Alhorani (2020) - Mathematical Models For The Optimal Design of I - and H-Shaped Crane Bridge Girders

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Asian Journal of Civil Engineering

https://doi.org/10.1007/s42107-020-00232-4

ORIGINAL PAPER

Mathematical models for the optimal design of I‑ and H‑shaped crane


bridge girders
Rana A. M. Alhorani1 

Received: 23 August 2019 / Accepted: 3 February 2020


© Springer Nature Switzerland AG 2020

Abstract
The design of crane bridge girders requires many revisions to ensure strength and stability, and limit the deflection of girders,
so design aids are needed. This research aims to develop mathematical models for the optimal design of crane bridge girders
by minimizing their weight and thus reducing the operating energy. I- and H-shaped crane girders with compact sections of
A36 steel are studied according to the American Institute of Steel Construction-Allowable Stress Design method. A code
was written to optimize girder sectional dimensions for spans of 5–50 m with crane capacities of 50–500 kN. The design
criteria are the allowable bending, shear stresses and deflections. The obtained dimensions were utilized as target values to
propose neural models for girder design using a back-propagation network method. The models were validated using two
network training performance parameters: the mean square error and regression plots. The obtained results of the mathemati-
cal models were used as input for finite-element analysis of double crane girders. The generated models should aid crane
designers in efficiently selecting the lightest girders and crane manufacturers in providing cost-effective design solutions for
a variety of crane configurations to suit the different needs of end users.

Keywords  I-crane girder section · H-crane girder section · Crane girder mathematical model · Lightest girder section ·
Crane bridge girder neural model

Introduction has led to developing of different methods for optimal design


of real-life structures. For instance, Kaveh and others have
Cranes are commonly found at industrial plants and require participated in various developments and applications of
structural steel for their construction and energy for opera- different methauristic algorithms for solving several opti-
tion. Designers usually attempt to minimize the costs of both mization problems in civil and structural engineering field
the materials and energy needed to build and operate cranes. during the last two decades (Kaveh 2017a, b; Kaveh and
Currently, there are several codes and standards that are spe- Bakhshpoori 2019).
cific to cranes (FEM 1.001 1998; AIST TR-06 2018; BS EN Many studies have been conducted on the optimal design
13001-3-1:2012+A1:2013 2012; CMAA Specification No. of beam structures in general and on the optimization of
70 2015; CMAA Specification No. 74 2015), or the design crane bridge girders. Vinot et al. (2001) solved the shape
of industrial buildings with cranes (AISC 316-89 1989; optimization problem of thin-walled structures with a beam-
AIST TR-13 2003; BS EN 1993-6:2007 2007; Fisher 2004). like dynamic behavior by determining the actual refined
Most industrial cranes are fabricated and operated geometry considering mechanical and topological design
in facilities constructed according to AISC Specifica- constraints. The beam characteristics were obtained using
tions. However, the design codes do not offer a systematic explicit equations, which are functions of the positions of
approach to efficiently selecting crane girder sections. The the nodes defining the beam cross section. A subset of these
progress in analysis and design of many engineering systems coordinates was selected as the design variables to solve for
the new target physical properties in a nonlinear optimiza-
* Rana A. M. Alhorani tion problem. Finally, an optimization algorithm based on a
ranaabdhorani@yahoo.com; r.alhorani@zuj.edu.jo sequential quadratic programming (SQP) method was imple-
mented in MATLAB to minimize the cost function with
1
Department of Civil and Infrastructure Engineering, Al- respect to the beam characteristics. Jarraya et al. (2007) used
Zaytoonah University of Jordan, Amman, Jordan

