.88 The selection of chromosomes relies on these fitness
values, as chromosomes with higher fitness values have
more chances to be selected for the next generation.
best techniques for predicting the maximum scour depth After the selection of chromosomes, genetic opera-
ratio (ds/b) at a given bridge pier. tions may be introduced to make modifications to the
selected chromosomes; these genetic operations include
inversion, mutation, transposition of gene, root insertion
4 Gene Expression Programming (GEP) sequence transposition (RIS), insertion sequence trans-
position (IS), recombination of the gene, and single or
GEP is a modern technology based on evolutionary artifi- double crossover/recombination, as described in more
cial intelligence (AI) that may be considered an extension detail in Ferreira [31]. Mutation is the most common
of genetic programming (GP). GEP combines the advan- genetic operation used to modify chromosomes. The
tages of both GP and genetic algorithms (GA) in terms process is then repeated until acceptable results have
of providing both modest and fixed length linear chro- been obtained or a certain number of generations has
mosomes in the form of genomes and different shapes passed [33].
and sizes of branched structures known as expression These processes as used to create a mathemati-
trees (ETs) in the form of phenotypes, similar to those cal model are represented in Fig. 2. GEP of this type
used in GP analysis trees [31, 32]. The goal of using the has recently been applied in several different fields,

Fig. 2  Gene expression algo-

Create Chromosome of Initial Population
rithm flow diagram [34]

Express Chromosome

Execute Each Program

Evaluate Fitness Iterate or Terminate End

Iterate

Keep Best Program

Select Program

Reproduction Steps

Prepare New Program of Next Generation

Vol:.(1234567890)
SN Applied Sciences (2021) 3:159 | https://doi.org/10.1007/s42452-020-04124-9 Research Article

including the transportation of sediment in sewer pipes The fourth stage managed the architecture of the
[34], the syntax of sentence ordering functions [35], the chromosomes based on the selection of head length and
development of rain runoff models [36], the prediction gene number, as incrementing gene number from one to
of hydraulic data [37], the modelling of time series [38], three in the chromosome helps with increasing the rate of
image compression, and multi-objective mining for clas- success according to Ferreira [33]. The adopted length of
sification base [39]. chromosome head (h) in this study was thus set to eight,
The powerful software package GeneXproTools 5.0 was while the selected number of genes per chromosome was
used in the current study to promote the development set to three. Five random floating types (Dc) in the range
of GEP-based models for predicting the ultimate depth of {−10,10} were then used to represent the random numeri-
scouring that may occur at the pier of a bridge. GEP pro- cal constants. In stage five, a connect function between
vides an arithmetical function of a scouring model to solve the sub expression trees (sub-ETs) was selected; as there
the problem using symbolic regression (function finding). were three genes per chromosome, the final equation
The resulting function can then be used to find a term that of these sub-ETs was linked by addition ( +), and a selec-
demonstrates the dependence satisfactorily. tion of genetic operators to make allow variations in both
type and rate of expression. A mix of all genetic opera-
tors was used in this study, including transposition (gene
transposition, IS, and RIS), mutation, (gene recombination,
5 Modelling the depth of local scouring both one-point and two-point), inversion, and Dc-specific
around a bridge pier genetic operations. The rates of these genetic operations
are illustrated in Table 2.
5.1 GEP Model
After determining all genetic parameters, the model
was simulated using GeneXproTools 5.0 for a number of
To develop the GEP model to predict local scouring occur-
generations in excess of 65,000. The resulting scour depth
ring around a bridge pier, 243 data sets were collected and
(ds/b) formula is represented in expression tree (ET) form
simulated numerically using Flow-3D software. These data
were then fed to the GEP in the form of a dependent out-
put variable (ds/b) and independent input variables
y
( VV , bB , b , Ks , Frp), with the output variable ds/b then devel- Table 2  GEP model parameters for the local scouring problem
c
oped using the GEP. The data were divided randomly by Parameters Values
the GEP into validation and testing data (20%) and training
Population size 50
data (80%) sets, with the training data used to build the
Set of function + , -, *, /, power
GEP model. The parameters and processes within the GEP
Set of terminals K𝜃 ,
y
, Ks , B
, V
,?
were then defined in six steps to facilitate the generation b b Vc

