Original Article

Proc IMechE Part B:

J Engineering Manufacture
227(7) 1023–1031
A fast optimization approach for Ó IMechE 2013
Reprints and permissions:
multipass wire drawing processes sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0954405413481780

based on the analytical model pib.sagepub.com

Guang-Liang Zhang1, Zhang-Wei Wang2, Shi-Hong Zhang2, Ming Cheng2

and Hong-Wu Song2

Abstract
A fast optimization approach is described for design optimization of the multipass wire drawing process with genetic
algorithm. The analytical models for calculation of the drawing force, power consumption, and the temperature rise are
addressed by considering the real steel wire drawing condition, and a new die wear factor indicating die wear and life is
established based on the slab method. The genetic algorithm has been implemented to build up a single objective optimi-
zer for minimizing the total power consumption and a multiobjective optimizer for minimizing both the total power con-
sumption and temperature rise under several deliberate design constraints. The numerical examples show that the
multiobjective optimizer presents better performance in finding the optimal solution, compared with the single objective
optimizer. Different types of optimization methods especially designed for real demands are demonstrated under multi-
objective optimization set. Compared with a reference design, it indicates that significant improvements in the total
power consumption and the control of maximum temperature, delta factor, and die life have been achieved by the opti-
mization. It has been proved that the optimizer for the multipass wire drawing processes is highly effective and efficient.

Keywords
Optimization, genetic algorithm, wire drawing, analytical model, pass schedule

Date received: 1 September 2012; accepted: 12 February 2013

Introduction both large viscoplastic strains and friction effects and

the experimental validation of drawing forces under
In the wire drawing process, initial rod after prepara- different drawing velocities. Jo et al.3 and Lee et al.4
tion is drawn in several passes through conical dies calculated the temperature in a multipass wire drawing
where its diameter is reduced successively. Similar to operation by analytical model for high carbon steels.
most product manufacturing processes, steel wire draw- Singh et al.5 presented a comprehensive finite element
ing process involves several process parameters and model for multipass wire drawing operation consider-
product properties. Selection of the drawing process ing both deformation and thermal effects during draw-
parameters such as pass schedule (pass number/partial ing and cooling of wires on the drawing capstan.
reduction), speed, die angle, and back-pull dramatically Optimal selection of the process parameters com-
influences the qualities of the final products, heat gen- bined with optimal product properties guarantees the
eration, die wear, power consumption, and risk of success of the production. With genetic algorithm (GA)
fracture. based on finite element analysis, Roy et al.6 developed
For multipass wire drawing, plenty of work has been
done in the study of the whole processes using both
analytical method and finite element method. Joun and 1
School of Mechanical Engineering, Taizhou University, Taizhou, China
Hwang1 proposed a rigid–viscoplastic finite element 2
Institute of Metal Research, Chinese Academy of Sciences, Shenyang,
model to calculate the energy consumption, which was China
applied to the pass schedule design problems in multi-
Corresponding author:
pass axisymmetric extrusion and drawing. Celentanoa Guang-Liang Zhang, School of Mechanical Engineering, Taizhou University,
et al.2 were devoted to the simulation of multipass wire 1139 Shi Fu Road, Taizhou, 318000, China.
drawing via finite element analysis that accounts for Email: [email protected]

