Hindawi

Mathematical Problems in Engineering

Volume 2018, Article ID 8430578, 12 pages
https://doi.org/10.1155/2018/8430578

Research Article
Parameter Optimization and Prediction Model of Induction
Heating for Large-Diameter Pipe

Xiurong Fang , Jia Lu, Junfeng Wang, and Jinhui Yang

School of Mechanical Engineering, Xi’an University of Science and Technology, Xi’an 710054, China

Correspondence should be addressed to Xiurong Fang; fangxr098@163.com

Received 26 July 2018; Revised 6 October 2018; Accepted 1 November 2018; Published 26 November 2018

Guest Editor: Arshad Afzal

Copyright © 2018 Xiurong Fang et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The parameters of induction heating of large-diameter pipes have a direct effect on the final processing quality of the elbow, and the
complexity of multifield coupling of magnetothermal force in induction heating can make it impossible to quantitatively optimize
the design parameters of the induction heating device. In this paper, X80 pipeline steel induction heating is taken as the research
object, and a corresponding numerical model is established. The influence of induction heating process parameters on the heating
temperature of pipeline steel under the skin effect is determined. First, the influence of process parameters on the heating effect of
pipeline steel is quantified by orthogonal test. Then, taking the optimum temperature difference between the inner and outer wall of
X80 pipeline steel during the induction heating process as a target, the optimal process parameter set of the pipe induction heating
is determined by using neural network genetic algorithm. Finally, comparing the relevant test criteria of the regression equation,
the optimum mathematical prediction model of the outer wall temperature of the pipe induction heating process is obtained, which
provides a theoretical basis for optimization of the process parameters of the pipe-based induction heating device.

1. Introduction forming process for large-diameter pipe is not yet clear. The
detailed theoretical guidance and basis for the selection and
The induction heating of pipe is one of the common tech- optimization of design parameters on the induction heating
niques for production of elbows thanks to its rapid heating, device are not yet available. Thus, it is impossible to ensure
easy cooling, small thinning rate of wall thickness, and a reasonable matching between the heating time of the pipe
easy adjustment of the bending radius of elbow. In the and the pushing speed of the hydraulic cylinder, resulting
actual project of oil and gas transmission engineering, in the fact that the reliability of the pipe quality cannot be
the bending of large-diameter pipes in oil and gas guaranteed [8]. Therefore, the induction heating device of
transportation pipe network is mostly processed by the pipe has become one of the important factors affecting
hydraulic push-type induction heating pipe bending the final processing quality of the elbow. In this paper, the
machine. The system diagram is shown in Figure 1 [1], optimization of the process parameters of pipe induction
and the specific induction heating diagram is shown in heating is investigated by means of numerical simulation and
Figure 2 [2]. It is required to realize continuous operation mathematical methods.
of propulsion󳨀→heating󳨀→bending󳨀→cooling󳨀→forming
while the induction heating device fulfills the local heating 2. Analysis on Induction Heating
of the pipe when the pipe bending machine is working Parameters of Pipes
[3]. At present, many researchers have conducted a lot
of studies on the bending process of different materials, The basic principle of induction heating is shown in Figure 3:
the mechanical properties, and the microstructure of the coil is connected to the heating power source, and an
the bent pipe [4–7]. However, the influence of various alternating magnetic field is generated in the pipe when
parameters on the heating effect of the induction heating the alternating current in the heating power source flows
2 Mathematical Problems in Engineering

movable crossbeam induction coil

hydraulic cylinder support roller tube blank front holder turning arm

Figure 1: Structure of the bend of induction heating pipe bending machine.

Figure 2: Specific induction heating.

Alternating magnetic field

Heating
Induction coil
power
supply

Pipe

Figure 3: Schematic diagram of pipe induction heating.

through the coil. Under the action of the alternating magnetic resistance, so that the pipe temperature rises rapidly to
field, vortex current is generated inside the pipe, and the achieve the purpose of heating [9].
vortex current and alternating current in the induction coil It can be seen that electromagnetic induction is used to
have opposite directions with the same frequency. When the generate vortex current heating in the conductor to realize the
vortex current passes through the pipe, the pipe generates a heating of the work piece. Therefore, by the electromagnetic
large amount of Joule heat under the action of the material induction law (1), the Joule-Electric law (2), and technical
Mathematical Problems in Engineering 3

