Machine Learning Intelligence To Assess The Shear Capacity of Corroded Reinforced Concrete Beams

www.nature.
com/scientificreports
OPEN Machine learning intelligence

to assess the shear capacity
of corroded reinforced concrete
beams
Aman Kumar 1,2*, Harish Chandra Arora 1,2, Nishant Raj Kapoor 1,3, Krishna Kumar 4,
Marijana Hadzima‑Nyarko 5,6 & Dorin Radu 6
The ability of machine learning (ML) techniques to forecast the shear strength of corroded reinforced
concrete beams (CRCBs) is examined in the present study. These ML techniques include artificial
neural networks (ANN), adaptive-neuro fuzzy inference systems (ANFIS), decision tree (DT) and
extreme gradient boosting (XGBoost). A thorough databank with 140 data points about the shear
capacity of CRCBs with various degrees of corrosion was compiled after a review of the literature.
The inputs parameters of the implemented models are the width of the beam, the effective depth
of the beam, concrete compressive strength (CS), yield strength of reinforcement, percentage of
longitudinal reinforcement, percentage of transversal reinforcement (stirrups), yield strength of
stirrups, stirrups spacing, shear span-to-depth ratio (a/d), corrosion degree of main reinforcement,
and corrosion degree of stirrups. The coefficient of determination of the ANN, ANFIS, DT, and XGBoost
models are 0.9811, 0.9866, 0.9799, and 0.9998, respectively. The MAPE of the XGBoost model is
99.39%, 99.16%, and 99.28% lower than ANN, ANFIS, and DT models. According to the results of
the sensitivity examination, the shear strength of the CRCBs is most affected by the depth of the
beam, stirrups spacing, and the a/d. The graphical displays of the Taylor graph, violin plot, and multi-
histogram plot additionally support the XGBoost model’s dependability and precision. In addition, this
model demonstrated good experimental data fit when compared to other analytical and ML models.
Accurate prediction of shear strength using the XGBoost approach confirmed that this approach is
capable of handling a wide range of data and can be used as a model to predict shear strength with
higher accuracy. The effectiveness of the developed XGBoost model is higher than the existing models
in terms of precision, economic considerations, and safety, as indicated by the comparative study.
One of the most important construction activities being carried out around the world today is the improvement of
existing infrastructures. In terms of sustainable development, it is becoming necessary to upgrade old structures
rather than demolish them. In many countries, design codes are constantly being updated due to increased load
requirements that demand greater strength in structural components. Aging structures deteriorate over time as
a result of environmental factors. Existing concrete structures can deteriorate due to reinforcement corrosion,
carbonation, freeze–thaw cycles, etc. One of the most common causes of reinforced concrete (RC) elements
degradation is corrosion, which reduces the cross-sectional area ( Acs) of reinforcement bars, degrades their
mechanical characteristics causes the concrete surface to crack or flake, and damages the steel to concrete bond1.
Due to the gradual loss of the steel area, the propagation of the damage in the form of cracks and the final spalling
of the concrete cover, and the deterioration of the connection between the steel reinforcement and the concrete,
the bearing capacity of the corroding element decreases with the time of corrosion. Corrosion reduces the ability
of the structures to support the load, which has an impact on both structural safety and in-service performance.
1
Academy of Scientific and Innovative Research (AcSIR), Ghaziabad 201002, India. 2Structural Engineering
Department, CSIR-Central Building Research Institute Roorkee, Roorkee 247667, India. 3Department of
Architecture and Planning, CSIR-Central Building Research Institute Roorkee, Roorkee 247667, India. 4Department
of Hydro and Renewable Energy, Indian Institute of Technology Roorkee, Roorkee 247667, India. 5Faculty of Civil
Engineering and Architecture Osijek, J. J. Strossmayer University of Osijek, Vladimira Preloga, Croatia. 6Faculty of
Civil Engineering, Transilvania University of Braşov, Braşov, Romania. *email: aman.civil16@outlook.com
Scientific Reports | (2023) 13:2857 | https://doi.org/10.1038/s41598-023-30037-9 1
Vol.:(0123456789)
www.nature.com/scientificreports/
A well-known form of failure of RC elements is a shear failure which is very fast, without warning, failure,
thus the RC elements must have sufficient shear capacity2. However, in service life, concrete structures undergo
various types of environmental changes. Considering the corrosion, in the shear area, the small diameter of the
stirrup and the brittle failure of the RC beam makes it more vulnerable. But still, these structure members take
the load but when these damages increase more then it becomes very difficult to retain the loads. As shear failures
are very dangerous, it is necessary to get early shear strength prediction using novel techniques.
A substantial amount of study has been conducted on many aspects of reinforcement corrosion relating to
the corrosion process, its beginning, and undesirable impacts such as strength decrease and prediction of the
residual strength of corroded components. The shear capacity of corroded RC beams (CRCBs) has recently been
calculated using a variety of analytical models. Several formulas proposed by researchers, namely, Xu and Niu3,
Yu4, Huo5, Zhao and J in6, Li et al.7, Higgins et al.8, Webster9, Xue et al.10, and Khan et al.11 have been listed in the
previous studies and used calculate the shear strength of the CRCBs. The shear performance of the CRCBs has
been studied well through experimentation, however, extensive testing may be costly and time-consuming. The
suggested models only take into account a small number of variables, leaving out certain significant ones that
define the level of corrosion and other important parameters. As the analytical models were built on various
model assumptions and various experimental databases, the accuracy and applicability of these models vary
greatly under various situations. Additionally, the ability of analytical models to precisely forecast the shear
capacity of degraded beams is constrained by the quick rise in corrosion levels in RC beams. Therefore, it is
crucial to create methodologies that take into account all controlling factors. It is undoubtedly difficult to create
an advanced shear prediction model that takes into account all the important factors since the shear capacity of
CRCBs depends on a number of variables that are difficult to quantitatively characterize without making several
assumptions.
Modern machine learning (ML) algorithms may offer a better solution to address these concerns because
they are effective at handling complex problems involving many different factors without making any assump-
tions. These algorithms might be used to construct a shear capacity prediction model. Numerous services and
sectors have seen significant productivity gains due to ML. Although it is still in its infancy in the construction
sector, its application has grown recently to address several difficulties, including concrete t echnology12–16 and
concrete durability17–19. The other various application in the field of civil and environmental engineering of ML
algorithms can be found in the published research work20–24. The following is a brief description of some selected
research related to the use of Extreme Gradient Boosting (XGBoost) and other ML (Classification and regression
tree (CART), adaptive boosting (AdaBoost), gradient boosted decision trees (GBDT), support vector regression
(SVR), random forest (RF), extremely random trees (ERT), artificial neural network (ANN), gene expression
programming (GEP), Kernel ridge regression (KNN), K-nearest neighbor (KRR), Gaussian process regression
(GPR) multivariate adaptive regression splines (MARS), support vector machine (SVM), linear regression (LR),
decision tree (DT) ensemble tree (ET), and evolutionary polynomial regression (EPR)) algorithms in concrete
technology and RC beams:
Wakjira et al.25 investigated the flexural capacity of FRP-reinforced RC beams. Six input parameters were
used to develop the ML models. Five ML algorithms (CART, AdaBoost, GBDT, XGBoost, and Super-learner)
were used to develop the most appropriate predictive model. The super-learner algorithm provides the highest
predictive performance among all the models studied, with the lowest RMSE and MAPE and as the highest
R2. In another study, the shear capacity of the shear-critical RC beams reinforced with novel composites was
investigated by Wakjira et al. using ML techniques26. The prediction models were built using SVR, CART, RF,
extremely random trees, GBDT, and XGBoost algorithms. According to the results of the study findings, the
XGBoost model performs well compared to other ML techniques and existing guidelines. Uddin et al.27 used
ANN, RF, GEP, and GBDT to predict the shear strength of RC beams. The performance of GBDT algorithm
was good compared to ANN, RF and GEP algorithms. In another study, Wakjira et al.28 investigated the flexural
capacity of FRCM-reinforced RC beams using KNN, KRR, SVR, CART, RF, GBDT and XGBoost methods. The
XGBoost model shows good performance and having the highest R 2-value of 99.3%, the lowest MAE, and the
MAPE. The proposed model has better predictive power and robustness as shown by its performance with that
of existing analytical models. Based on the above mentioned research work, the performance of the XGBoost
algorithm was higher compared to other ML algorithms.
Badra et al.29 predicted the punching shear strength of FRP-reinforced concrete slabs without shear rein-
forcement using ANN and SVM algorithms. The RMSE value of the ACI, CSA, JSCE, ANN, and SVM models
are 3.06 kN, 1.70 kN, 1.99 kN, 1.10 kN, and 1.32 kN, respectively. When comparing the performance metrics
the performance of ANN was superior. Deifalla and Salem30 investigated the torsional strength of externally
bonded FRP-reinforced concrete beams using ET, GPR and ANN models. The broad neural network model was
the most effective model for predicting the torsion strength of RC beams strengthened with EB-FRP. The R 2,
RMSE, and MAE values of the models were 0.93, 16,634 kN and 0.98 kN, respectively, and they reported the best
performance; however, they required the most training time. Mohammed and Ismail31 used MARS, XGBoost and
SVM models to predict the shear strength of RC beams. According to the research results, the developed MARS
and XGBoost models for simulating the shear strength of RC beams have potential. The results showed that all
the beam geometry and concrete properties criteria used were important for building the prediction model.
Numerically, the MARS model achieved the lowest possible RMSE (89.96 kN). Salem and D eifalla32 evaluated
the strength of slab-column connections with FRPs using ML algorithms (LR, DT, SVM, ET, and GPR). The
ideal hyper-parameters of the ML-based algorithms were selected during the training process using a grid search
with a 15-fold cross-validation. Among all the applied ML models, the ensemble boosted model was found to
be the most trustworthy and accurate model, with the best accuracy: R2, RMSE, and MAE were 0.97, 71.963 kN,
and 43.452 kN, for the test dataset, respectively. Ebid and D eifalla33 used ML procedures to predict the punch-
ing shear capacity of lightweight concrete slabs. The column dimensions, concrete density, slab effective depth,
Vol:.(1234567890)
CS, yield strength of steel, and flexural reinforcement ratio were considered as input parameters. The highest
prediction accuracy is shared by ANN and EPR (73.9% and 73.6%, respectively), while the GP model has the
lowest prediction accuracy (67.6%). Kaveh et al.34 used an XGBoost framework to calculate the shear capacity of
FRP-strengthened concrete beams. The correlation coefficient of the developed XGBoost model was 0.94, which
was higher than all the empirical models.
The aim of this article is to estimate the shear strength of CRCBs using ANN, ANFIS, DT, and XGBoost
algorithms. To the best knowledge of the authors, DT, ANFIS, and XGBoost algorithms have not been previously
used to estimate the shear strength of CRCBs. The influence of individual input parameters on the predicted
shear capacity of CRCBs is also determined.
Analytical models to estimate the shear capacity of CRCBs

Xu and Niu’s model. The limit equilibrium theory serves as the foundation for the formula used by Xu and
iu3 to determine the shear strength of CRCBs. The impact of reinforcing steel corrosion on shear capacity is
N
taken into consideration by adding a shear span-to-depth ratio (a/d) as well as the reduction in A
cs of the stirrup
and yield strength due to corrosion. The formulation is expressed in Eqs. (1–4):

(0.08 + 4ρl ) ho
Vu = Vc + Vs = ξ , ηw,sn × × fck bho + α ηw,sn × (0.25 + 0.4)Avc fyv (1)
( − 0.3) s