Downloaded from pib.sagepub.com at PENNSYLVANIA STATE UNIV on October 5, 2016

1024 Proc IMechE Part B: J Engineering Manufacture 227(7)

a design tool to minimize the difference between maxi- plastic deforming zone and stress analysis for wire
mum and minimum effective plastic strains in the final drawing. According to the slab method, a complex
products and the total deformation energy. Celano equation that permits the effects of friction, die angle,
et al.7 proposed an optimization approach consisting of and back-pull to be explicitly considered can be
simulated annealing (SA) algorithm and permitted to obtained9
the determination of both the optimal number of passes "  2B #  
and the optimal number of reduction rate sequence. slx 1+B D sq D 2B
= 1 + ð1Þ
Singh et al.8 developed an optimization method of mul- ss B D0 s s D0
tipass wire drawing by establishing a response surface
from finite element modeling. The optimization was where B = (f= tan a)  D0 and D1 are the wire diameters
carried out for energy, strain uniformity, temperature, at inlet and outlet of the die, respectively. ss is the aver-
and wire breakages during wire drawing. However, the age yield stress, sq is the back-pull tension during the
die wear was normally neglected. In addition, the opti- wire drawing processes, a is the semi-angle of the die,
mizers combined with finite element model must take and f is the Coulomb friction coefficient. When
many days to obtain an optimum result, which is time- D = D1 , the drawing stress sL can be obtained by
consuming and limits its application in industry, and  "  2B #  2B
the optimizers combined with a response surface based 1 D1 D1
sL = ss 1 + 1 + sq ð2Þ
on finite element modeling (FEM) are lack of accuracy B D0 D0
due to the difficulty in the establishment of an accurate
response surface. In the multipass wire drawing processes, the back-
Therefore, a fast approach for the optimization with pull is widely used to reduce the die load and thus
GA combined with analytical model is proposed. The enhance the die life. Eytelwein’s equation10 is used to
multipass drawing processes are analyzed using analyti- calculate the value of the back tension
cal methods, and the formulas used to calculate the
drawing force, power consumption, temperature rise, 1
sq = sL ð3Þ
and die wear are presented in our study. The aims of em2pm
the optimization are to minimize the power consump- where m is loops on capstans, sL is the drawing stress
tion and enhance the die life, besides keeping high qual- of previous pass, and m is the friction coefficient
ity of products. between the wire and capstan.
When the wire enters the reduction zone of a draw-
Theoretical analysis of the wire drawing ing die, material layers close to the surface would
undergo deformation because of the reduction in area
processes
and the change in flow direction. Consider the redun-
Calculation of drawing force dant deformation due to the internal distortion, an
The classic drawing force model based on slab method additional drawing stress s9L is added11
is well know and has been proven to be accurate for the 4ss
real application. Herein, based on the slab method, all s9L = pffiffiffi tan a ð4Þ
3 3
the factors according to the real condition in steel wire
drawing process, involving the homogenous deforma- In order to keep the properties and shape of the
tion, the redundant deformation, the back-pull, and the wires, the bearing zone is used in wire drawing pro-
bearing length, are considered. Figure 1 shows the cesses. The drawing stress would increase so as to

Figure 1. Plastic deformation and stress analysis for wire drawing.

Downloaded from pib.sagepub.com at PENNSYLVANIA STATE UNIV on October 5, 2016

Zhang et al. 1025

overcome the friction stress from the bearing zone. where V is drawing speed at the die outlet.
Somehow, traditionally, the effect of bearing zone is
neglected. Herein, the contribution of the friction on
the bearing zone to the drawing stress is calculated
Calculation of total power consumption
using an empirical formula The total work consumed at a drawing pass involves
three components: (a) the volumetric deformation
ld power (useful work required to reduce the cross sec-
sb = 0:1  f  ss ð5Þ
D1 tion), (b) the redundant deformation power (redundant
where ld is the bearing length. work required to change the flow direction), and (c) the
Taking equations (2)–(5) into consideration, the draw- friction power at the die–wire interface (work required
ing stress in the die can be expressed by equation (6) to overcome frictional resistance).
The calculated drawing force represents the total
( "  2B # )
1 D1 4 force required to overcome the volumetric deformation
sL = ss 1+ 1 + pffiffiffi tan a work, redundant work, and friction work. The power
B D0 3 3 consumption could be calculated by multiplied drawing
 2B force and drawing speed. The back-pull power should
D1 ld
+ sq + 0:1  f  ss ð6Þ be subtracted to obtain the effective power. The total
D0 D1
power is a summation of effective power required at
Then, the drawing force is easy to obtain by each pass, and it could be calculated by the following
equations
pD21 X  X 
F = sL  ð7Þ
4 WT = WFi  Wqi = sLi  sqi pD21i =4  V1i
ð12Þ