requirements of the large-diameter pipe induction heating differential form of the transient temperature field governing
forming device, we can obtain the main influential factors on equation in induction heating can be derived:
the heating effect of the pipe, such as the induction frequency,
current density, and air gap [10]. 𝜕𝑇 𝜕 𝜕𝑇 𝜕 𝜕𝑇 𝜕 𝜕𝑇
𝜌𝑐 − (𝜆 𝑥 ) − (𝜆 𝑦 ) − (𝜆 𝑧 )
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧 (7)
𝑑Φ
𝑒 = −𝑁 (1) − 𝑞V = 0
𝑑𝑡
𝑄 = 0.24𝑖2 𝑅𝑡 (2) where 𝜆 is thermal conductivity (W/(m⋅∘ C)); T is temperature
field distribution function; 𝑞V is intensity of the heat source of
where 𝑁 is the number of turns of the induction coil; Q induction vortex current in the pipeline (W/m3 ); 𝜌 is material
is Joule heat (unit is J); i is the effective value of induction density (Kg/𝑚3 ); c is the specific heat capacity of material
current (unit is A); R is conductor resistance (unit is Ω); and 𝑡 (J/(Kg⋅ K)).
is the time of the current flowing through the conductor (unit Use the Newton convection boundary as a temperature
is s). boundary condition:
Taking the elbow forming of X80 pipeline steel [11]
with diameter Φ1219mm and wall thickness 𝛿 27 mm as an 𝜕𝑇
example, the numerical modeling of induction heating is −𝜆 = ℎ (𝑇 − 𝑇0 ) (8)
𝜕𝑛
carried out.
where 𝑛 refers to the normal direction on the boundary; 𝜆 is
2.1. Mathematical Model of Induction Heating Electromagnetic thermal conductivity (W/m⋅∘ C); h is convective heat transfer
Field. In the numerical calculation of finite element, the coefficient (W/m2 ⋅K); T is work surface temperature (K); and
mathematical model of the induction heating vortex current 𝑇0 indicates ambient temperature (K).
field for the heated workpiece area can be obtained by
the control equation (Maxwell equations) and the magnetic 2.3. Coupling of Induction Heating Electromagnetic Field-
→
󳨀 Temperature Field. The induction heating is involved in the
vector-scalar potential ( 𝐴-Φ) theory of the induction heating
electromagnetic field [12–14]: interaction of two physical fields of electromagnetic field
and temperature field. The complexity of solving the magne-
→
󳨀 tothermal coupling problem makes it difficult to obtain accu-
1 →󳨀 1 →󳨀 𝜕𝐴 rate solutions for many engineering issues [18]. Hopefully,
∇× ∇ × 𝐴 − ∇ (∇ ⋅ 𝐴) + 𝜎 + 𝜎∇Φ = 0 (3)
𝜇 𝜇 𝜕𝑡 the large-scale finite element numerical simulation software
can be used for engineers to solve complex problems on
𝜕 →󳨀
∇2 Φ + (∇ ⋅ 𝐴) = 0 (4) the basis of theoretical analysis. This paper is aimed at the
𝜕𝑡 characteristics of induction heating magneto-thermal cou-
→󳨀 pling of large-diameter pipes. We adopt ABAQUS software
→
󳨀 𝜕𝐴 (5)
𝐽𝑒 = −𝜎 − 𝜎∇Φ to solve the magneto-thermal coupling issue by sequential
𝜕𝑡 coupling [19]. First, the vortex current is obtained in the
electromagnetic field; then the Joule heat of the workpiece
where ∇ is Hamilton operator; 𝜇 is magnetic conductivity under the action of vortex current field is got. Eventually, the
(H/m); 𝜎 is material conductivity (S/m); 𝜀 is capacitivity Joule heat is used as the temperature field input condition
→
󳨀
(F/m); 𝐴 is magnetic vector potential; Φ is scalar potential; to obtain the temperature distribution on the workpiece.
→󳨀
and 𝐽𝑒 is induction vortex current density vector (A/𝑚2 ). The analysis process of magnetothermal coupling during the
In order to ensure the continuity condition of the field induction heating is shown in Figure 4.
quantity at the interface of different media, consider the
boundary condition as a ferromagnetic boundary with sur- 2.4. Building of Numerical Model. The geometric model of
→󳨀 pipe induction heating consists of three parts: X80 pipe,
face current density 𝐽 . The expression is shown as
induction coil, and air. Since the pipe is only involved in
the solution of temperature field, it is assumed that, in the
󳨀𝑛 × ( 1 ∇ × →
→ 󳨀 →
󳨀
𝐴) = − 𝐽 solution of the electromagnetic field, the physical properties
(6)
𝜇 of the material corresponding to the induction coil and air are
not affected by the temperature change. Since the X80 pipe,
2.2. Mathematical Model of Induction Heating Temperature induction coil and air together constitute the electromagnetic
Field. It is assumed that the Joule heat generated by the path in the induction heating process, the thermal property
vortex current in the alternating magnetic field is completely parameters of the relative magnetoconductivity, resistivity,
converted into its own heat; that is, the induced current specific heat capacity, thermal conductivity, density, etc.
Joule heat obtained by the electromagnetic field is used as an corresponding to the three parts should be considered in
internal heat source to heat the workpiece, and the heating solving the electromagnetic field (as shown in Table 1). The
process is an unsteady heat conduction process [15, 16]. Based material of induction coil is T3 copper, its resistivity is 1.75 ×
on Fourier’s law and the first law of thermodynamics [17], the 10−8 Ω∙m, and the relative magneto-conductivity is 1. As air is
4 Mathematical Problems in Engineering

Table 1: The thermal properties of X80.