of the mathematical function required to predict local Random numerical constant (RNC) 05
scouring around a bridge pier. RNC type Floating point
The first step was to generate an initial population Range of RNC [-10, 10]
group. Any size of population may be used at this stage, Length of head 08
but the study done by Ferreira [33] suggested that a popu- Number of genes 03
lation of 30 to 100 offers optimal results. Several trials were Linking function +
thus done to select the optimum number, and the popula- Fitness function RMSE
tion used in the study was finally set to 50 chromosomes Rate of mutation 0.044
as this size of population offered the best results. The next Rate of inversion 0.1
stage was to measure the fitness function of each indi- Rate of IS transposition 0.1
vidual chromosome as calculated by RMSE. The third step Rate of RIS transposition 0.1
was determining groups of both functions and terminals Rate of Gene transposition 0.1
for each chromosome gene. The terminal set consisted of Rate of One-point recombination 0.1
the independent parameters and a random numerical con- Rate of Two-point recombination 0.3
y
stant (RNC), so that T = {VV , bB , b , Ks , Frp , X }, with X repre- Rate of Gene recombination 0.3
c Rate of Dc-specific mutation 0.044
senting the RNC, while basic arithmetical operations and
Rate of Dc-specific inversion 0.1
some mathematical functions were used for the function
Rate of Dc-specific IS transposition 0.1
set, in the form F = {+ , −, *, /, power}.
Rate of Random constant mutation 0.01

Vol.:(0123456789)
Research Article SN Applied Sciences (2021) 3:159 | https://doi.org/10.1007/s42452-020-04124-9

Fig. 3  GEP Expression Trees

(ET) to formulate the scour
depth formula

Table 3  Definition of Parameters Defini-

( )2 ( )2 b
ds V b V b V b B
parameters in ETs tion
= 7.24 ∗ ∗ − ∗ − ∗ + y
b VC B VC B VC B b
∗ V
− − b
B VC b B
d0 V/Vc
[( ( )2 ) ]
y b b V
d1 y/b + ∗ ∗ FrP + ∗ FrP ∗ Ks2 ∗
b B B VC
d2 b/B (3)
d3 KS
with the definition of parameters as used in Eq. 3 and
d4 FrP
the relevant ET represented in Table 3.
C1 7.24

5.2 NLR Predicting model

in Fig. 3, while the corresponding equation is offered in For the non-linear regression (NLR) model, the data
Eq. 3. required were divided in a similar manner to that used in
The depth of scour (ds/b) equation was thus the development of the GEP model (training and valida-
tion data sets). The first step in building this predicting
model was to try a linear model, which offered an ­R2 = 0.67,
suggesting that this failed to represent an accurate

Vol:.(1234567890)
SN Applied Sciences (2021) 3:159 | https://doi.org/10.1007/s42452-020-04124-9 Research Article

Training 6 Discussion
1.6

1.4
The main aim of this study was to evaluate GEP model effi-
ciency in predicting local depth of scouring at the pier of a
1.2
bridge by comparing the model’s performance with that of
ds/b predicted

1 the NLR model. A total data set of 243 cases were obtained
0.8 using Flow-3D numerical simulation software, and these
GEP
0.6
data were then used to develop new scour depth around
NLR
a bridge pier formula using GEP and NLR models. The
0.4
predicted scour depth was computed using both GEP
0.2 and NLR models and plotted against the measured scour
0 depth, as represented in the scatterplots in Figs. 4, 5 for
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
the training and testing data sets, respectively. The rea-
ds/b observed
son for using scatterplots was to investigate the degree
of similarity between the measured ​​and the predicted
Fig. 4  Comparison between GEP and NLR modes for training data values more intuitively. The statistical measures R ­ 2, MAE,
and RMSE were also calculated for all models, as illustrated
in Table 4. The statistical results shown in Table 4 identify
Tesng the equation that offers the fewest errors in prediction,
suggesting that GEP performs better than NLR. The GEP
1.6
model produced a higher value of ­R2 (0.901) and smaller
1.4 values for RMSE (0.141) and MAE (0.111), with less scatter
around the line of agreement than the NLR. The unique
ds/b ds/b predicted