Calculation of die wear factor where sLi is the drawing stress at each pass, sqi is the
back-pull stress, V1i represents the drawing speed exit-
The die wear and die life are essentially important and
ing the die at each pass, and D1i is the wire diameter
must be considered for the wire drawing design and
exiting the die at each pass.
optimization. In this study, a factor evaluating the die
The mass flow rate of wire through the die keeps
wear and die life is developed in the function of the die
constant for each pass, and this means that the volu-
pressure and drawing speed, as they are the main fac-
metric rate of wire entering a die must be the same as
tors influencing the die wear and life. First, based on
that exiting. Once the take-up speed Vf and final wire
the slab method, a die pressure model is established.
diameter Df are known, the drawing speed at each pass
The die angle and friction coefficient are very small,
can be expressed as
and the approximate yield criterion is considered
V1i pD21i Vf pD2f
slx + snx = ss ð8Þ = ð13Þ
4 4
From equations (1) and (8), the die pressure could be
obtained
"  2B #  2B
Calculation of temperature rise
1 D 1 D It is essential to calculate the wire temperature rise,
tnx = ss 1 +   sq ð9Þ
B D0 B D0 since temperature control often proves to be the limit-
ing factor with respect to productivity in the multipass
Equation (9) shows that the die pressure takes the max- wire drawing processes. The heat generation during
imum value at the point where the wire initially con- drawing would also influence wire properties, lubrica-
tacts the die. Thus, maximum wear usually occurs at tion effectiveness, and accelerate die wear. The tem-
the entrance of the die. Upon integration, the average perature control could prevent fracture of wire and the
die pressure during drawing can be calculated tendency for embrittlement from strain aging. The heat
  effects are a result of volumetric deformation, redun-
1+B D2B + 1  D2B +1
1 1
s n = ss  0 1
 2B  dant deformation, and friction work. The volumetric
Bð2B + 1Þ D0  D1 D0 B
deformation energy, the redundant deformation energy,
1 D2B + 1  D12B + 1 1 and friction energy can be obtained from the following
sq  0  2B ð10Þ equations, respectively4
2B + 1 D0  D1 D0
 2 
Therefore, the die wear factor can be expressed in pD0 pD21
Ev =  s s V1 z ð14Þ
function of die pressure and drawing speed as 4 4
4 pD21
DF = sn  V ð11Þ Es = pffiffiffi ss tan aV1 z ð15Þ
3 3 4

Downloaded from pib.sagepub.com at PENNSYLVANIA STATE UNIV on October 5, 2016

1026 Proc IMechE Part B: J Engineering Manufacture 227(7)

pD20 pD21 Table 1. Boundary conditions.

4  4 ld
Ef =  fss V1 z + 0:1  f  s s V1 z ð16Þ
sin a D1 Input variables Lower Upper Data
bound bound format
where z is the deformation time of the wire.
The temperature rise could be calculated on the basis TNP — 30 1, 2, ., 29, 30
of thermal equilibrium equation Partial reduction, r (%) 5 20 Continuous
DA 2a (°) 6 12 6, 7, ., 11, 12
  2  Ratio of bearing 5 40 5,10, ., 35,40
1 pD0 pD21
DT = pD2 3 A  ss + length-to-wire
4 4
4  rc
1 diameter (%)
Loops on capstan 0 4.5 0, 1.5, 2.5, 3.5, 4.5
pD20 pD21
4 pD21 4  4
pffiffiffi Ass  tan a + m  TNP: total number of pass; DA: die angle.
3 3 4 sin a

ld
fss V1 + 0:1  mf  ss ð17Þ instead of wire diameter, as the partial reduction of
D1
individual pass is independent. Similarly, in order to
where A is the mechanical equivalent of heat. In this avoid the pass dependence, the bearing length is charac-
study, about 55% of mechanical energy in the deforma- terized using the ratio of bearing length-to-outlet wire
tion energy is transferred into heat during the drawing diameter, named as the bearing length ratio. Finally,
process.12 DT is the temperature rise, r is the density of the design parameters involving the total number of
the wire, c is the specific heat of the wire, and m is the passes, partial reduction, die angle, bearing length
heat partition coefficient. The heat partition coefficient ratio, and loops in capstan are selected as the input
means that the quantity of heat that is transferred to design variables of the optimization, which are listed in
the wire due to friction at the interface, and Jo et al.3 Table 1.
proposed that about 60% of the heat generated by fric- It is very practical to limit the design variables vary-
tion is transferred to the wire and the rest is transferred ing in the defined bounds, such as a case study listed in
to the die in wet wire drawing. Table 1, where the maximum number of allowable
passes is 30, the partial reduction is limited between 5%
and 20%, the die angle varies between 6° and 12° with
Calculation of delta factor (D-factor)
a step of 1, the bearing length ratio ranges from 5% to
In wire drawing, the geometry of the working part of a 40% with a step of 5%, and the loops on capstans take
die is a critical factor. This geometry can be defined by five values. Based on the practical die manufacturing
the delta factor, which is the ratio of the circular arc and wire drawing operation, a step data format is used
spanning the midpoints of the die face to the length of to control the change in the die angle, bearing length
contact between wire and die.13 For conical dies, the ratio, and loops on capstan.
D-factor is
 