Temperature Relative Resistivity 𝜌 Specific Heat 𝐶𝑝 Thermal Conductivity Density Emissivity

(∘ C) Permeability 𝜇 (10−6 Ω⋅m) (J/Kg⋅∘ C) 𝐾 (W/m⋅∘ C) 𝜌(Kg/m3 ) 𝐶𝑠 (J/m2 sK4 )
0 200 0.195 450 47.4 7896 1
25 198 0.202 450 46.2 7896 0.998
100 194 0.251 479 43.7 7870 0.996
200 186 0.336 520 40.3 7849 0.992
300 179 0.434 564 38 7824 0.987
400 167 0.539 612 35.9 7796 0.983
500 156 0.659 668 33.8 7769 0.978
600 139 0.792 785 31.7 7744 0.978
700 93 0.947 936 28.4 7716 0.879
800 1 1.073 848 25.9 7691 0/648
900 1 1.164 655 23.5 7665 0.594
1000 1 1.199 620 23.9 7639 0.535

not affected by the current, the relative magnetoconductivity

Start should be set as 1 when performing numerical simulation.
It is one of the important influential factors on the quality
of the pipe to reasonably match the heating time of the pipe
and the pushing speed of the hydraulic cylinder during the
Establish Model bending process of the large-diameter pipe (see Figure 1).
Thus, in order to further study the problem of matching the
Enter electromagnetic
induction heating time with the pushing speed, the finite
field analysis options
element model is built in Cartesian coordinate system. At
Establish
the same time, considering that the large-diameter pipe and
electromagnetic field
its applied load are axisymmetric, 1/4 of the pipe is used to
analysis options
build a finite element model. The working loads are applied
Clear electromagnetic
to the model according to the characteristics of each physical
field analysis options,
enter temperature field
field; the corresponding analysis steps, solution variables,
analysis options and boundary conditions are setup; different grid types are
selected to mesh the model (the unit type of EMC3D8 is
Establish temperature
selected in the electromagnetic field and the unit type of
field analysis options
DC3D8 is selected in the temperature field). In order to
ensure the accuracy and convergence of the calculation, the
Read electromagnetic
pipe mesh is refined, as shown in Figure 5.
field analysis options

Compute
3. Results Analysis
electromagnetic field
3.1. Analysis of Numerical Simulation Results. In order to
visually display the dynamic changes of the electromagnetic
Read temperature field field and temperature field inside the pipe during the heating
Compute temperature
process, the main parameters such as current frequency
field (=1000Hz), current density (J=6 × 107 A/m2 ), and air gap
(d=30mm) are determined according to the engineering
practice, and the distribution clouds of the Joule heat and
temperature of the pipe under induction heating are obtained
End
through simulation (see Figures 6 and 7), and the curves of
temperature for the inner and outer wall of the pipe with
the times are shown in Figure 8. It can be seen that the
Figure 4: Analysis process of magnetothermal coupling during the distribution of Joule heat and temperature on the pipe is that
induction heating. the outer wall’s is much greater than that of the inner wall due
Mathematical Problems in Engineering 5

air
induction coil
pipe air
induction coil

X
pipe
Y

(a) (b) (c)

Figure 5: Numerical model of induction heating for thin-walled pipe. (a) Geometric model. (b) Finite element model. (c) Finite element
model of the pipe.

700. Outer Wall

EMJH 600.

Temperature (∘ C)
Inner Wall
(Avg: 75%) 500.
+1.307e+09 400.
+1.198e+09 300.
+1.089e+09 200.
+9.802e+08 100.
+8.713e+08
+7.624e+08 0.
+6.535e+08 −100.
+5.445e+08 0. 2. 4. 6. 8. 10. 12. 14. 16. 18. 20.
+4.356e+08 Time (s)
+3.267e+08
+2.178e+08 Figure 8: Temperature curves of the inner and outer wall of pipe
+1.089e+08
+4.984e-20 with the times.

It can be seen from Figure 8 that although there is

Figure 6: Distribution cloud of Joule heat in pipe (Unit: J). unevenness in the temperature rise of the inner and outer
walls in the pipe, the difference in temperatures between
the inner and outer walls tends to be stable as the increase
of heating time. Therefore, by parameter optimization, the
NT11
+7.574e+02 difference in temperature gradient between the inner and
+6.943e+02 outer walls of the pipe is reduced as much as possible.
+6.312e+02
+5.681e+02
+5.049e+02
In order to further explore the relationship between the
+4.418e+02 current frequency, current density, air gap, and pipe heating
+3.787e+02
+3.156e+02 effect, the control variable method is used to change the three
+2.525e+02
+1.894e+02 process parameters for simulation calculation. The results are
+1.262e+02
+6.312e+01 shown in Figure 9.
+1.554e-04 Figure 9 shows the effect of the current frequency, current
Step: Step-1 density, and the air gap between coil and pipe on the
Increment 982 temperature rise of the inner and outer wall of the pipe,
Step Time = 20.00 respectively. It can be seen from the figure although that
Primary Var: NT11 relationship between the variation of parameters and the
temperature of the inner and outer wall of the pipe is basically
linear; the influence of three parameters on the heating effect
of pipe cannot be quantified. Moreover, the temperature
Figure 7: Distribution cloud of temperature in pipe (Unit:∘ C). difference between the inner and outer wall of the pipe is the
key quantity; thus it is necessary to analyze the variation rule
of the temperature difference between the inner and outer
to the effects of skin effect, proximity effect, and ring effect wall of the pipe to find the optimal parameter group.
in the induction heating process [20], which ultimately leads
to the greater temperatures difference of the inner and outer 3.2. Parameter Optimization Based on Orthogonal Test Design
walls of the pipe, causing a great inconvenience to the pushing
speed of the elbow, which in turn affects the quality of the pipe 3.2.1. Orthogonal Test Design for Induction Heating of Pipe.
bending. The three process parameters of the current frequency,
6 Mathematical Problems in Engineering

1100 1100
1000 1200
1000 1100
900

Temperature (∘ C)
Temperature (∘ C)

Temperature (∘ C)
900 1000
800 800 900
700 700 800
600 600 700
500 500 600
400 400 500
300 300 400
200 300
200 200
100 100 100
0 0 0
600 800 1000 1200 1400 1600 5.0 5.5 6.0 6.5 7.0 10 20 30 40 50
Current Frequency (Hz) Current Density (107 A/Ｇ2 ) Air Gap (mm)

Outer Wall Outer Wall Outer Wall

Inner Wall Inner Wall Inner Wall
(a) (b) (c)

Figure 9: Relationship between three parameters and the inner and outer wall temperature of pipe. (a) Current frequency. (b) Current density.
(c) Air gap between pipe and induction coil.