1, ηw,sn ≤ ηcr,sn
0.069−0.43
ξ , ηw,sn = ηw,sn (2)
ηcr,sn , ηw,sn > ηcr,sn
cv fcu,150
ηcr,sn = 10.4 + (3)
φv2 φv

α ηw,sn = 1 − 1.077ηw,sn (4)
where, Vu , Vc , and Vs represents
the
shear resistance
of the RC beam, concrete resistance, and stirrups shear
resistance, respectively. ξ , ηw,sn and α ηw,sn considered as reduction factors. , fck , fyv Avc , fcu,150 , and φv
are a/d, CS of concrete, stirrups yield strength, residual area of the stirrups, CS of concrete (150 mm cube), and
diameter of stirrups, respectively. ρl , ηw,sn , and ηcr,sn are the percentage of longitudinal steel, percentage of stir-
rups corrosion, and crack initiation of stirrups corrosion ratio, respectively. b, ho, s and cv are the beam width,
effective depth, stirrups spacing, and concrete cover of stirrups, respectively.
Yu’s model. Yu has suggested a small change to the GB50010-200235 standard guideline. A new coefficient
was proposed by Yu4 to identify the impact of corrosion on longitudinal steel. The model also takes into account
how corrosion affects the A cs and stirrups yield strength4. The formulation to forecast the shear capacity of
CRCBs is expressed in Eqs. (5–7):
1.75ϕft bho 1.25fyc Avc ho
Vu = + (5)
( + 1) s
2
ϕ = −0.0354ηl,sn + 0.6256nl,sn − 1.2349 (6)

0.985 − 1.028ηw,sn fyv
fyc = (7)
1 − ηw,sn
where, ϕ , fyc , ft and nl,sn are the reduction factor, yield strength of stirrups (corroded), the tensile strength of
concrete, and section loss ratio of main steel.
Huo’s model. Huo5 introduced two reduction factors on the basis of the shear strength analytical model of
an un-corroded RC beam proposed by the China Academy of Building Research in 198536 to take into account
stirrup corrosion and longitudinal steel of CRCBs. In experimental tests of CRCBs, regression analysis is used to
determine both of these reduction v ariables5. The formulation to estimate the shear strength of CRCBs is given
in Eqs. (8–10):

0.08 100ρl α(0.4 + 0.3)Av fyv ho
Vu = ϕfck bho + + (8)
( − 0.3) .fck s

1.0, ηl,wt ≤ 5%
ϕ=
1.098 − 1.96ηl,wt , ηl,wt > 5% (9)
α = 1 − 1.059ηw.wt (10)
Vol.:(0123456789)
where, both ϕ and α are the reduction factors ( ηl,wt and ηw.wt ), and Av is the un-corroded area of the stirrup.
Zhao and Jin’s model. A methodology for estimating the shear capacity of un-corroded RC beams under
two-point loading was provided by Zararis37. Zhao and Jin6 suggested a variation of Zararis’s model, which is
applied to calculate the shear strength of CRCBs. Zhao and J in6 take into account a reduction factor that includes
all of the effects of corrosion on stirrups. The formulation to forecast the shear capacity of CRCBs is expressed
in Eqs. (11–13):
 � � � �2 � �2 
0.5Cs Cs a
 C s fcyl,150 1 − ho 0.5ρv fyv 1 − ho ho 
Vu = αVu0 = αbho  + � �  (11)
ho a
ho

1.0, ηw,sn ≤ 10%
α=
1.17 − 1.7ηw,sn , ηw,sn > 10% (12)
��
2 � 
1 + 0.27 hao + ρρvl � �2 � �
Cs 600ρ l 600ρ l 600ρ l
= � � �2 �  +4 −  (13)
ho fcyl,150 fcyl,150 fcyl,150
2 1 + hao + ρρv
l
where, Vu0 , α , Cs , ρv , and fcyl,150 are the ultimate shear resistance of uncorroded beams, shear span, compression
zone depth, percentage steel of stirrups, and CS of concrete (150 × 300 mm specimens), respectively.
Li et al.’s model. Li et al.7 also proposed an equation based on the Chinese Guideline (GB50010-2002)35 for
the shear capacity estimation of CRCBs. The equation takes into account the change in height and width of the
corrosion-damaged stirrups cross-section. Additionally, stirrups’ corrosion and yielding strength are taken into
consideration. The formulation to evaluate the shear capacity of CRCBs is expressed in Eqs. (14–17):
1.75ft bc hoc fyc Avc hoc
Vu = + (14)
( + 1) s

1 − 1.1219ηw,wt fyv
fyc = (15)
1 − ηw,wt
bc = b − Cv1 − Cv2 (16)
h0c = h0 − Csc (17)

where, Cv1 and Cv2 are concrete cover on both cross-sectional width directions.
Lu et al.’s model. Lu et al.,38 incorporate the impacts of stirrups, the level of corrosion of longitudinal
reinforcement, and the a/d in the shear strength estimation of the CRCBs. Overestimating the residual shear
strength of CRCBs with stirrups and longitudinal reinforcement corrosion would be risky. In addition, diagonal
tension failure and shear compression failure were also considered, and the maximum value is taken for final
calculations. The shear strength of CRCBs under purposive stress is shown in Eqs. (18–23):
Vu = φVc + Vs (18)
where, φ is a reduction coefficient linked with a/d.
Shear resistance of concrete ( Vc ) can be calculated using Eq. (19)39:
Vc = Max(Vc1 , Vc2 ) (19)
40
where, the terms Vc1 is shear resistance of concrete at diagonal tension and expressed in Eq. (20) :

3
4 10 1.4h0
3
Vc1 = 0.2 100fcyl,150 ρlc 0.75 + bh0 (20)
h0 a
where, ρlc is the percentage of corroded longitudinal steel, and Vc2 represent the compression failure and can be
determined by Eq. (21)41:
√
2
0.24 3 fcyl,150 1 + 100ρlc 1 + 3.33r
h0 bh0
Vc2 = 2 (21)
1 + ha0
Vol:.(1234567890)
where, r is the width of the loading plate (87.2 mm).

Shear resistance of the stirrups ( Vs ) is expressed in Eq. (22)10.
fyv Avc jh0
Vs = (22)
s
where, j is a coefficient and is generally taken as 1/1.1510.

0.008e(−0.122) − 0.003ηw,sn + 1.01, < 2.5
φ= (23)
0.1e(−0.122) − 0.003ηw,sn + 1.38, ≥ 2.5
Machine learning models establishment

Database setting up. A literature survey was conducted to collect experimental data on the shear capacity
of CRCBs. 140 datasets have been collected from the reviewed literature6,10,42–47. The parameters that affect the
shear strength of CRCBs are: (i) width of the beam (b), (ii) effective depth of the beam (d), (iii) CS of concrete
(fck), (iv) yield strength of reinforcement (fy), (v) percentage of longitudinal reinforcement ( ρl ) , (vi) percentage
of stirrups reinforcement ( ρv ) , (vii) yield strength of stirrups (fyv), (viii) stirrups spacing (s), (ix) a/d, (x) corro-
sion degree of longitudinal reinforcement ((ηl)), and (xi) corrosion degree of stirrups ( ηw ). The same parameters
have been used to develop ML models. The complete methodology to accomplish the objective of this work is
depicted in Fig. 1 and explained in subsequent sections. Table 1 lists the statistical properties of the amassed
database, and Fig. 2 displays the distribution of the collected parameters.
Pu
D d
Dataset Analytical Models (Six) L
Input parameters
ANN
Beam width, b (mm) R

ANFIS Training (70%)
Beam depth, d (mm) MAPE
DT Testing (30%) CS of concrete, fck (MPa) Scatter plot
MAE
Performance Metrics
Yield strength of steel, fy (MPa)
Best Fitted Model

2D Kernel Density Plot
XGBoost RMSE
Percentage of Long. Reinf., ρl (%) Taylor Plot
Percentage of Stirr. Reinf., ρv (%) NS
...
...
...
...
...
...
...
...
...
...
Violin Plot
Yield strength of stirrups, fyv (MPa) a20-index
10-fold cross-validation Stirrups spacing, s (mm) Multi-panel Histogram
n_estimators = 100 RRMSE
XGBoost
Shear span-to-depth ratio, λ

Hyper parameters tuning
alpha = 10 Feature Importance

subsample = 0.9
learning_rate = 0.4 Corrosion degree of Long. Reinf., ηl (%) Pi
random_state =123
Corrosion degree of stirrups, ηw (%) OBF
max_depth = 10 Longitudinal Reinforcement = Long. Reinf.
min_samples_split = 5 Stirrups Reinforcement =Stirr. Reinf.
Output parameter
DT
min_samples_leaf = 2
max_features = auto
max_leaf_nodes =100 Shear capacity of corroded RC beam Vu,exp (kN)
Figure 1. Methodology chart.
Index b (mm) d (mm) fck (MPa) fy (MPa) ρl (%) ρv (%) fyv (MPa) s (mm) λ ƞl (%) ƞw (%) Vu,exp (kN)
Mean 166.50 195.79 22.60 444.70 2.06 0.36 403.53 154.18 2.54 3.89 21.85 116.60
Std 45.60 92.84 5.59 96.22 0.62 0.19 105.06 46.89 0.98 6.25 25.20 89.47
Min 120 130 16.7 300 0.67 0.14 275 80 1 0 0 26.60
25% 120 150 17.33 390 1.65 0.23 332 120 1.76 0 0.19 80
50% 150 155 21.09 417.80 2.15 0.31 369.60 150 2.20 0 15.50 96.05
75% 200 207.50 27.89 443.10 2.58 0.45 467 200 3.10 8.73 32.60 122.40
Max 260 521 32.4 706 2.79 0.90 626 305 4.70 26 97.20 594
Skewness 0.83 1.93 0.36 1.26 − 0.86 1.53 0.73 0.40 0.66 1.35 1.31 3.66
Kurtosis 2.54 6.62 1.58 3.99 2.94 4.72 2.36 3.03 2.51 3.62 4.12 1.78
Table 1. Statistical parameters.
Vol.:(0123456789)
120 140 150 200 260

b (mm) 46 44 32 18
130 155 190 521

d (mm) 71 34 35
16.7 17.95 24.37 27.2 32.4

fck (MPa) 56 28 19 37
300 369 420 443.1 706

fy (MPa) 34 46 29 31
0.67 1.9 2.56 2.62 2.79

ρl (%) 51 54 17 18
0.14 0.19 0.3 0.39 0.9

ρv (%) 20 50 32 38
275 300 373 458 560 626

fyv (MPa) 22 59 18 10 31
80 120 170 200 305

s (mm) 39 63 27 11
1 1.76 2.6 3.5 4.7

λ 37 48 32 23
0 6 13.8 26
ƞl (%) 101 23 16
0 0.1 4 17.6 97.2

ƞw (%) 32 23 20 65
26.6 80 139 594

Vu,exp (kN) 40 83 17
Figure 2. Summary of statistical analysis of the CRCBs.
To determine the relationship between input parameters and shear capacity, and to show the dot distribution
a marginal plot is used, as shown in Fig. 3. A scatterplot with histograms, boxplots, or dot-plots in the x and
y-axes’ margins is referred to as a marginal plot. Figure 3a–k shows the marginal plot of all the input parameters
like b, d, fck, fy, ρl , ρv , fyv, s, , ηl and ηw , respectively.
Data preparation. In all ML algorithms, it is necessary to standardize the dataset in a certain form. The process
of standardization increases the efficiency and accuracy of ML algorithms. The commonly used standardization
ranges are: (i) 0–1, (ii) − 1 to + 1, and (iii) 0–9. In this study, the − 1 to + 1 standardized range has been adopted
to normalize the collected parameters. The formulation used for normalization is expressed in Eq. (24)48:

(z − zmin )
Znormalized = 2 × −1 (24)
(zmax − zmin )
where, Znormalized is the normalized outcome, z is the value to be standardized in the selected dataset, zmin and
zmax are the minimum and maximum values in the selected dataset, sequentially.
After normalization, the dataset has been processed for further processing in the different phases such as the
training, and testing phases. In ANN, ANFIS, DT, and XGBoost models, the dataset has been categorized only
in the training and testing phases with a percentage of 70% and 30%, respectively (Fig. 1).
Model evaluation. To estimate the performance of analytical and ML models, the used performance metrics
are: correlation coefficient (R), root mean square error (RMSE), Nash–Sutcliffe efficiency index (NSEI), mean
absolute percentage error (MAPE), and mean absolute error (MAE). In addition, the performance index (Pi) and
over-fitting analysis (OFA) have also been done to check the fitting of the ML algorithms. The formulation of all
the performance metrics is given in Table 249,50.
Artificial neural network. The study of biological neural links served as inspiration for ANN research.
ANN algorithm is a "black box" that houses a massively parallel system with numerous processing components
that are very good at information mining. The procedure inside the box enables a careful selection of the vari-
Vol:.(1234567890)
700 700 700

600 600 600
500 500 500
Vu (kN)
Vu (kN)
Vu (kN)
400 400 400
300 300 300
200 200 200
100 100 100
0 0 0
120 140 160 180 200 220 240 260 0 100 200 300 400 500 600 16 20 24 28 32
b (mm) d (mm) fck (MPa)
(a) (b) (c)
700 700 700

600 600 600
500 500 500
Vu (kN)
Vu (kN)
Vu (kN)
400 400 400
300 300 300
200 200 200
100 100 100
0 0 0
200 300 400 500 600 700 800 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
fy (MPa) ρl (%) ρv (%)
(d) (e) (f)
700 700 700

600 600 600
500 500 500
Vu (kN)
Vu (kN)
Vu (kN)
400 400 400

300 300 300
200 200 200
100 100 100
0 0 0
200 300 400 500 600 700 50 100 150 200 250 300 350 1 2 3 4 5
fyv (MPa) s (mm) λ
(g) (h) (i)
700 700
600 600
500 500
Vu (kN)
Vu (kN)
400 400
300 300
200 200
100 100
0 0
-5 0 5 10 15 20 25 30 0 20 40 60 80 100
ηl (%) ηw (%)
(j) (k)
Figure 3. Marginal box chart of input parameters with shear capacity of CRCBs (a) b, (b) d, (c) fck, (d) fy, (e) ρl ,
(f) ρv , (g) fyv, (h) s, (i) , (j) ηl and (k) ηw.
ables and a detailed investigation of their r elationships51,52. In different ANN algorithms, a back-propagation
network (BPN) is frequently used to solve engineering problems with the gradient descent technique to reduce
errors. A typical BPN contains three layers: (a) input layer (IL), (b) hidden layer (HL), and (c) output layer (OL)
Vol.:(0123456789)
S. no. Performance Metric Formulation Description

N
(Ei −E)(Pi −P )
i=1
1 R N 2 N 2
i=1 (Ei −E ) i=1 (Pi −P )

2 MAPE 1
N
N
i=1 |Ei − Pi |

3 RMSE 1 N
i=1 (Ei − Pi )
2
N
Where, E and P are the experimental and predicted
4 MAE 1 N Ei −Pi
N i=1 Ei × 100 output sets respectively, E and P are the mean of
N experimental and predicted output sets respectively, N
5 NSEI 1− i=1
(Ei −Pi )2 is the number of points in the dataset, and m20 is the
number of values obtained from measured/predicted
N 2
i=1 (Ei −P)
6 a20-index m20 value and lies in the range of 0.8–1.2

N
N
7 RRMSE RRMSE = 1 i=1 (Ei −Pi )
2
|r| N
8 Pi RRMSE
1+R

NTraining −NTesting NTesting
9 OFA N Pi,Training + 2 N Pi,Testing
Table 2. Description of performance metrics.
as presented in Fig. 4a. The HL neurons are connected to each input neuron, which represents an individual
input parameter. Depending on the kind of operation (linear or non-linear), these neurons sum the weighted
values or apply the activation function after receiving information from the appropriate IL to produce the desired
output. The extra node known as bias is present in both the HL and OLs.
Three layers of neurons are linked together by connections known as weights. To estimate the output of an
ANN algorithm for a specific pattern, the biases and weights must be adequate. Each neuron that receives a
numerical input from the preceding layer has its relevance determined by weighting variables. The shear strength
of the CRCBs can be evaluated using Eq. (25):
N

Vu = f(H−O) Wi(H−O) Ni + B(H−O) (25)
i=1
where, f(H−O) is the OL activation function as expressed in Eq. (26), Wi(H−O) are the OL weights, Ni are the input
variables and can be obtained from Eq. (27), and B(H−O) is the output bias.
f(H−O) = purelin = f (x) = x (26)
N

Ni = f(I−H) Wi(I−H) Xi + B(I−H) (27)
i=1
where, f(I−H) is activation function in the HL as expressed in Eq. (28), Wi(I−H) is the HL weights, Xi is the nor-
malized input values, and B(I−H) is the HL biases.
2
f(I−H) = TanSig = −1 (28)
1 − e−2z
ANN has been trained from three neurons to eleven neurons. A trial and error process has been adopted
to select the optimum neuron. On the basis of the R-value and MSE value the best-selected neuron is selected
as shown in Fig. 5. The best neuron is chosen based on performance indicators. In the range of three to eleven
neurons, neuron ten has the highest R-value and the lowest MSE value. The overall evaluation of neuron ten is
acceptable. The correlation coefficient and MSE values of the training phase, testing phase, and whole phase is
shown in Fig. 5.
The parrot-colored rectangular box depicts the location of the best neuron. The formulation to predict the
shear strength of CRCBs is expressed in Eqs. (29) and (30).
  
    b  
d1  −0.8071 0.0620 −0.9229 0.2073 0.5096 −1.3520 1.9673 0.0969 −0.0353 1.2739 0.4922   −1.9573
     d   
 d2   0.4208 −0.1356 −0.3379 −0.1556 0.2532 −0.4656 0.8207 −0.7836 2.0054 −0.1553 −0.2239     −1.2184 
     fck   
 d3



 −1.9789
 −0.8337 0.5156 −0.3473 0.4251 −0.5391 −0.6441 0.7573 −1.2279 0.4585 0.5222  
  f
  0.9513 
  
 d
 4


 0.0980
 −0.0709 0.2388 0.7224 −0.7378 −0.3546 −0.3905 1.6784 −2.9510 1.0597 −1.4663 
 
 y   −1.8569 
  
     ρl   
 d5

 
 = tansig  −2.2629 −0.3706 0.9284 0.5114 0.7825 −0.1382 −1.4820 1.7856 −0.1469 −0.6652 0.5620  
 ×  ρv
  1.7187 
+ 
 d
 6


 1.7002
 −0.0002 −0.2672 1.5503 1.0243 0.1634 0.3901 0.3129 1.9112 −0.4058 0.8284  
   0.4107 
  
     fyv   
 d7   1.5318 −0.9390 0.0097 −0.3094 −0.9207 0.9510 −0.7813 0.2750 −1.0490 −0.9694 2.0092  
  s
  −1.7598 
     
 d
 8


 0.3404
 −0.2200 0.7990 −1.7213 0.6330 0.8612 −0.2832 −0.8524 −2.2531 −1.7455 1.1284  
   −1.3854 
  
       
 d9   0.2358 0.8619 0.2756 0.4233 1.1082 −1.8456 0.5682 −0.0741 2.7171 1.8584 −0.6170     1.0573 
  ηl 
d10 −0.9259 −0.4344 −0.9707 0.4044 0.0502 −0.2438 0.6536 0.0834 −0.9066 −1.6718 −0.1058 −1.9978
ηw
(29)
Vol:.(1234567890)
d
Dataset X
fck
fy 1
Tree 1 {X, θ1} Tree 2 {X, θ2} Tree 3 {X, θ3} Tree Z {X, θz}
ρl
2 …
ρv Vu
...
fyv
f1(X,θ1) f2(X,θ2) f3(X,θ3) … fz(X,θz)
s 11
λ
Output Layer Σfz(X,θz)
ηl
Hidden Layer
ηw
Input Layer
(a) (b)
A1
b A2
B1
d B2
C1 b, d, fck, fy, ρl, ρv, fyv, s, λ, ηl, Output
ηw
fck C2 reference
D1 W1 W1
fy D2 П N П
W1f
E1
ρl f
E2 Σ Σ Vu
F1
ρv Output
F2
П N П W2f
G1 W2 W2
fyv
G2
Back propagation
s H1 b, d, fck, fy, ρl, ρv, fyv, s, λ, ηl, algorithm
H2 ηw
λ I1
I2
ηl J1
J2
ηw K1
K2 Layer 1 Layer 2 Layer 3 Layer 4 Layer 5
(c)
Figure 4. Architecture of ML models (a) ANN, (b) XGBoost and (c) ANFIS Model.
−0.1885d1 −0.4275d2 +0.4295d3 +0.5789d4 −0.7536d5 +0.9335d6 −0.4282d7 +0.5503d8 +0.4771d9 +0.4468d10 −0.3566
(30)
The values of d1 to d10 can be calculated using Eq. (29).
Adaptive neuro‑fuzzy inference system. ANFIS is the name of the hybrid neuro-fuzzy network that
simulates complex systems. A fuzzy inference system (FIS), which is employed with an ANN, and Takagi–Sug-
eno rule type make up the majority of the ANFIS model. An adaptive and feed-forward network derives fuzzy
rules from inputs using an ANFIS technique. A hybrid learning approach employs the fuzzy membership func-
tion (MF) parameters and looks for connections between the inputs and outputs based on the knowledge of
expert systems. The basic architecture of the ANFIS model is presented in Fig. 4c. The ANFIS structure con-
sists of five layers, namely, the “fuzzy layer”, “product layer”, “normalized layer”, “de-fuzzy layer”, and “total OL”
(Fig. 4c). The formulation of each layer and complete description is available in the literature53–55.
Layer 1: All nodes in this layer are adaptive nodes. MFs like the Gaussian MF and generalized bell MF are
employed as node functions.
Vol.:(0123456789)
(A) - Selection of best neuron based on correlation coefficient

1.000 0.995
1.00
0.995 0.95 0.990
0.990 0.90 0.985
R-value
R-value
R-value
0.985 0.85
0.980
0.80
0.980
0.75 0.975
0.975 0.70
0.970
0.970 0.65
3 4 5 6 7 8 9 10 11 3 4 5 6 7 8 9 10 11 3 4 5 6 7 8 9 10 11
Neuron Neuron Neuron
(a) Training dataset (b) Testing dataset (c) All dataset
(B) - Selection of best neuron based on MSE