D0 + D1
D=a ð18Þ Objective functions and constraints
D0  D1
The objectives are set considering the following aspects,
Low D values mean small die angle and higher reduc- which are highly interlinked:
tion in area, which results in high levels of friction work
due to longer contact length. While higher values of D 1. The ratio of drawing stress-to-yield stress (DS/YS).
represent large die angle and lower reduction in area, Any plastic deformation taking place outside of
which indicate more redundant deformation because of the die resulting from the drawing force is unallow-
excessive direction change during wire drawing. able. Thus, the ratio of DS/YS must be less than 1,
in order to avoid the necking during wire drawing.
Optimization approach 2. Maximum die wear factor. The die pressure and
drawing speed significantly influence die wear. The
Design variables maximum die wear factor among all the passes is a
The first step in the design optimization study is the indicator evaluating the shortest die life of a die
selection of variables. The main parameters in the wire set.
drawing process involve the inlet and outlet wire dia- 3. Total power consumption in the drawing process.
meters, die angle, and bearing length of each pass. It is The total power is a sum of power consumption
dependent on the die diameter between two neighbor per pass.
passes, where the outlet wire diameter must be smaller 4. Maximum temperature rise. As discussed earlier,
than that of its previous pass. Therefore, because of this thermal effect is particularly important for wire
dependence, it is a wise option to select the partial drawing processes. If the wire surface temperature
reduction as the design variable for the optimization excesses a desired value, it would adversely

Downloaded from pib.sagepub.com at PENNSYLVANIA STATE UNIV on October 5, 2016

Zhang et al. 1027

deteriorate the wire properties and lubricant and Optimization design

decrease the die life. Figure 2 presents the procedure of the optimization
5. Delta factor for each pass. Higher values of delta
design for a multipass wire drawing process. With giv-
factor are indicative of more inhomogeneous defor- ing initial and final wire diameter together with wire
mation and redundant deformation, leading to an material properties and setting the limits for the design
increase in power consumption.
variables as well as the constraints for the objective
variables, the optimizer is going to find an optimum
Optimization objectives in wire drawing include the solution with the best pass schedule. Both the single
desired product properties, enhanced die life, and objective and multiobjective optimizers have been
reduced power consumption. An optimal design prob- implemented as a MATLAB toolbox.
lem can be generally expressed as

minmize F½f1 ðxÞ, f2 ðxÞ, . . . , fn ðxÞ x e H Results and discussion

subject to hi ðxÞ40 i = 1, 2, 3, . . . , m Optimization reference
Lee et al.’s4 study on optimization of temperature to
where F(x) contains objectives to be minimized, and x
increase the drawing speed from 1100 to 2000 m/min
represents the design variables to be optimized. H is the
for wet wire drawing of low carbon steel is selected as a
design space that sets the bounds for the design vari-
reference for evaluating the performance of the optimi-
ables, hi (x) are the design constraints, and m is the total
zers. In Lee et al.’s4 study, in order to reduce the tem-
number of constraints. In this article, a single objective
perature rise, it adapted a straight–taper reduction rate
optimizer and a multiobjective optimizer are con-
alignment and increased the number of passes from 24
structed. For the single objective optimizer, total power
to 29, but it did not take the risk of fracture (delta fac-
consumption is used as the objective, and the maximum
tor), die wear (die wear factor), and power consumption
temperature rise is set as a constraint. While the multi- into consideration. A wire drawing design or optimiza-
objective optimizer is employed, two objectives, mini- tion must be able to minimize one or two objectives but
mizing the total power consumption and the maximum keep others in the allowable levels, by which the opti-
temperature rise, are concerned simultaneously. Some mized pass schedule can fit for use. Taking 24 and 29
general constraints are set for the optimization. passes with a drawing speed of 2000 m/min of Lee
et al.’s4 study as the references, their results in maxi-
mum ratio of DS/YS, maximum die wear factor, maxi-
Optimization algorithm mum temperature rise, and total power consumptions
The GA is utilized as the optimization algorithm in this are listed in Table 2. Based on the results of 24 and 29
article. The GA works on a population of potential passes, the constraints for the single objective optimizer
solutions using the survival of the fittest strategy to and multiobjective optimizer are also listed in Table 2,
improve successive generations, which involves several taking the larger values of ratio of DS/YS, die wear fac-
distinct operations such as selection, crossover, and tor, and delta factor, and the smaller temperature rise
mutation. The population type of design variables is between 24 and 29 passes as the upper limits for the
double vector. The selection function is tournament, optimization.
the crossover function is intermediate, and the muta- Four cases are designed to evaluate the performance
tion function is adaptive feasible. of the optimizers, as shown in Table 3. In Case 1, it
minimizes the power consumption with the single