Table 2: Test parameters and levels. the primary and secondary relationship of the influence of
various process parameters on the heating effect is obtained.
Factor Then we use the range analysis to verify the primary and
Level Current Frequency Current Density Air Gap secondary relationships again to provide a guidance for
𝑓(Hz) 𝐽(A/m2 ) 𝑑(mm) engineering process design.
1 600 5×107 10
2 800 5.5×107 20 (1) Variance Analysis. The analysis results of the influence of
3 1000 6×107 30 each parameter on temperatures of the inner and outer walls
of the pipe are shown in Tables 4 and 5.
4 1200 6.5×107 40
5 1500 7×107 50 (2) Range Analysis. First, the comprehensive mean and
range of each factor are obtained, and then the primary
and secondary relationships of the influence of the process
current density, and air gap are the factors of the orthogonal parameters on the heating effect are got by comparing the
test [21], and each factor is set at 5 levels, as shown in Table 2. extremes [22]. The range analysis of the inner and outer
It can be seen from Table 2 that there is a total of 125 wall temperatures of the pipe under different levels in the
experimental parameter combinations, and the orthogonal orthogonal test is shown in Table 6:
test is mainly used to analyze the primary and secondary According to the analysis results of Tables 4 and 5, when
relationship of the influence of various process parameters on the temperature of the inner and outer walls of the pipe is
the heating effect. Therefore, according to the orthogonality taken as an index, the influence of the current frequency and
and the engineering practice, some parameter groups with current density on the index is significant, and the influence
the characteristics of “uniform dispersion, neatness, and of the air gap on the index is extremely significant. And in the
comparability” are selected as the representative ones for the medium-frequency induction heating of the pipe, the influ-
experiment. ential degree of each process parameter on the heating effect
In the orthogonal experimental analysis, the Taguchi of the pipe is air gap>current density>current frequency.
design method is used to neglect the interaction between From the range R corresponding to each process parameter
different process parameters, the L25 (56 ) orthogonal table in Table 6, it can be concluded that the influential degree
is selected, and the empty column is adopted as the error of the process parameters on the heating effect of the pipe
column. According to the 25 sets of parameters selected in is consistent with the variance analysis result, which further
Table 3, the numerical simulation analysis of the induction verifies the accuracy of the obtained results. Compared with
heating of X80 pipe (Φ1219mm, wall thickness 27 mm) is the range analysis, the contribution rate of each process
carried out to obtain the temperatures of the inner and outer parameter to the survey indicators is defined more clearly by
walls of the pipe. The orthogonal experimental scheme is the variance analysis.
designed with the temperatures of the inner and outer walls It can be known from the engineering that when the
of the pipe as the test indexes (see Table 3). induction heating temperatures of the outer wall of the pipe
are 900∼1000∘C [23] and the difference between the inner and
3.2.2. Analysis of the Orthogonal Test Results of Induction outer wall temperature is small, the final processing quality
Heating of Pipe. The variance method is used to determine of the elbow is better. Combined with the above analysis, a
the significant degree and contribution rate of the influence of better process parameter group can be initially determined
various process parameters on the heating effect of medium- as current frequency 1500 Hz, current density 5.5 × 107 A/m2 ,
frequency induction heating of large-diameter pipes, and and air gap 10 mm.
Mathematical Problems in Engineering 7

Table 3: Orthogonal test design scheme and numerical calculation results.

Factor Test Index

Test No. Outer Wall Inner Wall
Current Current Density Interval
Temperature Temperature
Frequency (Hz) (A/m2 ) (mm)
(∘ C) (∘ C)
1 1(600) 1(5×107 ) 1(10) 649 127
2 1(600) 2(5.5×107 ) 2(20) 618 125
3 1(600) 3(6×107 ) 3(30) 574 119
4 1(600) 4(6.5×107 ) 4(40) 528 114
5 1(600) 5(7×107 ) 5(50) 474 105
6 2(800) 1(5×107 ) 2(20) 587 121
7 2(800) 2(5.5×107 ) 3(30) 569 120
8 2(800) 3(6×107 ) 4(40) 546 117
9 2(800) 4(6.5×107 ) 5(50) 508 113
10 2(800) 5(7×107 ) 1(10) 1526 215
11 3(1000) 1(5×107 ) 3(30) 532 114
12 3(1000) 2(5.5×107 ) 4(40) 528 115
13 3(1000) 3(6×107 ) 5(50) 508 113
14 3(1000) 4(6.5×107 ) 1(10) 1420 208
15 3(1000) 5(7×107 ) 2(20) 1349 205
16 4(1200) 1(5×107 ) 4(40) 509 112
17 4(1200) 2(5.5×107 ) 5(50) 508 114
18 4(1200) 3(6×107 ) 1(10) 1319 201
19 4(1200) 4(6.5×107 ) 2(20) 1297 202
20 4(1200) 5(7×107 ) 3(30) 1248 200
21 5(1500) 1(5×107 ) 5(50) 510 115
22 5(1500) 2(5.5×107 ) 1(10) 989 173
23 5(1500) 3(6×107 ) 2(20) 1255 200
24 5(1500) 4(6.5×107 ) 3(30) 1253 202
25 5(1500) 5(7×107 ) 4(40) 1239 204

Table 4: Variance analysis results of the orthogonal test of temperatures for pipe’s outer wall.