1.2

1
properties of the GEP model also provide a clear, easy-to-
use experimental expression of the bridge pier depth of
0.8 GEP
scour model, as represented in Eq. 3. Although the NLR
0.6 NLR model offered worse performance than the GEP model, it
0.4 still gave reasonably good results; GEP’s main contribution
0.2 is thus this compressed and explicit arithmetic expression,
0
which is likely to be useful for future designers.
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

ds/b observed
7 Sensitivity test
Fig. 5  Comparison between GEP and NLR modes for testing data
In order to determine the influence of each of independ-
ent parameter on the predicted scour depth, to ensure
prediction of scour depth. It was also found that all appro- the correct factors receive extensive attention in future
priate prediction models were nonlinear. Equation (4) thus studies, sensitivity testing was adopted to determine the
displays the relationship between the relative scour depth most sensitive parameters. Many factors affect the value
(ds/b) and other independent parameters. of local scouring depth around a bridge pier, including the

ds V y b
( y )2
= 0.496 − 9.7 + 0.01 − 3.29 − 0.388KS + 9.12Frp − 0.282 (4)
b VC b B b

flow intensity, width of pier, shape of pier, and depth of

flow. Various different input combinations were studied,

Table 4  The statistical results Models Training Testing

for GEP and NLR models
2
R RMSE MAE R2 RMSE MAE

GEP 0.917 0.139 0.106 0.901 0.141 0.111

NLR 0.858 0.168 0.135 0.832 0.170 0.148

Vol.:(0123456789)
Research Article SN Applied Sciences (2021) 3:159 | https://doi.org/10.1007/s42452-020-04124-9

Table 5  Sensitivity test for input parameters for the testing data Equation 3, obtained from the GEP model, can thus
GEP models Model parameters function R 2
RMSE help bridge pier designers to identify the maximum
depth of scour that may occur at a bridge pier under
Model 1 0.901 0.141 conditions such as those covered in this study. The
( )
ds y
b
= f Frp , b , Ks , Bb , VV
parameter limitations for Eq. 3 thus include intensity of
c

Model 2 0.734 0.246

( )
ds y B
b
= f Frp , b , Ks , b flow (0.55 to 1), ratio of flow depth (0.2 to 2.95), ratio of
Model 3 ds
(
y
)
0.757 0.194 pier width (0.11 to 0.2), pier shape factor (0.71 to 1.26)
= f Frp , b , Ks , VV
b c and for pier Froude number (0.12 to 0.47). In terms of
Model 4 0.889 0.159 future studies, the further development of Eq. (3) to
( )
ds y B V
= f Frp , b , b , V
b c
include the influence of sediment size, alignment angle,
Model 5 0.86 0.178 and channel width, by using different ranges of these
( )
ds y B V
b
= f b , Ks , b , V
parameters to study their effects on local depth of scour,
c

Model 6 0.338 0.388

( )
ds B V
b
= f Frp , Ks , b , V
c
is recommended.

Compliance with ethical standards

as illustrated in Table 4, with GEP models compared with
each independent parameter by eliminating one input Conflict of interest On behalf of all authors, the corresponding au-
thor states that there is no conflict of interest.
parameter in each case and determining the effect on the
expected scour depth in terms of the root mean square Open Access This article is licensed under a Creative Commons Attri-
error (RMSE) and coefficient of determination ­(R2), the bution 4.0 International License, which permits use, sharing, adap-
same statistical parameters used for evaluating the main tation, distribution and reproduction in any medium or format, as
performance criteria. The results in Table 5 and Fig. 6 indi- long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons licence, and indicate
cate that the depth of the flow approach has the most if changes were made. The images or other third party material in this
significant influence on the predicted depth of scouring article are included in the article’s Creative Commons licence, unless
after eliminating any remaining input parameters with no indicated otherwise in a credit line to the material. If material is not
significant effect on the predicted depth of scouring. included in the article’s Creative Commons licence and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this licence, visit http://creat​iveco​mmons​
8 Conclusion .org/licen​ses/by/4.0/.