0.006
0.0045 0.016
0.0040 0.014 0.005
0.0035
0.012
0.0030 0.004
0.010
MSE
MSE
MSE
0.0025
0.008
0.0020 0.003
0.0015 0.006
0.0010 0.004 0.002
0.0005 0.002
0.0000 0.000 0.001
3 4 5 6 7 8 9 10 11 3 4 5 6 7 8 9 10 11 3 4 5 6 7 8 9 10 11
Neuron Neuron Neuron
(a) Training dataset (b) Testing dataset (c) All dataset
Location of the best neuron
Figure 5. Selection of optimum neuron.
Layer 2: Each node output in this layer displays the firing rate of a rule.
Layer 3: The normalized firing strength of each rule is represented by each node.
Layer 4: Each node in this layer is adaptive and has a node function that describes how the rules contributed
to the final output.
Layer 5: The sum of all the rules outputs is computed by a single node.
The subtractive clustering approach and the grid partitioning method are used to choose the initial fuzzy
model based on the fuzzy rules specified. The subtractive clustering approach is adopted in the development of
the ANFIS model. Locating the cluster centres of the input–output data pairs is made easier by the cluster estima-
tion approach. This in turn aids in the identification of the rules that are dispersed across the input–output space,
as each cluster centre denotes the existence of a rule. Additionally, it aids in figuring out what the underlying
premise parameters should be set to. This is crucial because, during the neural network training session, a starting
value that is very near to the ultimate value will eventually force the model to quickly converge to that v alue56.
The potentials of all the input and output data points are determined using the Euclidian distances between them
and the other data points in this clustering approach.
In the subtractive clustering approach, the squash factor, reject ratio, and accept ratio are taken as constant
with values of 1.25, 0.5, and 0.15, sequentially. The cluster centre (r) is changed from 0.9 to 0.2 value. The best
cluster centre is chosen based on performance indicators. In the range of 0.9–0.2 cluster centre, cluster centre
0.45 has the highest R-value and the lowest RMSE, MAPE, and MSE values. The overall evaluation of cluster
centre 0.45 is acceptable. In Fig. 6, the performance of all the cluster centre is shown with a number of rules (n).
The number of rules and MFs are in Figs. 7 and 8, respectively.
The shear predictions using the ANFIS model is expressed in Eq. (31).
n
i=1 Wi Yi
Vu,pred. = n (31)
i=1 Wi
The values Wi and Yi are expressed in Eqs. (32) and (33), respectively.
Vol:.(1234567890)
(A) - Selection of best cluster based on R-value r n

0.9 9
0.9995
1.0 1.00 0.85 10
0.9990
0.80 10
0.9 0.95 0.75 10
0.9985 0.90 0.70 11
0.8
R-value
R-value
R-value
0.65 12
0.7 0.85 0.60 13
0.9980
0.80 0.55 16
0.9975 0.6 0.50 18
0.75 0.45 18
0.5
0.9970 0.70 0.40 23
0.4 0.35 31
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
r r r
0.30 38
0.25 48
(a) Training dataset (b) Testing dataset (c) All dataset 0.20 52
(B) - Selection of best cluster based on RMSE r n

0.9 9
140
4.8 70 0.85 10
4.6 120 0.80 10
4.4 60 0.75 10
4.2 100 0.70 11
50
RMSE
0.65 12
RMSE
RMSE
4.0
80
3.8 40 0.60 13
3.6 60 30 0.55 16
3.4 0.50 18
40 20
3.2 0.45 18
3.0 20 10 0.40 23
2.8 0.35 31
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
r r r 0.30 38
0.25 48
(C) - Selection of best cluster based on MAPE r n

0.9 9
3.4 40 0.85 10
14
3.2 0.80 10
3.0 35
12
0.75 10
2.8 0.70 11
2.6 30
MAPE
MAPE
0.65 12
MAPE
10
2.4 0.60 13
25
2.2 0.55 16
8
2.0 20 0.50 18
1.8
15 6 0.45 18
1.6
1.4
0.40 23
10 4 0.35 31
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
r r 0.30 38
r
0.25 48
(D) - Selection of best cluster based on MAE r n

0.9 9
3.5 60
18 0.85 10
50 16
0.80 10
3.0
0.75 10
14 0.70 11
2.5 40
0.65 12
MAE
MAE
12
MAE
30 10 0.60 13
2.0 0.55 16
8
20 0.50 18
1.5 6 0.45 18
10 4 0.40 23
1.0 0.35 31
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
r r 0.30 38
r
0.25 48
Figure 6. Selection of optimum cluster center (r) based on R and RMSE values. Selection of optimum cluster
center (r) based on MAPE and MAE values.
Vol.:(0123456789)
b = -0.571 d = -0.742 fck = -1 fy = -0.660 ρl = 0.5 ρv = -0.868 fyv = -0.675 s = 0.067 λ = -0.351 ηl = -0.338 ηw = -0.895 Vu,pred. = -0.845
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
-1 1 -1 1 -1 1 -1 1 -1 1 -1 1 -1 1 -1 0.5467 -1 1 -1 1 -1 0.93
-52.71 45.15
Figure 7. Rules of the established ANFIS model.

1 b − c1 2 1 d − c2 2 1 fck − c3 2
Wi = exp − × exp − × exp −
2 σ1 2 σ2 2 σ3
2 2
1 fy − c4 1 ρl − c5 1 ρv − c6 2
× exp − × exp − × exp −
2 σ4 2 σ5 2 σ6
2 2 (32)
1 fyv − c7 1 s − c8 1 − c9 2
× exp − × exp − × exp −
2 σ7 2 σ8 2 σ9
2 2
1 ηl − c10 1 ηw − c11
× exp − × exp −
2 σ10 2 σ11
where, b , d , fck , fy , ρl , ρv , fyv , s , , ηl , and ηw are the input variables (normalized values), and σ and c are the
Gaussian MF parameters.
Yi = k1 b + k2 d + k3 fck + k4 fy + k5 ρl + k6 ρv + k7 fyv + k8 s + k9 + k10 ηl + k11 ηw + k12 (33)
The values of k1, to k12 are given in Table 3, and c and σ values of input parameters are in Table 4.
Decision tree (DT). A decision tree is a structure that resembles a flowchart and is used to illustrate a
decision-making process. It is a kind of technique for supervised learning that may be applied to both classifica-
tion and regression applications. In order to create subsets (or "leaves") that are as homogeneous as feasible with
regard to the target variable, the dataset is recursively split into subsets based on the values of the input features.
The "root" node of the tree, which represents the complete dataset, is the first node in the tree. Then, based on
a selected feature and a threshold value, the root node is divided into two or more child nodes. Recursive split-
ting occurs on each child node until a halting requirement is satisfied. For instance, the tree can be terminated
when a node reaches a predetermined threshold for data points or when all of the data points in a node belong
to the same class.
Each leaf node of the tree represents a class label (in the case of a classification problem) or a predicted
value, and each internal node of the tree represents a test on an input characteristic (in the case of a regression
problem)57. A set of choices that result in a particular conclusion are represented by the route from the root to a
leaf node. A decision tree can be used for prediction by going from the root to a leaf node and selecting the class
label or predicted value linked to that leaf node. Decision trees have a number of benefits, one of which is their
readability and comprehension due to the clear and logical representation of the decision-making process. They
may, however, be prone to overfitting if they are not appropriately trimmed or regularized.
The simplicity of understanding and visualization, ease of data pre-processing, and insensitivity to outliers
are all advantages of DT over other ML m odels58. In this study, both tenfold cross-validations along with grid
search are used to optimize the decision tree. The values of tuning hyper-parameters are auto, 10, 2, 5, and 100
for parameters of max features, max depth, min samples leaf, min samples split, and max-leaf nodes, respectively.
eXtreme gradient boosting (XGBoost). XGBoost is a highly effective and scalable ML algorithm for
tree boosting and has been widely employed in various domains to produce cutting-edge outcomes on specific
Vol:.(1234567890)
(a) (b)
(c) (d)
(e) (f)
Figure 8. MF of the selected ANFIS model (a) b, (b) d, (c) fck, (d) fy, (e) ρl , and (f) ρv , (g) fyv, (h) s, (i) , (j) ηl
and (k) ηw.
Vol.:(0123456789)
(g) (h)
(i) (j)
(k)
Figure 8. (continued)
data difficulties. The gradient boosting framework is optimized in XGBoost, which is created to be extremely
effective, adaptable, and portable59. The basic task of the XGBoost method is to optimize the value of the objec-
tive function, which consists of the regularisation term and the loss function. Although the regularisation term
serves to smooth the final learned weights to limit overfitting, the loss function, which calculates the difference
between the estimated and actual label for a given training sample, minimizes the error of the entire model. A
few XGBoost tuning settings have a significant impact on the model’s performance and training efficiency. The
learning rate, maximum depth of a tree, minimal sum of instance weight, subsample, and the number of boost-
ing iterations are the some hyper tuning parameters. The basic architecture of the XGBoost model is depicted in
Fig. 4b. The pseudocode of the XGBoost algorithm is mentioned below:
Vol:.(1234567890)
Rule no k1 k2 k3 k4 k5 k6 k7 k8 k9 k10 k11 k12

1 0.5952 0.1915 − 0.8659 − 0.2932 0.3666 − 0.0850 0.7986 1.6790 0.7859 0.0832 0.6507 0.0010
2 0.1039 0.0932 − 0.0594 − 0.0392 0.0258 − 0.1442 − 0.0871 0.1397 0.0793 0.1397 − 0.1333 − 0.1397
3 − 3.0570 − 0.5361 1.2350 3.3750 − 1.2930 2.1300 − 2.0500 − 1.7170 5.3620 − 0.0709 0.1482 0.1249
4 − 4.1170 4.1170 − 0.0549 1.2150 4.1170 2.4120 1.8980 1.5550 − 2.2170 − 0.3409 − 19.7700 − 4.1170
5 0.0422 0.0269 0.0336 1.3010 − 0.0322 − 0.0205 − 0.0766 − 0.0441 0.0263 0.0402 − 0.0037 − 0.0399
6 1.1230 0.8158 − 3.4990 0.9940 − 0.9201 0.3232 0.9489 0.7169 − 5.6660 1.1250 − 0.7337 − 1.1240
7 0.1213 0.1525 0.0149 − 0.0858 − 0.0081 0.1676 − 0.2123 0.0425 − 0.2123 − 0.1178 0.0137 − 0.2123
8 − 0.9276 − 0.2090 − 4.7220 0.9276 0.3852 − 0.3173 − 0.7954 − 0.5978 2.6620 0.3315 − 0.9276 0.9276
9 0.3918 − 0.0266 − 0.7952 − 0.3876 − 0.1016 − 0.5806 0.2635 0.4002 0.1474 − 0.3483 2.5090 0.1854
10 0.3034 0.0745 0.1419 0.1586 − 0.2488 0.1242 0.2442 0.0143 0.0033 0.4008 − 0.1295 − 0.1128
11 − 0.0269 0.1290 0.1726 0.0773 0.0142 0.1332 0.0328 0.0139 0.0868 0.1890 0.0543 − 0.1890
12 − 0.0516 − 0.0204 0.1005 0.2010 − 0.1431 0.258 − 0.0219 − 0.0040 − 0.1024 0.0119 0.0787 − 0.3611
13 0.6952 − 0.8565 − 2.0150 0.0886 1.6950 0.4855 5.0140 0.9109 1.8460 − 1.1180 0.1374 1.1200
14 0.0915 0.1188 0.1601 0.1057 − 0.0802 0.1390 0.1081 − 0.0106 0.0562 − 0.0354 0.0683 − 0.1601
15 0.2795 0.0630 − 2.3860 − 0.2795 − 0.1161 0.0956 0.2397 0.1802 2.0360 − 2.3620 0.2795 − 0.2795
16 − 0.0359 − 0.0154 − 0.3115 0.1339 − 0.0952 − 0.1161 − 0.1673 − 0.2732 0.0083 − 0.3444 − 0.0251 − 0.2414
17 0.4053 0.5089 − 0.2638 0.1986 − 0.1725 0.6489 − 0.0627 − 0.0690 − 0.1327 − 0.1244 − 0.1952 − 0.1104
18 − 0.0230 − 0.0091 − 0.2675 0.0896 − 0.0638 0.1413 0.0389 0.0937 0.0811 − 0.1120 − 0.1021 − 0.1610
Table 3. Coefficients of membership cluster plot (normalized values).
b d fck fy ρl ρv fyv s λ ƞl ƞw
MF c1 σ1 c2 σ2 c3 σ3 c4 σ4 c5 σ5 c6 σ6 c7 σ7 c8 σ8 c9 σ9 c10 σ10 c11 σ11