Sources of Degree of Sum of Contribution

Mean Square F Significance
Variance Freedom Squares Rate (%)
Current
4 0.996 0.25 31.63 21.28 ∗
Frequency
Current
4 1.49 0.372 47.32 31.84 ∗
Density
Air Gap 4 2.1 0.525 66.68 44.86 ∗∗
Error 12 0.095 0.008 – 2.02
Total 24 4.671 1.155 – 100
Note: ∗ indicates that the difference is significant; ∗∗ indicates that the difference is extremely significant.

3.3. Process Parameter Optimization Based on Neural Net- neural network method to establish the mapping relationship
work and Genetic Algorithm. The orthogonal experiment can between the induction heating process parameters and the
determine the influence degree of the process parameters on temperatures of the inner and outer walls of the pipe [24, 25].
the temperature of the inner and outer walls of the pipe and The data is trained to predict the heating effect of all combi-
judge the better parameter group from the existing process nations of process parameters within the variation range, and
parameter group. However, it is difficult to obtain the global the process parameters are optimized by genetic algorithm
optimal process parameters for the multiparameter group [26, 27]. Thus, the global optimal process parameter group
problems which affect the actual engineering. We use BP is obtained.
8 Mathematical Problems in Engineering

Table 5: Variance analysis results of orthogonal test of temperatures for pipe’s inner wall.

Sources of Degree of Sum of Contribution

Mean Square F Significance
Variance Freedom Squares Rate (%)
Current
4 0.478 0.12 37.77 25.78 ∗
Frequency
Current
4 0.589 0.147 46.47 31.72 ∗
Density
Air Gap 4 0.751 0.188 59.26 40.45 ∗∗
Error 12 0.038 0.003 – 2.05
Total 24 1.856 0.458 – 100
Note: ∗ indicates that the difference is significant; ∗∗ indicates that the difference is extremely significant.

Table 6: Range analysis results of orthogonal test for pipe’s outer wall temperature.

Outer Wall Temperature (∘ C) Inner Wall Temperature (∘ C)

Current Current Current Current
Air Gap Air Gap
Frequency Density Frequency Density
K1 568.6 557.4 1180.6 118 117.6 184.8
K2 747.2 642.4 1021.2 137.2 129.4 170.6
K3 867.4 840.4 835.2 151 150 151
K4 976.2 1001.2 670 165.8 167.8 132.4
K5 1049.2 1167.2 501.6 178.8 185.8 112
R 480.6 609.8 679 60.8 68 72.8
Note: K1∼K5 represent the average temperature of each factor at each level, and R is range.

Current frequence f
Outer wall
temperature T1
.
Current density J .
.
Inner wall
temperature T2
Air gap d

Input Hidden layer Output

Figure 10: Structure of pipe induction heating BP neural network.

3.3.1. Prediction Model Based on BP Neural Network. The iteration is 274 times, the training is finished with the system
current frequency, current density, and air gap are selected error of 9.75 × 10−7 ; thus its error is within the expected range,
as the network model input, and the outer and inner wall and the network output is more accurate.
temperatures 𝑇1 and 𝑇2 of the pipe are taken as output to In order to verify the accuracy of the training prediction
build a three-layer BP neural network model. The network results, the five new sets of process parameters are rese-
structure is shown in Figure 10. lected as test samples for training prediction and numerical
The data (see Table 3) obtained by numerical simulation simulation. The training prediction results and numerical
is used as a training sample, and the sample is trained by simulation results are shown in Table 7.
running a BP neural network program. The training target is According to the above results in Table 7, the error
selected as 0.000001 and the learning rate is 0.1. The training between the prediction data and the simulation data is within
of the network model is shown in Figure 11. When the 3%, which indicates that the prediction accuracy of the BP
Mathematical Problems in Engineering 9

Table 7: Error between numerical simulation and BP neural network training.

Current Frequency Current Density Air Gap Outer Wall Temperature 𝑇1 (∘ C) Inner Wall Temperature 𝑇2 (∘ C)
𝑓(Hz) 𝐽(A/m2 ) 𝑑(mm) Simulation Prediction Error Simulation Prediction Error
1200 6×107 30 883 862.5 2.3% 164 163.29 0.43%
1500 6×107 30 1043 1036.5 0.62% 183 185.35 1.3%
1000 5.5×107 30 640 638.2 0.28% 132 130.67 1%
1000 6.5×107 30 900 894.5 0.61% 167 163.40 2.1%
1000 7×107 30 1070 1079.3 0.87% 184 180.71 1.8%

Fitness curve
 840

830
−
 820
System error

810

Adaptability
800
−
790

780
−
770

0 50 100 150 200 250 760

Number of iterations 750
0 10 20 30 40 50 60 70 80 90 100
Train Evolutionary algebra
Goal
Figure 12: Genetic algorithm optimization process.
Figure 11: BP neural network training error curve.