13
Vol.:(0123456789)
Asian Journal of Civil Engineering

an SQP method to optimize the thickness and shape of beam an algorithm that combined ant colony optimization with
structures having linear mechanical behavior. Bezier and a mutation-based local search to optimize the metal struc-
B-spline methods were applied to minimize the number of tural design of an overhead traveling crane. In the model,
variables. The optimization problem was analyzed consider- a continuous array-elements-code method was proposed to
ing both the volume of the structure and the mechanical con- avoid the need to round optimization design variables during
straints represented by the von Mises criteria for isotropic mixed-variable optimization and to facilitate the combina-
materials. Zuberi et al. (2008) used a generalized reduced- tion of heuristic information. Fan and Bi (2015) considered
gradient nonlinear optimization code to optimize various a sensitivity analysis of uncertain variables; including loads,
parameters associated with a welded box section girder and geometry, dimensions, and materials of crane metallic struc-
compared the results with those of a finite-element simula- tures, and constructed a reliability-based design optimiza-
tion. The size and thickness of plates, stiffeners and rein- tion (RBDO) model of an overhead traveling crane metal-
forcements were included as design parameters. The local lic structure using an advanced first-order second-moment
buckling of both web and compression flanges was consid- method to calculate the reliability indices of probabilistic
ered. The maximum allowable bending, shear stresses and constraints at each design point. An effective ant colony
permissible deflections were the design constraints used to optimization with a mutation local search was developed
compute the optimal volume of the girder. Sun et al. (2011b) to achieve the global optimal solution. Later, Fan and Zhi
proposed a modified particle swarm optimization model for (2017) proposed a RBDO method that combines the impe-
solving mixed-variable problems. The developed swarm rialist competitive algorithm with the performance measure
optimization model was later used by Sun et al. (2011a) to approach based inverse reliability strategy for the design of
find the parameters of the sectional size for a main bridge metallic structures of cranes to improve convergence speed
crane girder and minimize its self-weight. Djelosevic et al. and high computational cost of the existing RBOD model.
(2012) identified the local stress state parameters that influ- Lagrange multipliers and biologically inspired algorithms
ence the optimum design of the girder cross-sectional shape. were applied by Savković et al. (2013) and Savković et al.
The comparative analysis showed that a trapezoidal girder (2017) for the optimization of box-shaped crane girders. The
section has a considerably more favorable stress state than a reduction of the main girder mass was set as the objective
rectangular girder section based on the same carrying capac- function. Permissible stresses and strains, lateral stability
ity, depth, and thickness of vertical plates. It was concluded and dynamic stiffness criteria were considered as the con-
that the width and thickness of the top flange plate domi- straint functions. Ahmid et al. (2017) presented a mathe-
nantly influence the values of local stress and deformation, matical optimization procedure for welded I-beam cranes
but this was not the case for the bottom flange plate of the subjected to yield, buckling and deflection criteria based
girder. Tian et al. (2012) proposed a mathematical model on Crane Manufacturers Association of America specifica-
for the optimal design of overhead traveling crane box gird- tions (CMAA). Timoshenko beam theory was applied in
ers based on strength, rigidity and stability requirements combination with a genetic algorithm. The optimization
using the interior point penalty function method. Liu et al. algorithm and mathematical optimization procedure were
(2014) established a 3D finite-element model using ANSYS programmed in Mathcad and applied to welded I-beam gan-
software to perform an elastic–plastic stress analysis for the try cranes with different spans and rated loads to find the
main girder of an existing double-trolley overhead traveling optimal girder sectional dimensions (width and thickness).
crane. The arc-length algorithm and nonlinear stabilization Fan et al. (2019) developed Kriging surrogate models for the
algorithm were adopted to predict the limiting load bearing performance functions for the crane system design and then
capacity of the main girder. The strength analysis results for used the models for the reliability-based design optimization
the girder were used to conduct optimization calculations to address the computational challenges and thus enhance
considering the girder weight as the objective function, the design efficiency.
allowable stress and deflection as the state variables, and the In this paper, mathematical models are proposed to aid
thickness and height of the web plate as the design variables. designers in the selection process of optimizing the dimen-
Abid et al. (2015) developed a 3D finite-element model of a sions of I- and H-shaped crane bridge girders according to
crane girder with a 150-ton capacity and 32-m length using the design limitations imposed by the AISC 316-89 method.
ANSYS software. An optimization design model for the box A code was written in the C language to optimize the dimen-
girder was developed by changing the magnitude, shape and sions of I- and H-girders with lengths between 5 and 50 m
location of the horizontal and vertical stiffeners along the and crane capacities ranging from 50 to 500 kN. Mathemati-
girder. Then, the geometry was further optimized based on cal models were generated based on a regression analysis
the criterion of the maximum allowable bending stress by of the design outputs using the Microsoft Excel platform.
changing the thickness, height and width of the box girder The optimization results were then utilized as target values
with the smallest possible weight. Qu et al. (2015) developed to develop three artificial neural network models for I- and

13
Asian Journal of Civil Engineering

H-girders. A back-propagation algorithm was executed in A36 steel with compact sections in accordance with AISC
MATLAB to train the networks by calibrating the weights specifications. The crane loads are represented by vertical
of the input parameters and to accurately predict the output and lateral concentrated loads at the top flange level that
values. The STAAD Pro program was utilized to examine act in combination with the girder weight at the mid-span
the results of the mathematical equations proposed in this of the girder (Fig. 1). Crane bridge girders with lengths L
research. Ten girders were designed using these math- of 5–50 m and crane capacities P ranging between 50 and
ematical equations, and the design dimensions were then 500 kN are considered. Lateral loads H (%) of 10, 15, 20,
employed as inputs to analyze the finite-element models of and 25% of the vertical loads of the crane are applied.
double crane girders based on STAAD Pro. The deflections
and main stresses from both approaches were compared. Analytical models
The remainder of this paper is organized as follows: “Meth-
odology” section illustrates the analysis and design method The crane girder design variables are the web depth d, the
applied to construct the mathematical models proposed for flange width b, the flange thickness tf, and the web thick-
the optimal design of I- and H shaped crane bridge girders. ness tw. Initially, these variables are determined according
“Design constraints” section outlines the design constraints. to the bending stresses and deflection criteria. Then, the
The proposed models and discussion of the findings of this relationships among those dimensions (d, b, tf and tw) are
paper are shown in “Results and Discussion” section, followed obtained for the lightest main girder sections. Subsequently,
by concluding remarks in “Conclusion” section. the width-to-thickness ratios of the compression elements of
the girder section, (b⁄(2tf))) and (d⁄tw), are adjusted according
to AISC limitations for compact sections. The combined
Methodology stress ratio due to biaxial bending stresses is then validated
considering the allowable stresses in tension and compres-
Analysis and design method sion in accordance with AISC 316-89 method. The allow-
able compressive stress is computed based on the ((l⁄rt) ratio,
The overall performance of crane bridge girders, in terms where l is defined as the distance between cross sections
of the main girder strength, the section stability, and the braced against twisting or the lateral displacement of the
girder deflection, depends on the crane load, girder length compression flange and rt is the radius of gyration of a sec-
and cross-sectional dimensions of the girder. The design of tion comprising the compression flange plus one-third of the
crane bridge girders is performed for I- and H-members of compression web area about an axis in the plane of the web.

Fig. 1  Crane bridge girder dimensions and loading

13
Asian Journal of Civil Engineering

Then, the combined shearing stress caused by the vertical


loads and twisting moments is calculated according to the
allowable shear stress. Finally, the girder design section is
verified based on the vertical and horizontal deflection cri-
teria ∆vertical and ∆horizontal.