The local scouring that occurs around a bridge pier is a

complex phenomenon that can be difficult to measure; References
however, there is an urgent need to predict the depth
of such scouring accurately to ensure safety. Numeri- 1. Annandale GW (1995) Erodibility. J Hydraulic Res 33(4):471–494
2. Nou MRG, Moghaddam MA, Bajestan MS, Azamathulla HM
cal simulations were used to obtain 243 data sets that (2020) Control of bed scour downstream of ski-jump spillway
were then used to generate mathematical models that by combination of six-legged concrete elements and riprap. Ain
allowed a comparison of the use of GEP and NLR [5]. Shams Eng J 11(4):1047–1059
The results suggested that the performance of the local 3. Shepherd R, Frost JD (1995) Failures in civil engineering: struc-
tural, foundation and geoenvironmental case studies. ASCE,
scour depth formula derived from the GEP model was New York, NY
better than that derived from the conventional NLR 4. Cheremisinoff PN, Cheremisinoff NP, Cheng SL, Bert CW (1988)
model, as shown in Table 4. The GEP produced smaller Civil Engineering practice: water resources environmental, vol
values of RMSE (0.141) and MAE (0.111) and a greater 5. CRC. Technomic Publishing AG, USA, p 809
5. Jalal HK (2019) “Numerical Study of Optimum Pier Shape for Safe
value of ­R 2 (0.901). The GEP model was also charac- Bridge.” M.SC. thesis, University of Kerbala
terised by a clear and brief equation suitable for easy 6. Najafzadeh M, Saberi-Movahed F, Sarkamaryan S (2018)
use by bridge engineers. This feature makes the GEP NF-GMDH-Based self-organized systems to predict bridge pier
option more effective as compared to the NLR model. scour depth under debris flow effects. Mar Georesour Geotech-
nol 36(5):589–602
The GEP model thus offers an effective modelling tool 7. Najafzadeh M, Barani GA, Kermani MRH (2013) Abutment scour
for predicting local scour depth as well as providing a in clear-water and live-bed conditions by GMDH network. Water
simple empirical expression for the modelled response Sci Technol 67(5):1121–1128
function. Sensitivity analysis then showed that the flow 8. Melville BW (1975) Local scour at bridge sites. School of Engi-
neering, Report No. 117, University of Auckland
depth parameter has impact on scour depth prediction
among the input parameters examined.

Vol:.(1234567890)
SN Applied Sciences (2021) 3:159 | https://doi.org/10.1007/s42452-020-04124-9 Research Article