neural network for the heating effect is higher, and the

prediction result is close to the actual situation. 3.4. Pipeline Induction Heating Prediction Model Based on
Regression Analysis. Through the aforementioned analysis
3.3.2. Optimization of Induction Heating Parameters Based and calculation, the selection of optimal parameter group
on Genetic Algorithm. The population with a size of 50 is for the induction heating parameter design can be realized,
selected and the binary code is adopted. The crossover proba- but In order to further guide the engineering practice, a
bility and mutation probability are 0.75 and 0.01, respectively. prediction model of pipe induction heating is necessary.
The BP neural network prediction result is used as the fitness In this paper, the regression analysis of the data obtained
function to control the temperature of the inner and outer from numerical simulation is carried out to determine a
walls of the pipe [28, 29]. The genetic algorithm with 100 times reasonable mathematical model for the prediction of the
of iteration is designed to optimize the process parameters outer wall temperature, which provides a certain theoretical
of pipe induction heating. The software is used to iteratively guidance for the actual processing. The data obtained from
calculate the BP neural network and genetic algorithm joint the numerical simulation of medium-frequency induction
optimization program. After 100 iterations, the target fitness heating of large-diameter pipe are taken as samples, and
the process parameters that influence the induction heating
is shown in Figure 12.
effect of pipe materials, such as current frequency  (Hz),
It can be seen from the Figure 12 that, after 30 iterations, current density J (A/m2 ), and air gap d (mm), are seen as
the fitness value of the temperature difference between the independent variables, while the outer wall temperature of
inner and outer wall under the processable condition is the pipe is seen as the dependent variable. Subsequently,
maintained at 751; meanwhile the temperature 𝑇1 of the two regression models are established for prediction of the
outer wall and the temperature 𝑇2 of the inner wall is 923∘ C pipe outer wall temperature 𝑇 on the basis of combining
and 172∘ C, respectively. The process parameter group that the feature of sample data. It can be seen from Figure 9
is superior to the orthogonal test in the induction heating that the process parameters and the temperature of the
process of the pipe is obtained with the current frequency, inner and outer walls show a linear change law; thereby a
current density, and air gap which are 1460 Hz, 5.65 × regression Model I is established. However, considering the
107 A/m2 , and 15 mm, respectively. magnitude difference of the selected values of each parameter,
10 Mathematical Problems in Engineering

Table 8: Results of regression analysis.

Model I: 𝑇 = 𝑏0 𝑓 + 𝑏1 𝐽 + 𝑏2 𝑑 + 𝐶 Model II: 𝑇 = 𝐶𝑓𝑏0 𝐽𝑏1 𝑑𝑏2

−5
Regression Coefficients (0.525, 3.092×10 , -17.184, -1049.604) (-33.103, 0.642, 2.061, -0.481)
Sum of Residuals 181575.2825 0.295
Return Square Sum 3712511.687 4.915
Correlation Coefficient
0.953 0.943
r2
F-test Value 129.515 120.585
P 1.029×10−20 2.23×10−19

20 12

10
15
8
Frequency

Frequency
10 6

4
5
2

0 0
−4 −2 0 2 −4 −3 −2 −1 0 1 2
Regression Standardized Residual Regression Standardized Residual
(a) Model I (b) Model II

Figure 13: Regression standardized residual histogram.

the logarithmic conversion of the sample data is performed to It can be seen from Figure 13 that although the normalized
build a regression Model II [30]. residuals of Model I and Model II are basically normal distri-
bution, comprehensive analysis of the calculation results of
Regression Model I: 𝑇 = 𝑏0  + 𝑏1 𝐽 + 𝑏2 𝑑 + 𝐶 Table 8, including the correlation coefficient, sum of squared
Regression Model II: 𝑇 = 𝐶𝑏0 𝐽𝑏1 𝑑𝑏2 residuals, and F-test value. It can been seen that Model I has
an optimal fitting degree; thus the regression equation (1) is
Assume that the above two regression models meet the chosen as the mathematical model for predicting the outer
following conditions: wall temperature of large-diameter pipe during the medium-
(1) 𝜀𝑖 ∼ 𝑁(0, 𝜎2 ), 𝐶𝑜V(𝜀𝑖 , 𝜀𝑗 ) = 0, 𝑖 ≠ 𝑗 frequency induction heating. In order to verify the accuracy
(2) The selected variables are not random and have no of the regression models, five sets of process parameters
multiple linear relationship with each other are taken as test samples for the numerical simulation and
(3) There is a linear relationship between the dependent regression model prediction. The results are shown in Table 9.
variable 𝑇 and the transformed independent variable in the It can be seen from Table 9 that errors between the simu-
regression model II lated data and predicted data are within 5%. The error value
The two regression models are analyzed and solved by a needs to be determined according to industry standards.
multiple regression calculation program, and the results are However, it is worth noting that the mathematical model of
shown in Table 8. the outer wall temperature prediction can accurately reflect
The regression equations obtained by stepwise regression the mapping relationship between the process parameters of
calculation are as follows: the induction heating and the outer wall temperatures of the
pipe. The mathematical model of the outer wall temperature
Equation of Model I: 𝑇 = 0.525𝑓 + 3.092 × prediction for large-diameter pipe in the induction heating
10−5 𝐽-17.184𝑑-1049.604 process has certain significance for guiding engineering
Equation of Model II: 𝑇 = 𝑒−33.103 0.642 𝐽2.061 𝑑−0.481 practice.
Mathematical Problems in Engineering 11

Table 9: Errors between the numerical simulation and regression model prediction.