Artificial neural models

An artificial neural network (ANN) algorithm is a supervised


machine learning mathematical algorithm. ANNs require train-
ing with features and target data to compute the weights of
variables in the neural model. The training process aims to find
the weights of variables in the neural model to reduce the dif-
ference between the output value, known as predicted, and the
target data fed in the training. The weights are multiplied with
the features of each neuron in the network to obtain the output
value of each neuron. This process is called back-propagation,
because the values of the weights depend on the output of the
network. Which means that the model propagate from input
to output ‘forward’ to find the error and from the output to
the input ‘backward’ to find the correct values of the weights.
These two processes iterate until an error threshold value is
reached or an error of zero is reached. In this paper, artificial
neural back-propagation networks (ANBPNs) have been used
to construct mathematical models of I- and H-sections. Three
ANBPN models were developed to simplify the design process
of I- and H-shaped crane bridge girders. The models were con-
structed and trained in MATLAB to accurately predict the out-
put values (girders depth d and thickness t). To train the neural
models, girder shapes (S), girder lengths (L), and crane loads
(P and H %) were used as features; and the analytical design
results, as per “Analytical models” section, were applied as
Fig. 2  Pseudo-code of the ANBPN model for I-shaped girders
target data to predict the I- and H-sectional dimensions. The
neural models were validated by the mean square error (MSE),
as shown in Eq. (1), and regression plots. Figure 2 shows the of the matrix was normalized using Eq. (2). Finally, the nor-
pseudo-code of the ANBPN model of I- girders. malized data were divided into two matrices. The first matrix
represents an input matrix, and the second matrix represents
1 ∑( )2
N
the target matrix of the neural network. The 400 rows of this
MSE = yi − ỹi , (1)
N i=1 matrix were divided into three classes: 60% of the rows were
used for training, 20% were used for testing and 20% were
where N is the number of samples, yi is the target value and used for validation. The model was influenced by a fourth
ỹi is the predicted value. input known as bias. This neural model consisted of one hid-
den layer with 8 neurons and two output nodes (Fig. 3a).
I‑shaped crane girders xn − 𝜇
norm =
max(x)
, (2)
The ANBPN model was constructed to predict the cross-sec-
tional dimensions of I-shaped crane bridge girders. The capac- where xn is an input feature, μ is the mean of any column in
ities, lateral loads and lengths of crane bridge girders were the matrix and max ( x ) is the maximum value of any column
used as input features, and the analytical optimum dimensions in the matrix.
were used as target values to train the ANBPN model. These
values were arranged in a 400 by 5 matrix in which the first
three columns were the input values (P, H (%), L) and the last
two columns were the output values (d and t). Each column

13
Asian Journal of Civil Engineering

Fig. 3  ANBPNs for I- and H-shaped girders

H‑shaped crane girders length of the crane girder (L). The third input is the crane capac-
ity (P). The last input is the lateral load (H %), which given a
A model using the ANBPN concept was constructed to fixed value of 10% for H-shaped crane girders. Eight neurons
estimate the cross-sectional dimensions of H-shaped crane were used in the hidden layer as shown in Fig. 3c. A matrix of
bridge girders. The crane capacities and girder lengths were 500 records was leveraged as an input to the model. Another
used as input features, and the optimized girder dimen- matrix of 500 by 2 was used for the output. The model was
sions were used as target values to train the ANBPN model. trained in MATLAB. The input matrix was divided into train-
These values were arranged in a 100 by 4 matrix in which ing, testing and validation with 60%, 20% and 20% respectively.
the first two columns were the input values (P, L) and the
last two columns were the output values (d and t). As in the Finite‑element analysis
first model, the data were normalized and divided into two
matrices. The model was influenced by a third input known In this paper, STAAD Pro software was utilized to compare
as bias. The neural network consisted of one hidden layer the finite-element analysis results (stresses and deflections)
with 6 neurons and two output nodes (Fig. 3b). The gradi- and those results of the mathematical models developed for
ent descent method was applied to optimize the selected design of I- and H-shaped crane bridge girders. Ten crane
weights of the variables in the ANBPN model. bridge girders; five girders with I-shaped sections and five
girders with H-shaped sections, were designed using the pro-
I‑ and H‑shaped crane girders posed mathematical models. The loading and span details of
these girders are shown in Table 1. For the obtained girder
A model using feed-forward back-propagation neural network dimensions, the major moment of inertia and neutral axis
was constructed to calculate d and t values for I- and H-shaped values were calculated to determine the normal stress gen-
crane girders. Four input features have been utilized. The first erated in the girder. The vertical and horizontal deflection
feature is the shape of the crane girder (S) and it is given a value values were computed using the developed mathematical
of 1 for I-shape and 0 for the H-shape. The second input is the models of deflection. Then, the design dimensions of these

13
Asian Journal of Civil Engineering

Table 1  Loading and span P (kN) H (kN) L (m) where fb is the computed bending stress and Fb is the allow-
details of girders used in finite- able bending stress. The subscripts x and y indicate the axis
element analysis 100 10 20 on which a particular bending stress or design property is
100 10 27 considered. Chapter F in the AISC 316-89 manual explains
150 15 27 the required conditions for applying Eq. (7).
200 20 33
200 20 40 Allowable shearing stress ( Fv)

girders were used in the STAAD Pro finite-element models The allowable shearing stress, which is influenced by the
of double crane girders. girder depth and web thickness, is considered, as shown in
Eq. (8):

Design constraints Fv = 0.4Fy . (8)

The constraints used in the design of I- and H-shaped crane


bridge girders were as follows. Results and discussion
Local buckling Analytical models

Equations (3) and (4) provide the applicable limitations regard- The optimal design dimensions for I- and H-crane bridge gird-
ing the width–thickness ratios of the compression elements of ers are shown in Figs. 4a, b and 5a, b. Mathematical mod-
the compact sections in accordance with AISC 316-89 method. els are generated from these charts using regression analy-
Flanges of I- and H-shaped rolled or welded sections: sis via the MS Excel platform to determine the crane girder
√ dimensions (d and t). The developed models are provided
b E as functions of the girder length for different crane capaci-
≤ 0.38 . (3)
2tf Fy ties and lateral load percentages, as shown in Tables 2, 3 and
4. Tables 2 and 3 show the depth and thickness models of
Webs of I- and H-shaped sections: I-section girders, respectively. Table 4 provides the depth
√ and thickness models of H-section crane bridge girders. The
d
≤ 3.76
E
, (4) mathematical models used to calculate the vertical deflection
tw Fy of the optimized girders are shown in Table 4 for H-section
girders and Table 5 for I-section girders. These vertical deflec-
where E is the modulus of elasticity of steel and Fy is the tion models are useful for assessing the camber requirements
specified minimum yield stress of steel. established by the manufacturer. The horizontal deflection
models are shown in Table 6 for I-shaped girders and Table 7
Deflection criteria for H-shaped girders.
One of the findings of this research is that the lightest I-
Equations (5) and (6) show the applicable criteria for the ver- girder is achieved when the flange width equals half of the web
tical and horizontal deflections of crane girders, respectively. depth and both have the same thickness t. The thickness of the
lightest I-section varies in the range from d/12 to d/44. Addi-
Vertical deflection criterion: tionally, the lightest H-shaped girder is obtained when both the
L flange and web of the section have the same thickness t, which
Δvertical = . (5) varies from d/12 to d/20. This finding suggests that the H-girder
800
dimensions are not affected by varying the lateral load.
Horizontal deflection criterion: Comparing the I- and H-crane girders shows that eco-
nomical design solutions involve selecting H-sections when
L
Δhorizontal = . (6) crane spans are greater than 25 m with design capacities
500
between 50 and 100  kN. The H-section also represents
an optimal design solution for a crane girder of length
Combined bending stress ratio greater than 30 m for loads between 150 and 300 kN. For
spans greater than 35 m and crane capacities between 350
fbx fby
+ ≤ 1.0, (7) and 500 kN, H-sections should be selected. If the design
Fbx Fby