1 − 0.5717 0.3185 − 0.7418 0.3182 − 0.9998 0.3186 − 0.6600 0.3183 0.9999 0.3182 − 0.1813 0.3187 − 0.6753 0.3186 − 0.3532 0.1950 0.1348 0.3189 − 1.0000 0.3182 − 0.9123 0.2763
2 − 0.7143 0.3182 − 0.6479 0.3182 0.4255 0.3182 0.2808 0.3182 − 0.2354 0.3182 1.0000 0.3182 0.6239 0.3182 − 1.0000 0.2461 − 0.5892 0.3182 − 1.0000 0.3182 − 0.4126 0.3071
3 − 1.0013 0.3165 − 0.8361 0.3181 − 0.4639 0.3392 − 0.4317 0.3204 0.7878 0.3178 − 0.5412 0.3183 − 1.0010 0.3174 − 0.3755 0.2416 − 0.4842 0.3514 − 1.0001 0.3184 − 0.5561 0.2433
4 1.0000 0.3182 − 1.0000 0.3182 − 0.5282 0.3182 − 0.2951 0.3182 − 1.0000 0.3182 − 0.5857 0.3182 − 0.4610 0.3182 − 0.3778 0.2461 0.5385 0.3182 − 0.9885 0.3182 − 0.9969 0.3071
5 − 1.0000 0.3182 − 0.6479 0.3182 − 0.9198 0.3182 − 0.3349 0.3181 0.8397 0.3182 0.1049 0.3184 0.0755 0.3218 − 0.3791 0.2484 − 0.7297 0.3182 − 1.0000 0.3182 − 0.5862 0.2890
6 − 1.0000 0.3182 − 0.8356 0.3183 − 0.2665 0.2370 − 0.4302 0.3178 0.7884 0.3182 − 0.5415 0.3185 − 0.9998 0.3184 − 0.3782 0.2467 − 0.9963 0.3242 − 1.0000 0.3182 − 0.6014 0.3506
7 − 0.5714 0.3182 − 0.7183 0.3182 0.2866 0.3182 0.4039 0.3182 0.0381 0.3182 − 0.7895 0.3182 1.0000 0.3182 − 0.2000 0.2461 1.0000 0.3182 0.0846 0.3182 0.7819 0.3071
8 − 1.0000 0.3182 − 0.2254 0.3182 0.9312 0.3156 1.0000 0.3182 0.4153 0.3182 − 0.3421 0.3182 − 0.8576 0.3182 − 0.6444 0.2461 − 0.4663 0.3153 − 0.3047 0.2893 − 1.0000 0.3071
9 − 0.5716 0.3184 − 0.7418 0.3182 − 0.9998 0.3185 − 0.6600 0.3183 0.4995 0.3170 − 0.8684 0.3185 − 0.6753 0.3183 0.0677 0.2441 − 0.3522 0.3170 − 0.7842 0.3177 − 0.9672 0.3056
10 0.1428 0.3184 − 0.7418 0.3182 − 0.9134 0.3182 − 0.4089 0.3182 − 0.0750 0.3183 − 0.7632 0.3182 − 0.1738 0.3183 0.0667 0.2461 − 0.4595 0.3182 − 1.0001 0.3182 − 1.0000 0.3071
11 0.1429 0.3182 − 0.7418 0.3182 − 0.9134 0.3182 − 0.4089 0.3182 − 0.0751 0.3182 − 0.7632 0.3182 − 0.1738 0.3182 0.0667 0.2461 − 0.4595 0.3182 − 1.0000 0.3182 0.3362 0.3071
12 0.1429 0.3182 0.0563 0.3182 1.0000 0.3182 − 0.5567 0.3182 0.3965 0.3182 − 0.7105 0.3182 − 0.0997 0.3182 0.0667 0.2461 0.3514 0.3182 0.0615 0.3182 − 0.1996 0.3071
13 − 0.9889 0.3329 − 0.6506 0.3190 − 0.9217 0.3189 − 0.9809 0.3470 0.8440 0.3204 − 0.1100 0.3275 − 0.7191 0.3472 − 0.8120 0.2639 − 0.7095 0.3695 − 1.0000 0.3182 − 1.0460 0.2998
14 − 0.5714 0.3182 − 0.7418 0.3182 − 1.0000 0.3182 − 0.6601 0.3182 0.5002 0.3182 − 0.8684 0.3182 − 0.6752 0.3182 0.0667 0.2461 − 0.3514 0.3182 0.4616 0.3180 − 0.8292 0.3071
15 − 1.0000 0.3182 − 0.2254 0.3182 0.4545 0.3261 1.0000 0.3182 0.4153 0.3182 − 0.3421 0.3182 − 0.8576 0.3182 − 0.6444 0.2461 0.1841 0.3177 − 0.5307 0.3486 − 1.0000 0.3071
16 0.1429 0.3182 0.0563 0.3182 1.0000 0.3182 − 0.5567 0.3182 0.3965 0.3182 − 1.0000 0.3182 0.1453 0.3182 0.0667 0.2461 − 0.1892 0.3182 − 0.1846 0.3182 − 0.1626 0.3071
17 0.1429 0.3182 0.0563 0.3182 0.3376 0.3182 − 0.5567 0.3182 0.3965 0.3182 − 1.0000 0.3182 0.0427 0.3182 0.0667 0.2461 − 0.1892 0.3182 − 1.0000 0.3182 − 1.0000 0.3071
18 0.1429 0.3182 0.0563 0.3182 0.3376 0.3182 − 0.5567 0.3182 0.3965 0.3182 − 0.8684 0.3182 0.0427 0.3182 − 0.3778 0.2461 − 0.4595 0.3182 0.1923 0.3182 0.1440 0.3071
Table 4. Parameters of Gaussian MF (normalized values).
Vol.:(0123456789)
S. no Model R MAPE (%) MAE (kN) RMSE (kN) NS a20-index

1 Xu and Niu 0.8365 32.9710 38.8748 51.3782 0.6781 0.2286
2 Yu 0.7577 37.0414 44.6898 63.6309 0.5062 0.3071
3 Huo 0.8255 30.2997 37.1567 52.9301 0.6583 0.2929
4 Zhao and Jin 0.8099 62.9772 67.8452 86.0807 0.0964 0.1286
5 Li et al 0.7585 41.6435 51.1521 72.8369 0.3530 0.1571
6 Liu et al 0.8689 35.4832 36.5377 48.4388 0.7139 0.3500
7 Fu and Feng (GBRT)62 0.9889 – 9.5800 14.0500 – –
Table 5. Results of analytical models and existing ML model.
Start
Dataset: Training set and features
Parameters: n_estimators
alpha
subsample
learning_rate
random_state
Initialize for t = 1 to T do
Compute
Compute
The optimal value of is :
Update the current model with best suited tree:
Compute sub-trees along the T.

Estimated probability based on strong regression.
where, “ and are first and second-order gradient statistics on the loss function; is a differentiable convex
loss function that measures the difference between the prediction and the target . T is the number of leaves
in the tree; n_estimators is the number of trees in the forest; alpha is the L1 regularization term on weights,
subsample is the subsample ratio of the training instances; learning_rate is the step size shrinkage used in the
update to prevents overfitting and random_state; seed used by the random number generator”.

Compute the gradient gi and hessian hi of the loss function l yi , yi (t−1) with respect to the current model’s
(t−1)
predictions yi . Solve for the optimal value of the new tree ft (xi ) by minimizing an approximation of the
negative gradient, this step is known as the line search. This is done by computing the L(t) function which is a
combination of the gradient and hessian of the loss function and T. Update the current model by adding the new
(t−1)
tree ft (xi ) to the ensemble, with a step size of ϵ, yi (t) =
yi + ǫft (xi ) . The iteration process starts and stops when
the required criteria are met, such as a maximum number of trees or a minimum improvement in the error. In
the end, a final ensemble of decision trees is used to make predictions on new data.
Grid search, randomized search, and tenfold cross-validation are used to optimize the XGBoost hyperparam-
eters. Through a random search, the initial hyperparameters are obtained. Then, using a grid-search approach,
the resulting hyperparameters are optimized. Ten folds are randomly selected from the training dataset. Nine
folds are employed in this technique for model training, and one fold is used for performance evaluation. The
cross-validation procedure was then carried out ten times, using the validation data from each of the ten subsam-
ples exactly once each t ime60. The values of the grid and random search hyperparameters used in the XGBoost
model are 100, 10, 0.9, 0.4, and 123 for parameters of n estimators, alpha, subsample, learning rate, and random
state, respectively.
Vol:.(1234567890)
+30% +20%+10%
600
-10%
Predicted Vu (kN) 525 200
-20%
450 -30%
375 100
Error
300
0
225
150
-100
75
0
0 75 150 225 300 375 450 525 600 -25 0 25 50 75 100 125 150
Experimental Vu (kN) No. of datasets
(a) Xu and Niu
+30%+20%+10%
600 300
-10%
525 -20%
Predicted Vu (kN)
200
450 -30%
375
Error
100
300
225 0
150
-100
75
0
0 75 150 225 300 375 450 525 600 -25 0 25 50 75 100 125 150
(b) Yu
+30%+20% +10%
600 250
+10%
525 200
Predicted Vu (kN)
-20%
450 -30% 150

375 100
Error
300 50
225 0
150 -50
75 -100
0
0 75 150 225 300 375 450 525 600 -25 0 25 50 75 100 125 150
(c) Huo
Figure 9. Results of the analytical models (a) Xu and Niu, (b) Yu, (c) Huo, (d) Zhao and Jin, (e) Li et al., and (f)
Lu et al.
Vol.:(0123456789)
+30%+20%+10% 300
600 -10%
525 200
-20%
Predicted Vu (kN)
450 -30% 100
375
Error
0
300
-100
225
150 -200
75 -300
0
0 75 150 225 300 375 450 525 600 -25 0 25 50 75 100 125 150
(d) Zhao and Jin
+30%+20%+10%
600
-10% 300
525
Predicted Vu (kN)
-20%
450 -30% 200
375
Error
300 100
225
0
150
75 -100
0
0 75 150 225 300 375 450 525 600 -25 0 25 50 75 100 125 150
(e) Li et al.
150 kN
+30%+20%+10% 200
600
+10%
150 100 kN
525 -20%
Predicted Vu (kN)
100 50 kN
450 -30%
375 50
Error
300 0
225 -50
150 -100
75 -150
0
0 75 150 225 300 375 450 525 600 -25 0 25 50 75 100 125 150
(f) Lu et al.
Figure 9. (continued)
Results and discussion
The results and discussion section is categorized into four subsections. In the first subsection, the results of the
analytical models are explained with scatter plots, 2D kernel plots, and absolute error plots. The findings of the
ANN, ANFIS, DT, and XGBoost models are also discussed graphically in the second part. Additionally, the line
plot of the experimental, predicted, and error data is also utilized to make the produced models more visible.
The comparison between the analytical and ML-based models is explained in the discussion section with the
violin and Taylor plot. The influence of single parameters on the shear strength of the CRCBs is explained in
the last subsection.
Vol:.(1234567890)
S. no Model R MAPE (%) MAE (kN) RMSE (kN) NS a20-index Pi OFA