Figure 2. Optimization design procedure of the multipass wire drawing processes.

Downloaded from pib.sagepub.com at PENNSYLVANIA STATE UNIV on October 5, 2016

1028 Proc IMechE Part B: J Engineering Manufacture 227(7)

Table 2. Optimization reference and design of optimizers.

Objective function 24 Passes 29 Passes Single objective optimizer Multiobjective optimizer

Ratio of drawing stress-to-yield stress 0.367 0.324 \0.367 \0.367

Maximum die wear factor (MN/(mm s)) 80.458 85.800 \85.500 \85.500
Total power consumption (W) 13695 14678 Minimize Minimize
Maximum temperature rise (°C) 207 174 \174 Minimize
Delta factor 2.492 4.225 \4.225 \4.225

Table 3. Optimization designs for case study.

Name Optimizer Die angle BLRa Loops on capstan (m) Number of passes

Case 1 Single objective optimizer 6°\ 2a\12° 5%\BLR\40% 0\m\4.5 \30

Case 2 Multiobjective optimizer 6°\ 2a\12° 5%\BLR\40% 0\m\4.5 \30
Case 3 Multiobjective optimizer 2a = 10° BLR = 25% m = 3.5 \30
Case 4 Multiobjective optimizer 2a = 10° BLR = 25% m = 3.5 29

BLR: bearing length ratio.

a
BLR = bearing length/outlet wire diameter.

Table 4. Optimization results.

Name Power (W) T (°C) Die wear factor (MN/ (mm s)) Delta factor DS/YSa

Case 1 14,475 174 82.438 4.244 0.365

Case 2 14,456 171 79.552 3.070 0.363
Case 3 14,497 170 79.798 3.684 0.366
Case 4 14,589 170 79.725 4.165 0.392
Reference of 29 passes 14,678 174 85.800 4.226 0.324

DS/YS: drawing stress–to–yield stress ratio.

a
Ratio of DS/YS.

fixes the die angle and bearing length ratio and loops
on capstan, while keeping the number of passes free.
Based on Case 3, Case 4 fixes the total number of
passes as 29.

Comparison of single objective optimizer and

multiobjective optimizer
The optimization results including power consumption,
maximum temperature rise, die wear factor, delta fac-
tor, and ratio of DS/YS are listed in Table 4 as well as
the reference of 29 passes. Compared with the reference
of 29 passes, optimizations based on these four cases
reduce the power consumption by about 200 W, while
the other factors including temperature rise, die wear
factor, and delta factor are smaller than that in refer-
Figure 3. Temperature rise at the reference and the optimized ence, and a little increase in DS/YS will not lead to the
pass schedules for Cases 1 and 2. necking problem as they are less than 0.37. Therefore,
it can be concluded that both the single objective opti-
mizer and multiobjective optimizer have very good per-
objective optimizer, while setting the limits for the formance in finding a better pass schedule compared
design variables and objective functions, as shown in with the reference.
Tables 2 and 3, under the formats listed in Table 1. Figures 3 and 4 show the temperature rises and
Case 2 minimizes both the power consumption and reduction rate alignments optimized by the single objec-
temperature rise, while keeping the constraints the tive optimizer (Case 1) and multiobjective optimizer
same as that in Case 1. Case 3 is based on Case 2, it (Case 2), respectively. It indicates that the optimized

Downloaded from pib.sagepub.com at PENNSYLVANIA STATE UNIV on October 5, 2016

Zhang et al. 1029

Figure 4. Pass schedules at the reference and the optimization Figure 6. Pass schedules at the reference and the optimization
in Cases 1 and 2. in Cases 2 and 3.

sense. As can be observed from the comparison, results

obtained based on the single objective optimizations
are contained in those obtained according to the multi-
objective optimization. The multiobjective optimization
inherits all the favorable properties of the single objec-
tive optimizations and thus shows its priority on
application.