13
Asian Journal of Civil Engineering

Fig. 4  Design charts for I-section crane bridge girders

Fig. 5  Design charts for H-section crane bridge girders

Table 2  Web depth of I-section crane bridge girders: d = AH + BH L + CH L2 (cm)


P (kN) 50 100 150 200 250 300 350 400 450 500

A10 23.767 28.717 32.517 34.767 36.4 37.467 39.8 42.183 44.017 46.117
B10 2.6948 3.1574 3.4574 3.7209 3.9794 4.2006 4.2976 4.3726 4.4444 4.5205
C10 − 0.0191 − 0.022 − 0.0238 − 0.0258 − 0.0279 − 0.0297 − 0.0297 − 0.0298 − 0.0295 − 0.0298
A15 25.95 32.033 35.433 36.95 39.8 42.95 44.867 47.417 49.833 51.917
B15 2.9141 3.3652 3.7512 4.1129 4.2976 4.3923 4.5491 4.6256 4.6924 4.8065
C15 − 0.0205 − 0.0227 − 0.0258 − 0.0289 − 0.0297 − 0.0295 − 0.0303 − 0.0303 − 0.03 − 0.0311
A20 28.45 33.733 36.7 40.617 44.017 46.933 49.833 51.7 53.7 56.35
B20 3.0411 3.6212 4.0767 4.2953 4.4444 4.6012 4.6924 4.8597 5.0039 5.0444
C20 − 0.0208 − 0.0248 − 0.0288 − 0.0295 − 0.0295 − 0.0303 − 0.03 − 0.0312 − 0.0324 − 0.032
A25 29.75 35.217 39.017 43.867 47.683 50.283 53.167 55.35 58.567 59.867
B25 3.2141 3.8044 4.2444 4.4048 4.5977 4.7929 4.9221 5.0444 5.0864 5.2594
C25 − 0.0223 − 0.0259 − 0.0295 − 0.0291 − 0.0302 − 0.0314 − 0.0318 − 0.032 − 0.0312 − 0.0327

13
Asian Journal of Civil Engineering

Table 3  Thickness of I-section crane bridge girders: t = AH + BH L + CH L2 + DH L3 (mm)


P (kN) 50 100 150 200 250 300 350 400 450 500

A10 3.8333 6.3667 8.3 12.433 13.2 15.3 17 16.8 20.333 20.133
B10 1.206 1.1968 1.127 0.6227 0.6363 0.6011 0.505 0.5709 0.1845 0.3296
C10 − 0.0232 − 0.0181 − 0.0127 0.0088 0.0112 0.0114 0.016 0.0158 0.0299 0.0246
D10 0.0007 0.0005 0.0004 0.0001 0 0 0 0 − 0.0002 − 0.0001
A15 4.8667 7.2 11.667 14.81 15.067 18.333 18.567 19.567 21.833 21.333
B15 0.9341 0.927 0.4949 0.2142 0.3262 − 0.0354 0.0983 0.0849 − 0.0686 0.1159
C15 − 0.0146 − 0.0113 0.0082 0.0207 0.0179 0.0309 0.0263 0.0277 0.032 0.0245
D15 0.0006 0.0004 0.0001 0 0 − 0.0002 − 0.0001 − 0.0002 − 0.0002 − 0.0001
A20 7.4 7.8 11.9 13.767 15.167 15.733 16.333 15.567 16.967 18.367
B20 0.5257 0.9261 0.628 0.574 0.5096 0.5871 0.7205 0.9745 0.876 0.8495
C20 − 0.0014 − 0.0155 − 0.0033 − 0.001 0.0033 0.0013 − 0.0045 − 0.016 − 0.0101 − 0.0105
D20 0.0004 0.0005 0.0003 0.0002 0.0002 0.0002 0.0002 0.0004 0.0003 0.0003
A25 6.6667 9.3333 10.767 11.8 12.133 12.6 12.967 13.3 13.867 14.5
B25 0.7988 0.8824 0.9999 1.0968 1.2129 1.308 1.4201 1.4558 1.6107 1.6634
C25 − 0.0173 − 0.018 − 0.0229 − 0.0243 − 0.0276 − 0.0309 − 0.0344 − 0.0319 − 0.0398 − 0.0408
D25 0.0006 0.0005 0.0005 0.0005 0.0005 0.0005 0.0006 0.0005 0.0006 0.0006

Table 4  Web depth, thickness P d = A + BL + CL2 t = A + BL + CL2 + DL3 Δvertical = A + BL + CL2


and vertical deflection of (kN) (cm) (mm) (mm)
H-section crane bridge girders
A B C A B C D A B C