Current Frequency Current Density Outer Wall Temperature T (∘ C)

Air Gap 𝑑(mm)
𝑓(Hz) 𝐽(A/m2 ) Simulation Return Error
1200 6×107 30 883 920.1 4.2%
1500 6×107 30 1043 1077.6 3.2%
1000 5.5×107 30 640 660.5 3.1%
800 6×107 30 683 710.1 4.0%
1000 6×107 40 626 643.2 2.7%

4. Conclusions References
(1) Based on the basic theory of induction heating and [1] R. X. Fang and W. S. Cao, “The research of hydraulic technology
analysis of induction heating forming process for large- for large caliber medium frequency heating pipe hydraulic
diameter X80 pipe, the magnetic-thermal coupling issue bending machine,” Chinese Hydraulics & Pneumatics, 2010.
in the heating process is simulated by sequential coupling [2] X. Fang, J. Wang, and W. Cao, “Induction heating of large-
method. The influential law of each group’s induction heating diameter thin-walled elbow based on skin effect,” Oil & Gas
process parameters (current frequency, current density, and Storage and Transportation, vol. 36, no. 8, pp. 958–963, 2017.
air gap) on the temperature of the inner and outer walls of [3] X. Fang, X. Meng, and H. Fan, “Research on Ultra-low and
the pipe has been obtained. Constant Speed Hydraulic Control System of Large Push-
(2) Taking the induction heating process parameters as bending Style Pipe Bender,” Machine Tool & Hydraulics, vol. 19,
pp. 61–64, 2017.
the test factors, we carry out the orthogonal test design
of the process parameters with the difference of the inner [4] C. Pandey, M. M. Mahapatra, P. Kumar, N. Saini, and J. G.
Thakre, “Nano-size Particle Evolution During Heat Treatment
and outer wall temperature of the pipe as the evaluation
of P91 Steel and Their Effect on Micro Hardness,” Transactions
index, the influential degree of each process parameter on of the Indian Institute of Metals, pp. 1–8, 2017.
the heating effect is determined as air gap>current density>
[5] C. Pandey and M. M. Mahapatra, “Effect of Heat Treatment on
current frequency, and the optimized process parameter sets
Microstructure and Hot Impact Toughness of Various Zones
for the heating bending of the pipe have been obtained of P91 Welded Pipes,” Journal of Materials Engineering and
using the neural network genetic algorithm, with the current Performance, vol. 25, no. 6, pp. 2195–2210, 2016.
frequency, current density, and air gap which are 1460 Hz, [6] C. Zheng, H. Liu, X. Ding, and Q. Fu, “Horizontal Vibration of
5.65 × 107 A/m2 , and 15 mm, respectively. a Large-Diameter Pipe Pile in Viscoelastic Soil,” Mathematical
(3) The linear regression models of two types of pipe Problems in Engineering, vol. 2013, Article ID 269493, 13 pages,
temperature have been built based on the simulation results 2013.
of different heating parameters. By comparing and analyzing [7] C. Pandey, M. Mahapatra, P. Kumar et al., “Some studies on P91
the relevant test standards of each regression equation, we steel and their weldments,” Journal of Alloys & Compounds, vol.
have obtained the mathematical model of the outer wall 743, pp. 332–364, 2018.
temperature prediction during the induction heating process [8] S. Q. Pan, X. U. Zhe, and H. E. Bi-Shi, “A Study on Temperature
of the pipe. Control for the Billet Electromagnetic Induction Heating Based
(4) From the bend forming practice of X80 pipe, the on Simulation,” Industrial Heating, vol. 41, no. 6, pp. 25–28, 2012.
numerical simulation process parameters sets are somewhat [9] H. Du, J. Li, and Y. Qu, “Mathematical Modeling of Eddy-
limited. Hopefully, our mathematical analysis method is of a Current Loss for a New Induction Heating Device,” Mathemat-
referential significance for practical engineering. ical Problems in Engineering, vol. 2014, Article ID 923745, 7
pages, 2014.
Data Availability [10] Y. Wang, C. Liu, and X. Wang, “Influence of induction coil
parameter on temperature field of steel plate and choosing of
The data used to support the findings of this study are coil size,” Jinshu Rechuli/Heat Treatment of Metals, vol. 41, no. 4,
available from the corresponding author upon request. pp. 178–182, 2016.
[11] T. D. S. Antonino, P. B. Guimarães, R. D. A. Alécio et al.,
Conflicts of Interest “Measurements of the Thermophysical Properties of the API 5L
X80,” Materials Sciences & Applications, vol. 5, no. 8, pp. 617–627,
The authors declare that they have no conflicts of interest. 2014.
[12] Y. H. Zhang, Numerical Simulation of Induction Heating Tem-
Acknowledgments perature Field, Jiangnan University, 2008.
[13] Z. QianZhe, Numerical Simulation and Experimental Study on
This work was supported by National Natural Science Foun- Induction Heating Process of Steel Bar Production Line, North
dation of China (NSFC) (no. 51775427) China Electric Power University, 2013.
12 Mathematical Problems in Engineering