Training 0.9908 5.2482 5.7648 10.1523 0.9812 0.9489 0.0437
1 ANN Testing 0.9962 11.1211 9.5234 14.5629 0.9890 0.9524 0.0580 0.0567
All 0.9905 7.4703 7.0135 12.2962 0.9809 0.9357 0.0530
Training 0.9987 2.3935 2.0639 3.7146 0.9975 1 0.0159
2 ANFIS Testing 0.9894 12.623 11.755 18.8090 0.9741 0.8571 0.0813 0.0551
All 0.9933 5.4623 4.9713 10.7610 0.9854 0.9571 0.0463
Training 0.9905 4.8066 5.8488 12.3964 0.9720 0.9796 0.0534
3 DT Testing 0.9898 10.0982 11.2110 21.7765 0.9654 0.8810 0.0941 0.0778
All 0.9899 6.3940 7.4575 15.8062 0.9685 0.9500 0.0681
Training 0.9999 0.0542 0.1349 0.5890 0.9999 1 0.0025
4 XGBoost Testing 0.9999 0.0267 0.0291 0.0603 1 1 0.0003 0.0012
All 0.9999 0.0459 0.1031 0.4939 0.9999 1 0.0021
Table 6. Results of ANN, ANFIS, DT, and XGBoost model.
+30% +20% +10%

Training dataset
600 75
Testing dataset -10%
525 -20% 50
Predicted Vu (kN)
450 -30%
25
375
Error
300 0
225 -25
150
-50
75
-75
0
0 75 150 225 300 375 450 525 600 -25 0 25 50 75 100 125 150
(a) (b) (c)
Training dataset Testing dataset
(d)
Figure 10. Results of the ANN model (a) scatter plot, (b) 2D Kernel density plot, (c) absolute error plot, and
(d) line plot of the experimental and predicted values with errors.
Outcomes of analytical models. Six analytical models and one existing ML (Gradient Boosted Regres-
sion Trees (GBRT)) model have been used to assess the performance of the ML models. When comparing ana-
lytical models, the correlation coefficient of Lu et al.’s model is the highest and the values of the models by Xu and
Niu, Huo, Zhao and Jin, Yu, and Li et al. decreasing sequentially. However, Huo’s model has the lowest MAPE
value and this value is 14.61% lower than that of Lu et al.’s model. The other performance metrics of Lu et al.’s
confirmed the accuracy of this model compared to other analytical models. Table 5 shows the values for each
performance metric.
The scatter plot (Fig. 9 (left-side), 2D kernel density (Fig. 9 (middle-plot) and absolute error plot (Fig. 9 (right-
side) of all the analytical models are shown in Fig. 9. In Huo, Liu et al., Yu, Xu and Niu, Li et al., and Zhao and Jin
model’s 55%, 52.14%, 49.29%, 46.42%, 36.42%, and 22.89% data lie in the range of − 30 to + 30 kN, respectively.
Vol.:(0123456789)
+30% +20%+10%
600 Training dataset 100
525 75
Predicted Vu (kN)
-20%
450 -30% 50
375
Error
25
300
0
225
150 -25
75 -50
0
0 75 150 225 300 375 450 525 600 -25 0 25 50 75 100 125 150
(a) (b) (c)
(d)
Figure 11. Results of the ANFIS model (a) scatter plot, (b) 2D Kernel density plot, (c) absolute error plot, and
(d) line plot of the experimental and predicted values with errors.
The range of the error of the above mentioned analytical model is − 100 to 201.28 kN, − 99.50 to 246.13 kN,
− 71.41 to 207.89 kN, − 268.43 to 195.69 kN, − 62.42 to 273.36 kN and − 129.27 to 144.10 kN. According to the
absolute error plot (Fig. 9 (right-side)), in the model of Liu et al., approximately 80% of the dataset is inside the
50 kN error limit. Therefore, it can be inferred that the Liu et al. model performs well in comparison to other
analytical models.
Outcomes of XGBoost, DT, ANFIS, and ANN models. In all the models, the dataset is divided into
two categories: (i) Training (70%), and (ii) Testing (30%) dataset. The R-values of the ANN training and testing
dataset are 0.9908, and 0.9962, sequentially. The MAPE, RMSE, MAE, NSEI, and a20-index of the whole dataset
(ANN) are 7.4703%, 12.2962 kN, 7.0135 kN, 0.9809, and 0.9357, respectively. The overfitting value of the ANN
model is 0.0567 as shown in Table 6. In the ANFIS model, the R-value of the training and testing dataset is 0.9987
and 0.9894, respectively. The MPAE value of the ANFIS model is lower than the ANN model which is 5.4623%.
Similarly, the MAPE values of the ANFIS model is 14.58% less than the DT model. The overall MAE and RMSE
values of the ANFIS models is also less than the ANN and DT models. The ANFIS model has a higher NSEI
and an a20-index than the ANN and DT models. The overfitting values of the ANFIS and ANN models are very
close to each other. The MAPE value of the XGBoost model for the whole dataset is minimal which is 0.0459%
and R-value is approximately equal to one. The a20-index and NS index also approaches to the value of one. The
overfitting value of the XGBoost model is 0.0021, as shown in Table 6.
The scatter plot, 2D kernel plot, absolute error plot and line plot of the ANN model is presented in Fig. 10a–d,
sequentially. According to the scatter plot (Fig. 10a), only 25.72% of values directly lie over the fitting line whereas
80% of the values are inside the 10 kN absolute error limit. As per Fig. 10b,c, the error range is between − 36.39
and 61.60 kN. The line plot of the measured and predicted value with the distribution of the errors is presented
in Fig. 10d.
Figure 11 shows the scatter plot, 2D kernel plot, absolute error plot, and line plot of the ANFIS model. Accord-
ing to the scatter plot (Fig. 11a), 45.72% dataset lie over the fitting line, and 87.14% of values are inside the 10
kN absolute error limit. The range of the error between − 40.24 and 80.19 kN is shown in Fig. 11b,c. The line
plot of the phases of the train and test dataset of the predicted and experimental values is displayed in Fig. 11d.
The scatter plot, 2D kernel plot, absolute error plot, and line plot of the XGBoost model is shown in Fig. 12.
According to Fig. 12a, 13.57% of the dataset lie over the fitting line, and 82.14% of values inside the 10 kN absolute
Vol:.(1234567890)
Training dataset
+30% +20% +10% 150
600
125
525
Predicted Vu (kN)
-20%
100
450 -30%
75
375
Error
50
300
25
225
0
150
−25
75
−50
0
0 75 150 225 300 375 450 525 600 −25 0 25 50 75 100 125 150
(a) (b) (c)
(d)
Figure 12. Results of the DT model (a) scatter plot, (b) 2D Kernel density plot, (c) absolute error plot, and (d)
line plot of the predicted and experimental values with errors.
error limit. According to the 2D kernel plot (Fig. 12b) the range of the errors is between − 29.13 and 120.18 kN.
Figure 12c,d show the absolute error and line plot of the developed DT model.
Similarly, the scatter plot, 2D kernel plot, absolute error plot, and line plot of the XGBoost model is shown
in Fig. 13. According to Fig. 13a, 97.86% dataset lie over the fitting line, and 99.29% of values inside the 2.88 kN
absolute error limit. According to the 2D kernel plot (Fig. 13b) the range of the errors is between − 0.49 and 4.80
kN. Figure 13c,d show the absolute error and line plot of the developed XGBoost model.
The developed XGBoost model has greater performance and reliability when compared to ANN ANFIS, and
DT models, according to performance metrics and graphical representations.
Discussion. The results of the developed ML models have been compared with analytical models and exist-
ing ML-based models (Fu and Feng61). The R-value of the XGBoost model is 15.07%, 1.11%, 0.95%, 0.66%, and
1.01% higher than Lu et al., Fu and Feng, ANN, ANFIS, and DT models, sequentially. Similarly, the NSEI and
a20-index of the XGBoost model is 40.06%, 1.94%, 1.47%, and 3.24% and 185.71%, 6.87%, 4.48%, and 5.26%
higher than Lu et al., ANN, ANFIS and DT models, respectively. On the other hand, the MAPE, RMSE, and
MAE values of the XGBoost model is the lowest as shown in the last row of Table 7. The overfitting value of the
XGBoost model is 97.88% 97.82%, and 98.46% lower than ANN, ANFIS, and DT models, respectively.
The violin plot and multi-histogram of all the analytical and developed ML models is shown in Fig. 14. From
Fig. 14, it is clearly depicted that the accuracy of the XGBoost model is higher as compared to other models
(analytical, ANN, ANFIS and DT). The Taylor diagrams of the analytical and ML models are shown in Fig. 15a,b,
respectively. Taylor diagram is the graphical representation of the predicted values in relation to the original
data. Taylor diagram plotted between the R, RMSE, and standard deviation. In Fig. 15a, two analytical models
(Zhao and Lu et al.) crossed the reference line of the standard deviation, and Xu and Niu, Huo, and Liu et al.
models lie below the 60 kN RMSE value. On the other hand, in Fig. 15b, the XGBoost model directly lies over
the reference line of the original dataset. This ensures the reliability and precision of the XGBoost model among
all the analytical and ML models.
Feature importance. Lundberg and Lee developed a new technique to interpret black-box models called
S HAP62. The SHAP is defined as “SHapley Additive exPlanations”. SHAP method uses game theory to character-
ize how well a machine-learning model performs.
Vol.:(0123456789)
+30%+20% +10% 6
600 Training dataset
5
525 -20%
Predicted Vu (kN)
4
450 -30%
375 3
Error
300 2
225 1
150 0
75 -1
0
0 75 150 225 300 375 450 525 600 -25 0 25 50 75 100 125 150
(a) (b) (c)
(d)
Figure 13. Results of the XGBoost model (a) scatter plot, (b) 2D Kernel density plot, (c) absolute error plot,
and (d) line plot of the predicted and experimental values with errors.
S. no Model R MAPE (%) MAE (kN) RMSE (kN) NS a20-index Pi OFA