50 16.283 1.8868 − 0.0138 4.3 1.7923 − 0.0588 0.0009 12.505 − 1.3829 0.1323
100 19.883 2.1729 − 0.015 5.8667 2.0213 − 0.0569 0.0007 10.797 − 0.9608 0.1031
150 21.833 2.4185 − 0.0167 6.7667 2.0536 − 0.0495 0.0006 10.475 − 0.8063 0.0901
200 24.2 2.5606 − 0.0176 8.3333 1.8916 − 0.0346 0.0003 10.478 − 0.7266 0.0822
250 25.567 2.71 − 0.0185 9.3667 1.9371 − 0.0336 0.0003 9.7433 − 0.5739 0.0746
300 25.867 2.9079 − 0.0206 10.733 1.8242 − 0.0258 0.0002 8.8133 − 0.4187 0.068
350 27.7 2.9506 − 0.0203 11.533 1.8552 − 0.0262 0.0002 8.1467 − 0.3024 0.063
400 27.933 3.1042 − 0.0218 12.033 1.9749 − 0.031 0.0003 7.6417 − 0.21 0.0589
450 28.633 3.2161 − 0.0227 12.533 1.8574 − 0.0246 0.0002 6.96 − 0.1145 0.0553
500 29.167 3.2979 − 0.0233 15.6 1.762 − 0.0214 0.0001 6.3783 − 0.0348 0.0522

Table 5  Vertical deflection of P (kN) 50 100 150 200 250 300 350 400 450 500
I-section crane bridge girders:
Δvertical = AH + BH L + CH L2 A10 10.685 7.3817 5.5567 4.8667 4.2233 3.33 2.8583 2.6067 2.0517 1.5533
(mm)
B10 − 0.9552 − 0.3901 − 0.1018 0.0446 0.149 0.2509 0.3133 0.3682 0.4354 0.492
C10 0.0732 0.0505 0.0399 0.0339 0.0298 0.0266 0.0243 0.0223 0.0203 0.0187
A15 9.06 6.2083 4.7533 3.5767 3.1367 2.155 1.7567 1.3183 0.66 0.2717
B15 − 0.826 − 0.3417 − 0.0985 0.067 0.1448 0.2783 0.3289 0.3754 0.4553 0.5056
C15 0.0631 0.0435 0.0344 0.0287 0.0254 0.0216 0.0197 0.0181 0.016 0.0146
A20 7.7283 4.795 3.475 2.325 1.765 1.4083 0.8833 0.4383 0.2367 − 0.0367
B20 − 0.7403 − 0.2714 − 0.0638 0.086 0.173 0.221 0.2798 0.3367 0.3592 0.3874
C20 0.0566 0.0384 0.0304 0.0253 0.0221 0.0201 0.0181 0.0164 0.0154 0.0145
A25 6.48 3.775 2.5717 2.0017 1.3367 1.0833 0.8417 0.6733 0.4133 0.325
B25 − 0.6562 − 0.2335 − 0.0606 0.0255 0.1147 0.1467 0.1867 0.2009 0.2282 0.2383
C25 0.0515 0.0349 0.028 0.0243 0.0211 0.0195 0.0179 0.017 0.016 0.0154

13
Asian Journal of Civil Engineering

Table 6  Horizontal deflection of I-section crane bridge girders: Δhorizontal = AH + BH L + CH L2 (mm)


P (kN) 50 100 150 200 250 300 350 400 450 500

A10 0.685 − 0.7533 − 1.645 − 1.465 − 1.5917 − 2.3083 − 2.7317 − 2.52 − 3.1933 − 3.6033
B10 1.5331 1.7029 1.7917 1.795 1.7944 1.835 1.8545 1.8461 1.9025 1.9251
C10 − 0.0218 − 0.0216 − 0.0211 − 0.0199 − 0.0189 − 0.0186 − 0.0181 − 0.0175 − 0.0179 − 0.0179
A15 0.2817 − 1.4067 − 2.335 − 3.2333 − 3.3067 − 5.02 − 5.1467 − 5.58 − 6.445 − 7.3367
B15 2.0032 2.1851 2.2894 2.3615 2.3527 2.5261 2.5127 2.5165 2.6047 2.6757
C15 − 0.0286 − 0.0275 − 0.027 − 0.0263 − 0.025 − 0.027 − 0.0257 − 0.0249 − 0.0259 − 0.0263
A20 − 1.285 − 4.0517 − 5.1167 − 6.6017 − 7.17 − 7.3933 − 8.1583 − 9.1433 − 9.115 − 9.525
B20 2.4418 2.6983 2.7773 2.8776 2.9097 2.8944 2.9204 2.9722 2.963 2.9649
C20 − 0.0346 − 0.0339 − 0.0324 − 0.0317 − 0.0306 − 0.029 − 0.0284 − 0.0239 − 0.0272 − 0.0264
A25 − 4.1633 − 6.9467 − 8.3017 − 8.53 − 9.255 − 8.9217 − 9.235 − 8.8683 − 9.365 − 9.555
B25 2.8902 3.1146 3.1586 3.1101 3.1333 3.0399 3.0427 2.9844 2.948 2.9234
C25 − 0.0405 − 0.0387 − 0.0355 − 0.0321 − 0.0307 − 0.0273 − 0.0264 − 0.0284 − 0.0226 − 0.0213

Table 7  Horizontal deflection of H-section crane bridge girders: Δhorizontal = AH + BH L + CH L2 (mm)