9. Najafzadeh M, Rezaie Balf M, Rashedi E (2016) Prediction of max- 27. Azmathullah HM, Deo MC, Deolalikar PB (2005) Neural networks
imum scour depth around piers with debris accumulation using for estimation of scour downstream of a ski-jump bucket. J
EPR, MT, and GEP models. J Hydroinformatics 18(5):867–884 Hydraul Eng 131(10):898–908
10. Yang Q (2005) Numerical investigations of scale effects on local 28. Najafzadeh M, Oliveto G (2020) Riprap incipient motion for over-
scour around a bridge pier. Master thesis at the Department of topping flows with machine learning models. J Hydroinformat-
Civil Engineering in the Florida. State University ics 22(4):749–767. https​://doi.org/10.2166/hydro​.2020.129
11. Little PR (1978) Scour at bridge piers and abutments. Doctoral 29. Abdallah Y, Moussa M (2013) Modeling of local scour depth
dissertation, University of Cape Town downstream hydraulic structures in trapezoidal channel using
12. Gacek JD (2007) “Numerical Simulation of Flow through A Spill GEP and ANNs. Ain Shams Eng J 4(4):717–722. https​://doi.
Way and Diversion Structure” M.SC. Thesis org/10.1016/j.asej.2013.04.005
13. Vasquez J and Walsh B (2009) “CFD simulation of local scour in 30. Khan M, Azamathulla HM, Tufail M (2012) Gene-expression pro-
complex piers under tidal flow.” Proc thirty-third IAHR Congr gramming to predict pier scour depth using laboratory data. J
Water Eng a Sustain Environ 604: 913–920 [Online]. Available: Hydroinformatics 14(3):628–645
http://flow3​d2.propa​gatio​n.net/pdfs/tp/wat_env_tp/cfd-simul​ 31. Ferreira C (2006) Gene expression programming: mathemati-
ation​-of-local​-scour​-in-compl​ex-piers​-under​-tidal​-flow-23-09. cal modeling by an artificial intelligence, vol 21. Springer,
pdf. Netherlands
14. Fattah MY, Hassan WH, Rasheed SE (2018) Behavior of flexible 32. Hassan WH (2019) Application of a genetic algorithm for the
buried pipes under geocell reinforced subbase subjected to optimization of a location and inclination angle of a cut-off wall
repeated loading. International Journal of Geotechnical Earth- for Anisotropic foundations under hydraulic structures. Geo-
quake Engineering (IJGEE) 9(1):22–41 tech Geol Eng 37(2):883–895. https​://doi.org/10.1007/s1070​
15. Kais Jalal H, Hassan WH (2020) Three-dimensional numerical 6-018-0658-9
simulation of local scour around circular bridge pier using Flow- 33. Ferreira C (2001) Gene expression programming: a new adaptive
3D software. MS&E 745(1):012150 algorithm for solving problems. arXiv preprint cs/0102027, pp
16. FLOW-3D manual, “FLOW-3D user manual, version 11.” Flow Sci- 1–22
ence Santa Fe, NM, 2014. 34. Ab Ghani A, Md Azamathulla H (2011) Gene-Expression Pro-
17. Jalal HK, Hassan WH (2020) Effect of Bridge Pier Shape on Depth gramming for Sediment Transport in Sewer Pipe Systems. J
of Scour. In: IOP Conference Series: Materials Science and Engi- Pipeline Syst Eng Pract 2(3):102–106. https​://doi.org/10.1061/
neering, vol 671, No 1. Institute of Physics Publishing, p 012001 (asce)ps.1949-1204.00000​76
18. Azamathulla HM, Ghani AA, Zakaria NA, Guven A (2010) Genetic 35. Xie Z, Li X, Di Eugenio B, Xiao W, Tirpak TM, Nelson PC (2004)
programming to predict bridge pier scour. J of Hydraul Eng Using gene expression programming to construct sentence
136(3):165–169 ranking functions for text summarization. In: COLING 2004:
19. Mohamed TA, Pillai S, Noor MJMM, Ghazali AH, Huat BK, Yusuf B Proceedings of the 20th International Conference on Compu-
(2006) Validation of some bridge pier scour formulae and mod- tational Linguistics, pp 1381–1384
els using field data. J King Saud Univ Eng Sci 19(1):31–40 36. Fernando AK, Shamseldin AY, Abrahart RJ (2009) Using gene
20. Richardson EV and Davis SR (1995) “Evaluating scour at bridges.,” expression programming to develop a combined runoff esti-
United States. Federal Highway Administration. Office of Tech- mate model from conventional rainfall-runoff model outputs. In:
nology Applications Proc., 18th World IMACS Congress and MODSIM09 International
21. Laursen EM, Toch A (1956) Scour around bridge piers and abut- Congress on Modelling and Simulation, pp 2377–2383
ments, vol 4. Iowa Highway Research Board, Ames, IA. Hunter 37. Eldrandaly KA, Negm AA (2008) Performance evaluation of gene
Rouse, Director, Iowa City, p 63 expression programming for hydraulic data mining. Int Arab J
22. Melville BW, Sutherland AJ (1989) Design Method for Local Inf Technol 5(2):126–131
Scour at Bridge Piers. J Hydraulic Eng 114(10):1210–1226 38. Bărbulescu A, Băutu E (2009) Time series modeling using an
23. Muzzammil M, Alama J, Danish M (2015) Scour prediction at adaptive gene expression programming algorithm. Int J Math
bridge piers in cohesive bed using gene expression program- Models Methods Appl Sci 3(2):85–93
ming. Aquat Procedia 4:789–796 39. Dehuri S, Cho SB (2008) Multi-objective classification rule min-
24. Najafzadeh M, Barani G (2011) Comparison of group method of ing using gene expression programming. In: 2008 Third Inter-
data handling based genetic programming and back propaga- national Conference on Convergence and Hybrid Information
tion systems to predict scour depth around bridge piers. Sci Iran Technology, vol 2. IEEE, pp 754–760
18(6):1207–1213. https​://doi.org/10.1016/j.scien​t.2011.11.017
25. Najafzadeh M, Azamathulla HM (2015) Neuro-fuzzy GMDH to Publisher’s Note Springer Nature remains neutral with regard to
predict the scour pile groups due to waves. J Comput Civ Eng jurisdictional claims in published maps and institutional affiliations.
29(5):4014068
26. Najafzadeh M, Barani G, Reza M, Kermani H (2013) GMDH based
back propagation algorithm to predict abutment scour in cohe-
sive soils. Ocean Eng 59:100–106. https​://doi.org/10.1016/j.
ocean​eng.2012.12.006

Vol.:(0123456789)