Multiobjective optimization with and without fixed die

angle and bearing length
In the real production, in order to facilitate the die
manufacturing and operation, the die angle, bearing
length ratio, and loops on capstan are fixed as the same
for all the passes. Here, Case 3 is an example to fix
these three parameters, while optimizing the partial
Figure 5. Temperature rise at the reference and the optimized reduction rates and number of passes to minimize both
pass schedules for Cases 2 and 3. the temperature rise and the power consumption, with
multiobjective optimizer. From Table 4, the results in
Case 3 are close to that in Case 2, except that Case 3
pass schedules between Cases 1 and 2 are very similar has a slight higher value in delta factor. As smaller
as a shape of increasing–decreasing with fluctuating in delta factor contributes to lower risk of fracture; there-
the middle passes. Compared with the reference pass fore, the pass schedule obtained in Case 2 may be bet-
schedule, the reduction rates of the final 10 passes ter than that in Case 3. The reason leading to the lower
increase, while the inlet passes have smaller reduction value of delta factor is that die angle of Case 3 is fixed
rate. Temperature results show that the optimized pass at 10°, while the die angle of Case 2 is varying from 6°
schedule presents the same temperature rise in the final to 11°. Optimization will find a smaller die angle in
10 passes, which can meet the requirements of high- order to reduce the delta factor, which is a constraint in
speed wire drawing. the optimization.
Comparing single objective optimizer (Case 1) and Figures 5 and 6 plot the optimized temperature rise
multiobjective optimizer (Case 2), it can be found that results and optimized pass schedules in Cases 2 and 3,
the multiobjective optimizer presents a better perfor- respectively, as well as the reference results. In Figure 5,
mance, where all the values in the power consumption, compared with the reference, it is clear that Cases 2 and
temperature rise, delta factor, and DS/YS are smaller 3 present a more uniform temperature rise along the
than that in the single optimizer. The multiobjective passes. As it can be found from Figure 6, in Case 3, a
optimizer is able to find a global optimum result since near straight–taper profile in pass schedule is obtained,
it gets a set of solutions that are optimal in an overall which is quite similar to the reference pass schedule. In

Downloaded from pib.sagepub.com at PENNSYLVANIA STATE UNIV on October 5, 2016

1030 Proc IMechE Part B: J Engineering Manufacture 227(7)

Figure 7. Temperature rise at the reference and the optimized Figure 8. Pass schedules at the reference and the
pass schedules for Cases 2–4. optimization in Cases 2–4.