P (kN) 50 100 150 200 250 300 350 400 450 500

A10 − 2.055 − 1.1983 − 0.5417 − 0.0617 0.0283 0.0283 0.0283 0.0283 − 0.0617 − 0.1517
B10 0.7392 0.6019 0.5098 0.4473 0.4355 0.4355 0.4355 0.4355 0.4416 0.4477
C10 − 0.0081 − 0.004 − 0.0017 − 0.0003 0 0 0 0 0 − 0.0002
A15 − 3.0917 − 1.8067 − 0.8533 − 0.1333 0.0467 − 0.1333 − 0.2233 − 0.3133 − 0.4033 − 0.4033
B15 1.1089 0.9029 0.7691 0.6753 0.6518 0.664 0.6701 0.6762 0.6823 0.6823
C15 − 0.0121 − 0.0061 − 0.0027 − 0.0005 0 − 0.0001 − 0.0002 − 0.0003 − 0.0004 − 0.0004
A20 − 4.1533 − 2.44 − 1.16 − 0.5 − 0.44 − 0.53 − 0.71 − 0.8 − 0.89 − 0.98
B20 1.4789 1.2043 1.0248 0.9219 0.9028 0.9089 0.9211 0.9272 0.9333 0.9394
C20 − 0.0161 − 0.0081 − 0.0035 − 0.001 − 0.0004 − 0.0005 − 0.0007 − 0.0008 − 0.0009 − 0.001
A25 − 5.1967 − 3.055 − 1.805 − 0.995 − 0.875 − 1.055 − 1.145 − 1.415 − 1.505 − 1.745
B25 1.8506 1.5073 1.3058 1.1777 1.1508 1.163 1.169 1.1873 1.1934 1.207
C25 − 0.0202 − 0.0101 − 0.0048 − 0.0016 − 0.0009 − 0.001 − 0.0011 − 0.0014 − 0.0015 − 0.0017

conditions do not meet the requirements for H-shaped gird- load deflection. Alternatively, the deflection due to the
ers, the designer may opt for I-shaped girders. girder weight is greater than the live load deflection when
In most cases, the vertical deflection criterion governs the crane beam length is greater than 40 m. The latter
the crane girder design. However, the combined stress finding may help designers develop girder camber profiles
ratio controls the girder design for cranes with short according to the recommendations of crane manufacturers.
lengths. For instance, the design of I-section girders with
lengths ranging between 5 and 15 m and H-section girders Artificial neural models
with lengths ranging from 5 to 10 m are governed by the
combined stress ratio criterion. I‑shaped crane girders
Furthermore, an analysis of the optimal design results
shows that the girder deflection due to the girder weight The ANBPN model of I-shaped crane girders and the
is less than that due to the live load when the length is weights of the relevant variables are given in Eqs. (9) and
less than 20 m and load varies between 50 and 500 kN. In (10), respectively. The MSE values and regression analysis
addition, the computed self-weight deflection is less than for the model outputs are shown in Figs. 6a and 7a. The
the live load deflection for cranes with spans of 25 m or crane girder dimensions are predicted by the developed
less and loads between 100 and 500 kN. For cranes with ANBPN model with an MSE of 0.000048 and an R value
spans of 30 m or less and capacities from 150 to 500 kN, of 0.99944 for training, validation and testing.
the computed self-weight deflection is less than the live

13
Asian Journal of Civil Engineering

Fig. 6  Mean square error of the ANBPN models

Fig. 7  Regression values for the ANBPN models

where
1
d= gi = ai1 + ai2 ∗ L + ai3 ∗ P + ai4 H i = 1, 2, 3, 4, 5, 6, 7, 8
1 + e(−h1 )
1 ∑2
∑∑
2 9
t= , hr = br1 + gi−1 ∗ bri r = 1, 2
1 + e(−h2 ) r=1 r=1 i=2
(9)

13
Asian Journal of Civil Engineering

⎡ 1.7607 −1.2304 0.15439 0.82957 ⎤


⎢ 1.2486 0.22343 0.012471 0.055505 ⎥
⎢ ⎥
⎢ −2.5869 −6.7141 −8.7885 5.187 ⎥
⎢ 2.6161 6.8557 8.9718 −5.2546 ⎥
[a] = ⎢ ⎥
⎢ 1.6518 −1.2376 0.13608 0.8576 ⎥ (10)
⎢ −3.5869 −4.7141 8.7885 3.187 ⎥
⎢ 2.6161 6.8557 8.9718 −5.2546 ⎥
⎢ ⎥
⎣ 3.21 −1.9677 −5.0544 6.1869 ⎦
� �
−14.6129 9.2007 24.1227 −6.2293 −6.1054 −6.9467 −1.1578 −5.5036 2.7668
[b] = ,
−13.2433 −32.1033 16.3867 3.232 3.1677 29.0065 −0.17121 1.3155 −1.7353

1
where h1 and h2 is the hidden layer value for d and t, respec- d=
tively, [a] represents the weight matrix of the input layer and 1 + e(−h1 )
[b] represents the weight matrix of the hidden layer. 1
t= ,
1 + e(−h2 )
H‑shaped crane girders
where
The ANBPN model of H-shaped crane girders and the gi = ai1 + ai2 ∗ L + ai3 ∗ P i = 1, 2, 3, 4, 5, 6
weights of the input data and hidden layer are shown in ∑2
∑∑
2 7
(11)
Eqs. (11) and (12), respectively: hr = br1 + gi−1 ∗ bri r = 1, 2
r=1 r=1 i=2

⎡ −12.6291 6.6913 −8.6024 ⎤


⎢ −7.3866 3.2282 −4.8613 ⎥
⎢ ⎥
7.9465 −4.2787 −2.3863 ⎥
[a] = ⎢
⎢ −0.21963 0.83954 0.57061 ⎥
⎢ −1.9677 (12)
−5.0544 6.1869 ⎥
⎢ ⎥
⎣ 6.719 4.9733 0.40293 ⎦
� �
0.21827 0.35789 −1.1578 −5.5036 2.7668 −0.076917 3.4412
[b] = ,
−0.17121 1.3155 −1.7353 −5.2414 2.7052 −0.090186 3.5449

Fig. 8  Comparison of calculated and predicted d and t values for H-shaped crane bridge girders