[14] Z. Tianxu, Numerical analysis and experimental study on [30] C. Pandey, A. Giri, and M. M. Mahapatra, “On the prediction
dynamic medium frequency induction heating process of high of effect of direction of welding on bead geometry and residual
frequency straight seam welded pipe, Yanshan University, 2016. deformation of double-sided fillet welds,” International Journal
[15] Y. Chenguang, Numerical Simulation of Variable Power Induc- of Steel Structures, vol. 16, no. 2, pp. 333–345, 2016.
tion Heating of 42CrMo Steel Shaft Parts, Yanshan University,
2010.
[16] S. Wenjie, Numerical simulation and experimental study on elec-
tromagnetic field and temperature field of large-diameter molten
silicon single crystal growth equipment, Zhejiang University,
2013.
[17] J. L. Fleming and J. Moser, “A Fourier Approach to Model Elec-
tromagnetic Fields Scattered by a Buried Rectangular Cavity,”
Mathematical Problems in Engineering, vol. 2009, Article ID
937327, 22 pages, 2009.
[18] X. Yi, L. Li, J. Yue et al., “Simulation of coupling of electromag-
netic and thermal of intermediate frequency induction heating
of pipeline,” Heat Treatment of Metals, vol. 41, no. 7, pp. 154–158,
2016.
[19] H. U. Jun, J. Chen, and L. I. Ming, “Research on Drawing
Technology of High-strength Steel Rectangular Box Based on
ABAQUS,” Hot Working Technology, 2014.
[20] S. Liu, H. Shi, and L. Feng, “Transformer harmonic loss model
considering skin effect and proximity effect,” Dianli Zidonghua
Shebei/Electric Power Automation Equipment, vol. 35, no. 3, pp.
133–139, 2015.
[21] Z. Zhao and S.-Y. Zhang, “Self-adaptive step sequential-
interaction method and its application in transient-fields cou-
pling problems,” Jisuanji Jicheng Zhizao Xitong/Computer Inte-
grated Manufacturing Systems, CIMS, vol. 17, no. 7, pp. 1404–
1414, 2011.
[22] Z. Wang, W. Yang, and Y. Song, “The R/S analysis and appli-
cation in the oil-gas gathering and transportation process,”
Petroleum Exploration & Development, vol. 37, no. 5, pp. 618–
622, 2010.
[23] C. Pandey, A. Giri, and M. M. Mahapatra, “Evolution of phases
in P91 steel in various heat treatment conditions and their
effect on microstructure stability and mechanical properties,”
Materials Science & Engineering A, vol. 664, pp. 58–74, 2016.
[24] S. Wang, N. Zhang, L. Wu, and Y. Wang, “Wind speed
forecasting based on the hybrid ensemble empirical mode
decomposition and GA-BP neural network method,” Journal of
Renewable Energy, vol. 94, pp. 629–636, 2016.
[25] Z. Sun, H. Yang, and Z. Tang, “Microstructural evolution model
of TA15 titanium alloy based on BP neural network method
and application in isothermal deformation,” Computational
Materials Science, vol. 50, no. 2, pp. 308–318, 2010.
[26] C. Bharathi, D. Rekha, and V. Vijayakumar, “Genetic Algorithm
Based Demand Side Management for Smart Grid,” Wireless
Personal Communications, vol. 93, no. 2, pp. 481–502, 2017.
[27] H. S. Madraswala and A. S. Deshpande, “Genetic algorithm
solution to unit commitment problem,” in Proceedings of the
2017 International Conference on Nascent Technologies in Engi-
neering, ICNTE 2017, India, January 2017.
[28] S. Pal, S. K. Pal, and A. K. Samantaray, “Artificial neural network
modeling of weld joint strength prediction of a pulsed metal
inert gas welding process using arc signals,” Journal of Materials
Processing Technology, vol. 202, no. 1-3, pp. 464–474, 2008.
[29] J. Tang, “Study on temperature control effect with pipe-cooling
of thick inverted T-shaped concrete pier-wall during construc-
tion,” Water Resources & Hydropower Engineering, 2017.
Advances in Advances in Journal of The Scientific Journal of
Operations Research
Hindawi
Decision Sciences
Hindawi
Applied Mathematics
Hindawi
World Journal
Hindawi Publishing Corporation
Probability and Statistics
Hindawi
www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 http://www.hindawi.com
www.hindawi.com Volume 2018
2013 www.hindawi.com Volume 2018

International
Journal of
Mathematics and
Mathematical
Sciences

Journal of

Hindawi
Optimization
Hindawi
www.hindawi.com Volume 2018 www.hindawi.com Volume 2018

Submit your manuscripts at

www.hindawi.com

International Journal of
Engineering International Journal of
Mathematics
Hindawi
Analysis
Hindawi
www.hindawi.com Volume 2018 www.hindawi.com Volume 2018

Journal of Advances in Mathematical Problems International Journal of Discrete Dynamics in

Complex Analysis
Hindawi
Numerical Analysis
Hindawi
in Engineering
Hindawi
Differential Equations
Hindawi
Nature and Society
Hindawi
www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018

International Journal of Journal of Journal of Abstract and Advances in

Stochastic Analysis
Hindawi
Mathematics
Hindawi
Function Spaces
Hindawi
Applied Analysis
Hindawi
Mathematical Physics
Hindawi
www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018