1 Liu et al 0.8689 35.4832 36.5377 48.4388 0.7139 0.3500 – –
2 Fu and Feng (GBRT)61 0.9889 – 9.5800 14.0500 – – – –
3 ANN 0.9905 7.4703 7.0135 12.2962 0.9809 0.9357 0.0530 0.0567
4 ANFIS 0.9933 5.4623 4.9713 10.7610 0.9854 0.9571 0.0463 0.0551
5 DT 0.9899 6.3940 7.4575 15.8062 0.9685 0.9500 0.0681 0.0778
6 XGBoost 0.9999 0.0459 0.1031 0.4939 0.9999 1 0.0021 0.0012
Table 7. Comparison of analytical and existing ML model with developed ML models.
It is crucial to carry out a variety of studies that are AI-based, adaptable, and capable of doing well on a variety
of data. Through assessing its reliance on physical processes, sensitivity analysis (SA) and parametric studies
help to confirm the robustness, effectiveness, and reliability of the generated MEP models. The influence of the
individual parameter is shown in Fig. 16. According to the best-fitted model, the width of the beam, stirrups
spacing, and the a/d is the most influencing factor with values of 60.86%, 21.67%, and 12.33% affect the shear
capacity. The degree of corrosion of stirrups is only a 1.59% impact on the shear strength of the CRCBs.
Conclusions
Estimating the shear capacity of the CRCBs is a very challenging issue in the civil engineering sector. To neutralize
this issue, four ML-based algorithms (ANN, ANFIS, DT, and XGBoost) have been developed. The considered
parameters that can influence the shear strength of the CRCBs are the width of the beam, the effective depth of
the beam, CS of concrete, yield strength of steel, percentage of longitudinal steel, percentage of stirrups steel,
yield strength of stirrups, stirrups spacing, a/d, corrosion degree of longitudinal steel, and corrosion degree of
stirrups. Following is a summary of the conclusions drawn from the results of the analysis:
Vol:.(1234567890)
80 80
Count
Count
5 10
300 40 40
0 0
−100 0 100 200 300 0 2 4
200 60 80
Count
Count
30
4 40 9
100 0 0
−200 0 200 0 50 100
Range
40 60
Count
Count
3 8
0 20 30
0 0
0 100 200 −50 0 50
−100 80 60
Count
Count
40
2 30
7
−200 0 0
−100 0 100 200 −50 0 50
40
Count
Count
20 1 6
−300 20
0 0
−100 0 100 200 −100 0 100
1 Xu and Niu 2 Yu 3 Huo 4 Zhao and Jin 5 Li et al.
6 Lu et al. 7 ANN 8 ANFIS 9 DT 10 XGBoost
Figure 14. Performance comparison of analytical and ML-based models with violin and multi-histogram plot.
Xu and Niu Yu Huo Zhao and Jin ANN ANFIS XGBoost Reference
Li et al. Lu et al. Reference DT
0.0 0.0
120 120
105 105
Standard deviation
Standard deviation
90 90
75 75
60 60
45 45
30 30
15 15
1.0
1.0
0 0
0 15 30 45 60 75 90 105 120 0 15 30 45 60 75 90 105 120
Standard deviation Standard deviation
(a) (b)
Figure 15. Taylor graph of (a) analytical models and (b) ANN, ANFIS, DT, and XGBoost models.
• Among analytical models, the prediction accuracy of Lu et al. is highest based on the performance metrics.
The R-value and MAE of the Liu et al. model are 0.8689 and 36.54 kN, respectively.
• A single hidden layer with ten neurons has been used in the ANN and the model shows the good accuracy
of the developed model. The R-value of the training, and testing, data is 0.9908, and 0.9962, sequentially. The
MAPE value of the whole dataset is 7.47%.
• With a cluster radius 0.45 and eighteen rules, the performance of the ANFIS model is good. The R-value of
the training and testing dataset is 0.9987 and 0.9894, respectively. The MAPE, MAE, and RMSE of the whole
dataset are 5.46%, 4.97 kN, and 10.76 kN, respectively.
• The correlation coefficient of the DT model for the whole dataset is 0.9899. In addition, the error performance
metrics of the DT model are higher than the ANFIS model.
• The correlation coefficient of the training and testing dataset of the XGBoost model is 0.9999 and 0.9999,
respectively. The MAPE, MAE, and RMSE values of the whole dataset are 0.05%, 0.10 kN, and 0.49 kN,
respectively.
Vol.:(0123456789)
dd
ss
λλ
Contributing parameters
ηƞw
w
fck
fck
ffyv
yv
ffyy
ρρll
ηƞl
l
ρρvv
bb
0 5 10 15 20 25 30 35 40 45 50 55 60 65
Relative importance (%)
Figure 16. Feature importance.
• The excellent effectiveness of the XGBoost model in calculating the shear capacity of CRCBs was also shown,
along with Taylor’s graphical representation and violin plot.
• The developed model is very flexible and robust for engineers, requiring relatively few trial experiments. As
a result, it saves more time and money throughout the CRCB strengthening process.
In addition to the experimental data gathered for this study, further research should utilize larger datasets.
A GUI that enables interactive button-based task execution is also necessary to aid users in the practical and
design interpretation of the shear capacity e stimation28. Therefore, to develop multi-dimensional validation and
improve the methodology employed in this work, the aforementioned elements should be taken into account
and dealt with in later investigations.
Data availability
All data generated or analysed during this study are included in this published article (and its supplementary
information file).
Received: 4 January 2023; Accepted: 14 February 2023
References
1. Fang, C., Lundgren, K., Plos, M. & Gylltoft, K. Bond behaviour of corroded reinforcing steel bars in concrete. Cem. Concr. Res. 36,
1931–1938. https://doi.org/10.1016/j.cemconres.2006.05.008 (2006).
2. Xu, T. & Li, J. Experimental investigations of failure modes of reinforced concrete beams without web reinforcement. Eng. Struct.
185, 47–57. https://doi.org/10.1016/j.engstruct.2019.01.102 (2019).
3. Xu, S. & Niu, D. The shear behavior of corroded simply supported reinforced concrete beam. J. Build. Struct 25, 98–104 (2004).
4. Yu, F. The test research and analysis on the shear strength of diagonal section in corroded reinforced concrete beam, 455 Master’s
thesis. Hohai University, China 456 (2005).
5. Huo, Y. Research on shear capacity of simply supported concrete beam with corroded reinforcement (Nanchang University Nanchang,
2007).
6. Zhao, Y.-X. & Jin, W.-L. Analysis on shearing capacity of concrete beams with corroded stirrups. J. Zhejiang Univ. Eng. Sci. 42, 19
(2008).
7. Shi-bin, L. & Xin, Z. Analysis for shear capacity of reinforced concrete beams with corrosion stirrups. J. Eng. Mech. 28, 60–063
(2011).
8. Higgins, C. et al. Shear capacity assessment of corrosion-damaged reinforced concrete beams (Oregon. Dept. of Transportation.
Research Unit, 2003).
9. Webster, M. P. The assessment of corrosion-damaged concrete structures, University of Birmingham, (2000).
10. Xue, X., Seki, H. & Chen, Z. in Proceedings of the Thirteenth East Asia-Pacific Conference on Structural Engineering and Construction
(EASEC-13). C-6–2 (The Thirteenth East Asia-Pacific Conference on Structural Engineering).
11. Khan, I., François, R. & Castel, A. Experimental and analytical study of corroded shear-critical reinforced concrete beams. Mater.
Struct. 47, 1467–1481 (2014).
12. Khan, N. M. et al. Application of machine learning and multivariate statistics to predict uniaxial compressive strength and static
Young’s modulus using physical properties under different thermal conditions. Sustainability 14, 9901 (2022).
13. Nazar, S. et al. Development of the new prediction models for the compressive strength of nanomodified concrete using novel
machine learning techniques. Buildings 12, 2160. https://doi.org/10.3390/buildings12122160 (2022).
14. Kovačević, M., Lozančić, S., Nyarko, E. K. & Hadzima-Nyarko, M. Application of artificial intelligence methods for predicting the
compressive strength of self-compacting concrete with class F fly ash. Materials 15, 4191 (2022).
15. Czarnecki, S., Hadzima-Nyarko, M., Chajec, A. & Sadowski, Ł. Design of a machine learning model for the precise manufacturing
of green cementitious composites modified with waste granite powder. Sci. Rep. 12, 13242. https://doi.org/10.1038/s41598-022-
17670-6 (2022).
Vol:.(1234567890)
16. Asteris, P. G., Skentou, A. D., Bardhan, A., Samui, P. & Pilakoutas, K. Predicting concrete compressive strength using hybrid
ensembling of surrogate machine learning models. Cement Concr. Res. 145, 106449. https://doi.org/10.1016/j.cemconres.2021.
106449 (2021).
17. Rathakrishnan, V., Beddu, S. & Ahmed, A. N. Predicting compressive strength of high-performance concrete with high volume
ground granulated blast-furnace slag replacement using boosting machine learning algorithms. Sci. Rep. 12, 9539. https://doi.org/
10.1038/s41598-022-12890-2 (2022).
18. Cai, R. et al. Prediction of surface chloride concentration of marine concrete using ensemble machine learning. Cement Concr.
Res. 136, 106164. https://doi.org/10.1016/j.cemconres.2020.106164 (2020).
19. Taffese, W. Z. & Espinosa-Leal, L. Prediction of chloride resistance level of concrete using machine learning for durability and
service life assessment of building structures. J. Build. Eng. 60, 105146. https://doi.org/10.1016/j.jobe.2022.105146 (2022).
20. Nguyen, T.-A. & Ly, H.-B. Estimation of the bond strength between FRP and concrete using ANFIS and hybridized ANFIS machine
learning models. J. Sci. Transp. Technol. 1(4), 36–47 (2021).
21. Kainthura, P. & Sharma, N. Hybrid machine learning approach for landslide prediction, Uttarakhand India. Sci. Rep. 12, 20101.
https://doi.org/10.1038/s41598-022-22814-9 (2022).
22. Ahmad, J. et al. Effects of waste glass and waste marble on mechanical and durability performance of concrete. Sci. Rep. 11, 21525.
https://doi.org/10.1038/s41598-021-00994-0 (2021).
23. Martínez-Álvarez, F., Troncoso, A. & Riquelme, J. C. Data science and big data in energy forecasting. Energies 11(11), 3224 (2022).
24. Amini Pishro, A. et al. Application of artificial neural networks and multiple linear regression on local bond stress equation of
UHPC and reinforcing steel bars. Sci. Rep. 11, 15061. https://doi.org/10.1038/s41598-021-94480-2 (2021).
25. Wakjira, T. G., Abushanab, A., Ebead, U. & Alnahhal, W. FAI: Fast, accurate, and intelligent approach and prediction tool for
flexural capacity of FRP-RC beams based on super-learner machine learning model. Mater. Today Commun. 33, 104461. https://
doi.org/10.1016/j.mtcomm.2022.104461 (2022).
26. Wakjira, T. G., Ebead, U. & Alam, M. S. Machine learning-based shear capacity prediction and reliability analysis of shear-critical
RC beams strengthened with inorganic composites. Case Stud. Constr. Mater. 16, e01008. https://doi.org/10.1016/j.cscm.2022.
e01008 (2022).
27. Uddin, M. N. et al. Developing machine learning model to estimate the shear capacity for RC beams with stirrups using standard
building codes. Innov. Infrastruct. Solut. 7, 227. https://doi.org/10.1007/s41062-022-00826-8 (2022).
28. Wakjira, T. G., Ibrahim, M., Ebead, U. & Alam, M. S. Explainable machine learning model and reliability analysis for flexural
capacity prediction of RC beams strengthened in flexure with FRCM. Eng. Struct. 255, 113903. https://doi.org/10.1016/j.engst
ruct.2022.113903 (2022).
29. Badra, N., Aboul Haggag, S. Y., Deifalla, A. & Salem, N. M. Development of machine learning models for reliable prediction of
the punching shear strength of FRP-reinforced concrete slabs without shear reinforcements. Measurement 201, 111723. https://
doi.org/10.1016/j.measurement.2022.111723 (2022).
30. Deifalla, A. & Salem, N. M. A machine learning model for torsion strength of externally bonded FRP-reinforced concrete beams.
Polymers 14, 1824 (2022).
31. Mohammed, H. R. M. & Ismail, S. Proposition of new computer artificial intelligence models for shear strength prediction of
reinforced concrete beams. Eng. Comput. 38, 3739–3757. https://doi.org/10.1007/s00366-021-01400-z (2022).
32. Salem, N. M. & Deifalla, A. Evaluation of the strength of slab-column connections with FRPs using machine learning algorithms.
Polymers 14, 1517 (2022).
33. Ebid, A. & Deifalla, A. Using artificial intelligence techniques to predict punching shear capacity of lightweight concrete slabs.
Materials 15, 2732 (2022).
34. Kaveh, A., Mohammad Javadi, S. & Mahdipour Moghani, R. Shear strength prediction of FRP-reinforced concrete beams using
an extreme gradient boosting framework. Period. Polytech. Civ. Eng. 66, 18–29. https://doi.org/10.3311/PPci.18901 (2022).
35. GB50010-2002. Code for design of concrete structures. China Construction Industry (2002).
36. China Academy of building Research, Design and Construction of Reinforced Concrete Structure: Compilation of Background
Data for Design Code-1985, Beijing Sanhuan Printing Plant, 1985 (in Chinese).
37. Zararis, P. D. Shear compression failure in reinforced concrete deep beams. J. Struct. Eng. 129, 544–553. https://doi.org/10.1061/
(ASCE)0733-9445(2003)129:4(544) (2003).
38. Lu, Z.-H., Li, H., Li, W., Zhao, Y.-G. & Dong, W. An empirical model for the shear strength of corroded reinforced concrete beam.
Constr. Build. Mater. 188, 1234–1248. https://doi.org/10.1016/j.conbuildmat.2018.08.123 (2018).
39. Tanabe, T., Higai, T., Umehara, H. & Niwa, J. Concrete structure 2nd edn. (Asakura Publishing Co., 2000).
40. Futaha, Jun., Yamada, K., Yokozawa, K. & Okamura, H. Re-evaluation of shear strength formula of RC beams without shear
reinforcement. J. Japan Soc. Civ. Eng. 1, 372. https://doi.org/10.2208/jscej.1986.372_167 (1986).
41. Niwa, J. Shear equation of deep beams based on analysis. In Proceedings of JCI 2nd Colloquium on Shear Analysis of RC Structures,
Tokyo (1983).
42. Rodriguez, J., Ortega, L. M. & Casal, J. Load carrying capacity of concrete structures with corroded reinforcement. Constr. Build.
Mater. 11, 239–248. https://doi.org/10.1016/S0950-0618(97)00043-3 (1997).
43. Higgins, C. & Farrow, W. C. III. Tests of reinforced concrete beams with corrosion-damaged stirrups. ACI Mater. J. 103, 133 (2006).
44. Xia, J., Jin, W.-L. & Li, L.-Y. Shear performance of reinforced concrete beams with corroded stirrups in chloride environment.
Corros. Sci. 53, 1794–1805. https://doi.org/10.1016/j.corsci.2011.01.058 (2011).
45. Imam, A. & Azad, A. K. Prediction of residual shear strength of corroded reinforced concrete beams. Int. J. Adv. Struct. Eng. 8,
307–318. https://doi.org/10.1007/s40091-016-0133-x (2016).
46. Juarez, C. A., Guevara, B., Fajardo, G. & Castro-Borges, P. Ultimate and nominal shear strength in reinforced concrete beams
deteriorated by corrosion. Eng. Struct. 33, 3189–3196. https://doi.org/10.1016/j.engstruct.2011.08.014 (2011).
47. Liu, S. The research on shear capacity of corroded rc beams, PhD Thesis, Master’s thesis, Central South University, China, (2013).
48. Singh, R. et al. Enhancing sustainability of corroded RC structures: Estimating steel-to-concrete bond strength with ANN and
SVM algorithms. Materials 15, 8295 (2022).
49. Kumar, A., Arora, H. C., Kapoor, N. R. & Kumar, K. Prognosis of compressive strength of fly-ash-based geopolymer-modified
sustainable concrete with ML algorithms. Struct. Concrete. https://doi.org/10.1002/suco.202200344.
50. Kumar, A., Arora, H. C., Kumar, K. & Garg, H. Performance prognosis of FRCM-to-concrete bond strength using ANFIS-based
fuzzy algorithm. Expert Syst. Appl. 216, 119497. https://doi.org/10.1016/j.eswa.2022.119497 (2023).
51. Liu, Q.-F. et al. Prediction of chloride diffusivity in concrete using artificial neural network: Modelling and performance evaluation.
Constr. Build. Mater. 268, 121082. https://doi.org/10.1016/j.conbuildmat.2020.121082 (2021).
52. Ebid, A. M., Deifalla, A. F. & Mahdi, H. A. Evaluating shear strength of light-weight and normal-weight concretes through artificial
intelligence. Sustainability 14, 14010 (2022).
53. Kurtgoz, Y. & Deniz, E. in Exergetic, Energetic and Environmental Dimensions (eds Ibrahim Dincer, C. Ozgur Colpan, & Onder
Kizilkan) 133–148 (Academic Press, 2018).
54. Amirkhani, S., Nasirivatan, S., Kasaeian, A. B. & Hajinezhad, A. ANN and ANFIS models to predict the performance of solar
chimney power plants. Renew. Energy 83, 597–607. https://doi.org/10.1016/j.renene.2015.04.072 (2015).
55. Kumar, K. & Saini, R. P. Adaptive neuro-fuzzy interface system based performance monitoring technique for hydropower plants.
ISH J. Hydraul. Eng. 1, 1–11. https://doi.org/10.1080/09715010.2022.2115320 (2022).
Vol.:(0123456789)
56. Buragohain, M. & Mahanta, C. A novel approach for ANFIS modelling based on full factorial design. Appl. Soft Comput. 8, 609–625.
https://doi.org/10.1016/j.asoc.2007.03.010 (2008).
57. Zhang, J., Li, J., Hu, Y. & Zhou, J. Y. The identification method of igneous rock lithology based on data mining technology. Adv.
Mater. Res. 466–467, 65–69. https://doi.org/10.4028/www.scientific.net/AMR.466-467.65 (2012).
58. Wakjira, T. G., Al-Hamrani, A., Ebead, U. & Alnahhal, W. Shear capacity prediction of FRP-RC beams using single and ensenble
ExPlainable Machine learning models. Compos. Struct. 287, 115381. https://doi.org/10.1016/j.compstruct.2022.115381 (2022).
59. Chen, T. & Guestrin, C. in Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining
(pp. 785–794).
60. Wakjira, T. G., Alam, M. S. & Ebead, U. Plastic hinge length of rectangular RC columns using ensemble machine learning model.
Eng. Struct. 244, 112808. https://doi.org/10.1016/j.engstruct.2021.112808 (2021).
61. Fu, B. & Feng, D.-C. A machine learning-based time-dependent shear strength model for corroded reinforced concrete beams. J.
Build. Eng. 36, 102118. https://doi.org/10.1016/j.jobe.2020.102118 (2021).
62. Chen, T., & Guestrin, C. Xgboost: A scalable tree boosting system. In Proceedings of the 22nd acm sigkdd international conference
on knowledge discovery and data mining (pp. 785–794). https://doi.org/10.1145/2939672.2939785 (2016).
Author contributions
A.K.: conceptualization, methodology, investigation, software, writing—original draft. H.C.A.: conceptualization,
methodology, investigation, resources, validation, formal analysis, writing—review and editing, supervision,
project administration. N.R.K.: visualization, software, formal analysis, validation, writing—review and editing.
K.K.: methodology, software, validation, formal analysis, writing—review and editing. M.H.-N.: investigation,
formal analysis, validation, software, writing—review and editing. D.R.: visualization, validation, writing—review
and editing, acquired the funding for this research. All authors contributed extensively to discussion about the
work and in reviewing the manuscript.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary Information The online version contains supplementary material available at https://doi.org/
10.1038/s41598-023-30037-9.
Correspondence and requests for materials should be addressed to A.K.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not 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://creativecommons.org/licenses/by/4.0/.
© The Author(s) 2023
Vol:.(1234567890)