13
Asian Journal of Civil Engineering

where h1 and h2 is the hidden layer value for d and t, respec- where h1 and h2 is the hidden layer value for d and t, respec-
tively, [a] represents the weight matrix of the input layer and tively, [a] represents the weight matrix of the input layer and
[b] represents the weight matrix of the hidden layer. [b] represents the weight matrix of the hidden layer.
The proposed ANBPN model estimates H-girder dimen-
sions with an MSE value of 0.000138 and an R value of Finite‑element models
0.99832 (Figs. 6b and 7b).
A comparison of the calculated and predicted values for The STAAD Pro program was used to examine the proposed
the d and t dimensions of H-shaped crane bridge girders mathematical equations for the design of I- and H-shaped
shows very good agreement (Fig. 8a, b). crane bridge girders. Tables 8 and 9 show the dimensions
and analysis results for selected I- and H-shaped crane
I‑ and H‑shaped crane girders bridge girders, respectively. A comparison of the math-
ematical models and STAAD Pro analysis showed the
The ANBPN model of I- or H-shaped crane girders and extent of matching for the model results. Small variances
the weights of the input data and hidden layer are shown in in the stresses and vertical deflections of girders were
Eqs. (13) and (14), respectively. The proposed ANBPN model found because of small differences in the section dimen-
estimates I- and H-girder dimensions with an MSE value of sions applied in the two methods. The horizontal deflec-
0.004 and an R value of 0.92 (Fig. 9a, b). tions of single girders, as determined by the mathematical
1 models, were more than double those values for the double
d= girders, as determined by STAAD Pro. This difference was
1 + e(−h1 ) occurred due to the larger lateral resistance of the double
1
t= , girders. Figure 10 a–c show input data of the finite-element
1 + e(−h2 ) model in STAAD Pro for a girder with I-shaped section of
where length L = 20 m and loads of P = 100 kN and H = 10 kN and
Fig. 10d shows the result global stress distribution ( fbx ) in
gi = ai1 + ai2 ∗ S + ai3 ∗ L + ai4 ∗ P + ai5 the girder.
∗H i = 1, 2, 3, 4, 5, 6, 7, 8

2

2

9
hr = br1 + gi−1 ∗ bri r = 1, 2
r=1 r=1 i=2 (13)

⎡ −2.7042 0.72558 −0.73062 −1.9661 0.75332 ⎤


⎢ −0.49715 1.1969 0.47772 0.072767 0.0074412 ⎥
⎢ ⎥
⎢ 1.6883 −2.2127 −0.8579 −0.089953 −0.13563 ⎥
⎢ −0.21609 −1.3623 0.73827 −0.19406 −0.13811 ⎥
[a] = ⎢
⎢ 3.9089 5.0026 0.10069 0.11243 −0.57448 ⎥⎥
(14)
⎢ 1.2437 −2.8771 −0.9494 −1.2518 0.51799 ⎥
⎢ −3.5483 −0.15211 0.93345 0.10508 −1.1875 ⎥
⎢ ⎥
⎣ 2.3933 −3.5485 −0.053023 0.04929 −0.3337 ⎦
� �
−16.7993 −0.26837 9.1421 3.1841 −1.1302 7.4458 0.49786 −0.61775 −4.8826
[b] = ,
6.4628 0.15197 2.2673 0.48788 2.2008 −2.0598 −0.13299 5.8957 −2.4633

13
Asian Journal of Civil Engineering

Conclusions

The design process of crane bridge girders usually involves


verifying the applicable stress, stability, and deformation
limitations for the selected cross-sectional dimensions. The
objective of this research was to develop mathematical mod-
els for the optimal design of I- and H-shaped crane bridge
girders. In a crane design environment, these models save
time and reduce costs. The girder weight is chosen as the
optimization criterion, and structural safety and operational
requirements are fulfilled. The mathematical models devel-
oped are suitable for most industrial applications. The results
obtained in this paper show that the proposed ANBPN can
be used as a design tool and provide good predictions for
crane bridge girders. The design of I- and H-shaped crane
bridge girders was investigated using the proposed math-
ematical equations. The design results were used to analyze
finite-element models constructed in STAAD Pro software
for double crane bridge girders to verify the agreement
between the mathematical equation results and the finite-
element analysis results for the studied girders. The deflec-
tions and stresses generated in crane girders in both applica-
tions were found to be comparable. The third ANBPN model
considered the shape of the girder section as an input feature.
A regression value of more than 93% has been recorded for
the model. This allowed constructing a single model for both
design sections. For future work, the model can be trained
to predict the values of other crane shapes. Moreover, other
unsupervised machine language algorithms could be lever-
aged to reduce the training complicity.

Fig. 9  Performance parameters of the ANBPN model for I-and


H-shaped crane girders

13

Table 8  Comparing the mathematical model and finite-element analysis results for I-section crane bridge girders
Mathematical model analysis
fbx

13
P (kN) H (kN) L (m) b (cm) d (cm) t (mm) ∆Vertical (mm) ∆Horizontal (mm)
(MPa)

100 10 20 41.54 83.07 27.06 55.12 19.78 24.66


100 10 27 49 97.93 35.33 52.60 33.66 29.48
150 15 27 54.26 108.52 37.34 55.14 31.9 31.35
200 20 33 64.73 129.46 46.16 57.65 43.26 36.1
200 20 40 71.16 142.32 57.82 59.74 60.89 38.5
Finite-element analysis
P (kN) H (kN) L (m) b (cm) h (cm) t (mm) fbx ∆Vertical (mm) ∆Horizontal (mm)
(MPa)

100 10 20 42 84 27 55.5 24.52 12.49


100 10 27 49 98 35.33 49.9 36.99 12.33
150 15 27 54 108 37.34 49.5 34.07 13.67
200 20 33 65 130 46.16 49.2 45.18 16.33
200 20 40 70 142 57.8 47.8 56.14 17.77

Table 9  Comparing the mathematical model and finite-element analysis results for H-section crane bridge girders
Mathematical model analysis
P (kN) H (kN) L (m) b (cm) d (cm) t (mm) fbx ∆Vertical (mm) ∆Horizontal (mm)
(MPa)

100 10 20 57.34 57.34 29.13 63.01 32.82 9.24


100 10 27 67.62 67.62 32.74 66.53 60.02 12.14
150 15 27 74.96 74.96 37.94 65.94 54.39 11.98
200 20 33 89.53 89.53 43.86 71.65 76.02 14.37
200 20 40 98.46 98.46 47.84 78.77 112.93 17.35
Finite-element analysis
P (kN) H (kN) L (m) b (cm) h (cm) t (mm) fbx ∆Vertical (mm) ∆Horizontal (mm)
(MPa)