Case 3, the optimized pass schedule has 28 passes,

which is different with the reference, and its taper slope may take 1 h to finish one iterative generation under
is smaller than that in the reference. the same condition, and thus, a total number of 100–
200 h are needed to accomplish the final results.
Multiobjective optimization with and without fixed Therefore, it is firmly expected that this new approach
number of passes will be used as a solid basis for process design and con-
trol in wire drawing.
Optimization under fixed number of passes is also con-
ducted with multiobjective optimizer in Case 4, as
shown in Table 3. The optimized temperature rise and Conclusions
pass schedules with 29 passes are plotted in Figures 7
and 8, respectively. Similar to Case 3, the temperature In this study, a reliable analytical model employed to
rises of the final 10 passes approach to 170 °C. An calculate the drawing force, die pressure, power con-
increasing–decreasing pass schedule is obtained in Case sumption, and temperature rise in the multipass wire
4, as shown in Figure 8. From Table 4, it can be found drawing processes is proposed. Based on this analytical
that Case 4 presents higher value in power consump- model, a new approach to optimal design of the multi-
tion and delta factor, compared with Cases 2 and 3. It pass wire drawing processes with GAs is presented. In
means that the restriction in number of passes will limit this approach, the total number of passes, partial reduc-
the searching scope of optimization, which will lead to tion, die angle, bearing length, and loops on capstan are
the difficulty in finding a global optimum solution. selected as design variable to be optimized. The optimi-
zation is conducted under various design constraints.
Limitations are set on the maximum drawing stress,
Performance of the optimization approach maximum die wear factor, and delta factor. The mini-
Currently, trial-and-error methods are widely used to mum of total power consumption and the minimum of
design the multipass wire drawing processes in the maximum temperature rise are used as goals for optimi-
industrial practice. Depending on their own experience zation. Both single objective and multiobjective optimi-
and historical data, the designers attempt to arrive at a zation are chosen and compared. The multiobjective
desirable solution after numerous experiments. optimization has been proved to enjoy more benefits,
However, our new approach, optimization with GA, and thus, it is in preference to the single objective opti-
provides an extremely cost saving and high-efficiency mization. Based on the multiobjective optimization, dif-
way to design the drawing processes. It required less ferent kinds of optimization methods designed for
than 100 iterative generations for single objective opti- actual requirements are compared and discussed. It is
mization and about 200 iterative generations for multi- shown that the optimal solutions are much better than
objective optimization to obtain optimal solutions. The the reference design, in terms of the total power con-
whole optimization process could be accomplished on sumption, delta factor, heat generation, and die wear.
a PC (CPU is Quad 3.3 GHz Intel Pentium III and A significant merit of this approach is its capability in
Memory is 3242M) within less than 3 min. For the optimizing the multipass drawing processes in a very
optimization process based on finite element model, it short time (less than 3 min in a normal PC).

Downloaded from pib.sagepub.com at PENNSYLVANIA STATE UNIV on October 5, 2016

Zhang et al. 1031

Funding for 0.72 wt% C steel wire. J Mater Process Tech 2010;
210: 776–783.
This study was financially supported by NV Bekaert.
5. Singh SK, Gautham BP, Goyal S, et al. Development of
a virtual wire drawing tool for process analysis and opti-
Acknowledgement mization. Wire J Int 2007; 40: 72–78.
6. Roy S, Ghosh S and Shivpuri R. Optimal design of pro-
The authors thank NV Bekaert for providing the pearli-
cess variables in multi-pass wire drawing by genetic algo-
tic steel wires and computers and also thank Ali Sarban
rithms. J Manuf Sci E: T ASME 1996; 118: 244–251.
from Bekaert Technology Center for his efforts and 7. Celano G, Fichera S, Fratini L, et al. The application
time in preparing the article. of AI techniques in the optimal design of multi-pass
cold drawing processes. J Mater Process Tech 2001;
Conflict of interest 113: 680–685.
8. Singh SK, Gautham BP, Goyal S, et al. Optimization of
The authors declare that there is no conflict of interest. multi-pass steel wire drawing operation. Wire J Int 2009;
42: 82–88.
References 9. Sachs OA and Hoffman G. Introduction to the theory of
plasticity for engineers. New York: McGraw-Hill Com-
1. Joun MS and Hwang SM. Pass schedule optimal design pany, 1953.
in multi-pass extrusion and drawing by finite element 10. Attaway SW. The mechanics of friction in rope rescue.
method. Int J Mach Tool Manu 1993; 33: 713–724. In: International technical rescue symposium. Fort Collins,
2. Celentanoa DJ, Palaciosb MA, Rojasb EL, et al. Simula- USA, 5–7 November, paper 2: 1–16. La Fayette, GA:
tion and experimental validation of multiple-step wire ITRS, 1999.
drawing processes. Finite Elem Anal Des 2009; 45: 163– 11. Wen JL. Theory of metal extrusion and drawing pro-
180.
cesses. Shenyang: Northeastern University, 2003; p.178.
3. Jo HH, Lee SK, Kim MA, et al. Pass schedule design
12. Macdougall D. Determination of the plastic work con-
system in the dry wire-drawing process of high carbon
verted to heat using radiometry. Exp Mech 2000; 40:
steel. Proc IMechE, Part B: J Engineering Manufacture
298–306.
2002; 216: 365–373.
13. Wistreich JG. The fundamentals of wire drawing. Metall
4. Lee SK, Lee SB and Kim BM. Process design of multi-
Rev 1958; 3: 97–141.
stage wet wire drawing for improving the drawing speed

Downloaded from pib.sagepub.com at PENNSYLVANIA STATE UNIV on October 5, 2016