Machine Learning Intelligence To Assess The Shear Capacity of Corroded Reinforced Concrete Beams

Uploaded by

Copyright:

Available Formats

Machine Learning Intelligence To Assess The Shear Capacity of Corroded Reinforced Concrete Beams

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Machine Learning Intelligence To Assess The Shear Capacity of Corroded Reinforced Concrete Beams

Uploaded by

Copyright:

Available Formats

www.nature.

OPEN Machine learning intelligence

Scientific Reports | (2023) 13:2857 | https://doi.org/10.1038/s41598-023-30037-9 1

Scientific Reports | (2023) 13:2857 | https://doi.org/10.1038/s41598-023-30037-9 2

Analytical models to estimate the shear capacity of CRCBs

Scientific Reports | (2023) 13:2857 | https://doi.org/10.1038/s41598-023-30037-9 3

bc = b − Cv1 − Cv2 (16)

h0c = h0 − Csc (17)

Scientific Reports | (2023) 13:2857 | https://doi.org/10.1038/s41598-023-30037-9 4

where, r is the width of the loading plate (87.2 mm).

Machine learning models establishment

Dataset Analytical Models (Six) L

Beam width, b (mm) R

Yield strength of steel, fy (MPa)

Best Fitted Model

Shear span-to-depth ratio, λ

alpha = 10 Feature Importance

Scientific Reports | (2023) 13:2857 | https://doi.org/10.1038/s41598-023-30037-9 5

120 140 150 200 260

130 155 190 521

16.7 17.95 24.37 27.2 32.4

300 369 420 443.1 706

0.67 1.9 2.56 2.62 2.79

0.14 0.19 0.3 0.39 0.9

275 300 373 458 560 626

80 120 170 200 305

1 1.76 2.6 3.5 4.7

0 0.1 4 17.6 97.2

26.6 80 139 594

Figure 2. Summary of statistical analysis of the CRCBs.

Scientific Reports | (2023) 13:2857 | https://doi.org/10.1038/s41598-023-30037-9 6

700 700 700

(a) (b) (c)

700 700 700

(d) (e) (f)

700 700 700

400 400 400

Scientific Reports | (2023) 13:2857 | https://doi.org/10.1038/s41598-023-30037-9 7

S. no. Performance Metric Formulation Description

6 a20-index m20 value and lies in the range of 0.8–1.2

Table 2. Description of performance metrics.

Scientific Reports | (2023) 13:2857 | https://doi.org/10.1038/s41598-023-30037-9 8

Scientific Reports | (2023) 13:2857 | https://doi.org/10.1038/s41598-023-30037-9 9

(A) - Selection of best neuron based on correlation coefficient

0.990 0.90 0.985

(a) Training dataset (b) Testing dataset (c) All dataset

(B) - Selection of best neuron based on MSE

(a) Training dataset (b) Testing dataset (c) All dataset

Location of the best neuron

Figure 5. Selection of optimum neuron.

The values Wi and Yi are expressed in Eqs. (32) and (33), respectively.

Scientific Reports | (2023) 13:2857 | https://doi.org/10.1038/s41598-023-30037-9 10

(A) - Selection of best cluster based on R-value r n

(B) - Selection of best cluster based on RMSE r n

(C) - Selection of best cluster based on MAPE r n

(D) - Selection of best cluster based on MAE r n

Scientific Reports | (2023) 13:2857 | https://doi.org/10.1038/s41598-023-30037-9 11

Figure 7. Rules of the established ANFIS model.

Scientific Reports | (2023) 13:2857 | https://doi.org/10.1038/s41598-023-30037-9 12

Table 2. Description of performance metrics.

Table 3. Coefficients of membership cluster plot (normalized values).

Table 4. Parameters of Gaussian MF (normalized values).

Table 5. Results of analytical models and existing ML model.

Table 6. Results of ANN, ANFIS, DT, and XGBoost model.

Table 7. Comparison of analytical and existing ML model with developed ML models.