100 10 20 58 58 29.13 64.7 35.40 4.05


100 10 27 68 68 32.74 69.2 63.23 5.60
150 15 27 75 75 37.94 67.7 58.74 5.87
200 20 33 90 90 43.86 67.8 79.37 8.08
200 20 40 99 99 47.84 71.6 116.77 10.12
Asian Journal of Civil Engineering
Asian Journal of Civil Engineering

Fig. 10  Finite-element model in STAAD Pro of the I-shaped crane bridge girder: L = 20 m, P = 100 kN and H = 10kN

13
Asian Journal of Civil Engineering

Compliance with ethical standards  FEM 1.001 Rules for the Design of Hoisting Appliances. (1998).
European materials handling federation (FEM).
Fisher, J. M. (2004). Steel design guide series 7: Industrial build-
Conflict of interest  The author declares that she has no conflict of in-
ings-roofs to anchor rods. Milwaukee, WI: American Institute
terest.
Steel Construction Inc.
Jarraya, A., Dammak, F., Abid, S., et al. (2007). Shape and thick-
ness optimization performance of a beam structure by sequential
References quadratic programming method. Journal of Failure Analysis and
Prevention (JFAP), 7(1), 50–55.
Zuberi R. H., Kai, L., & Zhengxing, Z. (2008). Design optimization
Abid, M., Akmal, M., & Wajid, H. (2015). Design optimization of
of EOT crane bridge. In EngOpt 2008: International confer-
box type girder of an overhead crane. Iranian Journal of Sci-
ence on engineering optimization. EngOpt 2008, Rio de Janeiro,
ence and Technology, Transactions of Mechanical Engineering,
Brazil (pp. 187–196).
39(M1), 101–112.
Kaveh, A. (2017a). Advances in metaheuristic algorithms for optimal
Ahmid A., Le V., & Dao, T. (2017). Optimization procedure for
design of structures (2nd ed.). Cham: Springer.
an I-beam crane subjected to yield and buckling criteria. In
Kaveh, A. (2017b). Applications of metaheuristic optimization algo-
2017 World congress on advances in structural engineering and
rithms in civil engineering. Cham: Springer.
mechanics (ASEM17) (pp. 1–12). Ilsan, Seoul, Korea: ASEM.
Kaveh, A., & Bakhshpoori, T. (2019). Metaheuristics: Outlines,
American Institute of Steel Construction, Chicago, IL. (1989). AISC
MATLAB codes and examples. Cham: Springer.
manual of steel construction: Allowable stress design (AISC
Liu, P. F., Xing, L. J., Liu, Y. L., et al. (2014). Strength analysis
316-89), 9th edn.
and optimal design for main girder of double-trolley overhead
Association of Iron and Steel Technology, Warrendale, PA. (2003).
traveling crane using finite element method. Journal of Failure
Guide for the design and construction of mill buildings (AIST
Analysis and Prevention (JFAP), 14(1), 76–86.
TR-13).
Qu, X., Xu, G., Fan, X., et al. (2015). Intelligent optimization meth-
Association of Iron and Steel Technology, Warrendale, PA. (2018).
ods for the design of an overhead travelling crane. Chinese
AIST Specification for electrical overhead traveling cranes for
Journal of Mechanical Engineering (CJME), 28(1), 187–196.
steel mill service (AIST TR- 06).
Savković, M. M., Bulatović, R. R., Gašić, M. M., et al. (2017). Opti-
BS EN 1993-6:2007. (2007). Eurocode 3: Design of steel structures-
mization of the box section of the main girder of the single-
crane supporting structures. London, UK: BSI.
girder bridge crane by applying biologically inspired algo-
BS EN 13001-3-1:2012+A1:2013. (2012). Cranes: General design-
rithms. Engineering Structures, 148, 452–465.
limit states and proof competence of steel structure. Dublin,
Savković, M. M., Gašić, M. M., Ćatić, D. M., et al. (2013). Opti-
Ireland: NSAI.
mization of the box section of the main girder of the bridge
CMAA Specification No. 70. (2015). Top running bridge and gantry
crane with the rail placed above the web plate. Structural and
type multiple girder electric overhead traveling cranes. Alex-
Multidisciplinary Optimization, 47(2), 273–288.
andria, VA: CMAA.
Sun, C., Tan, Y., Zeng, J., et al. (2011a). The structure optimiza-
CMAA Specification No. 74. (2015). Top running and under running
tion of main beam for bridge crane based on an improved PSO.
single girder electric overhead cranes utilizing under running
Journal of Computers (Taipei), 6(8), 1585–1590.
trolley hoist. Alexandria, VA: CMAA.
Sun, C., Zeng, J., & Pan, J. (2011b). A modified particle swarm
Djelosevic, M., Gajic, V., Petrovic, D., et al. (2012). Identification of
optimization with feasibility-based rules for mixed variable
local stress parameters influencing the optimum design of box
optimization problems. International Journal of Innovative
girders. Engineering Structures, 40, 299–316.
Computing, Information and Control, 7(6), 3081–3096.
Fan, X., & Bi, X. (2015). Reliability-based design optimization
Tian, G., Zhang, S., & Sun, S. (2012). The optimization design of
for crane metallic structure using ACO and AFOSM based on
overhead traveling crane’s box girder. Advanced Materials
China standards. Mathematical Problems in Engineering, Arti-
Research, 538–541, 2850–2855.
cle ID 828930, 12 p.
Vinot, P., Cogan, S., & Piranda, J. (2001). Shape optimization of
Fan, X., Wang, P., & Hao, F. (2019). Reliability-based design
thin-walled beam-like structures. Thin-Walled Structures, 39(7),
optimization of crane bridges using Kriging-based surrogate
611–630.
models. Structural and Multidisciplinary Optimization, 59(3),
993–1005.
Fan, X., & Zhi, B. (2017). Design for a crane metallic structure Publisher’s Note Springer Nature remains neutral with regard to
based on imperialist competitive algorithm and inverse reli- jurisdictional claims in published maps and institutional affiliations.
ability strategy. Chinese Journal Mechanical Engineering, 30,
900–912.

13

You might also like