This article was downloaded by: [Northeastern University]

On: 03 December 2014, At: 06:56

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Vehicle System Dynamics: International

Journal of Vehicle Mechanics and
Mobility
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/nvsd20

Dynamic analysis and design

optimisation of the seat belt
pretensioner
a b a
Sung Pil Jung , Tae Won Park & Chan Seung Park
a
Graduate School of Mechanical Engineering , Ajou University ,
Suwon, Republic of Korea
b
Department of Mechanical Engineering , Ajou University ,
Suwon, Republic of Korea
Published online: 21 Jan 2010.

To cite this article: Sung Pil Jung , Tae Won Park & Chan Seung Park (2010) Dynamic analysis and
design optimisation of the seat belt pretensioner, Vehicle System Dynamics: International Journal of
Vehicle Mechanics and Mobility, 48:1, 65-78, DOI: 10.1080/00423110903259519

To link to this article: http://dx.doi.org/10.1080/00423110903259519

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
 Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Downloaded by [Northeastern University] at 06:56 03 December 2014
 Vehicle System Dynamics
Vol. 48, No. 1, January 2010, 65–78

Dynamic analysis and design optimisation of the

seat belt pretensioner
Sung Pil Junga , Tae Won Parkb * and Chan Seung Parka
Downloaded by [Northeastern University] at 06:56 03 December 2014

a Graduate School of Mechanical Engineering, Ajou University, Suwon,

Republic of Korea; b Department of Mechanical Engineering,
Ajou University, Suwon, Republic of Korea

(Received 15 December 2008; final version received 12 August 2009 )

In this paper, the dynamic model of the high-power pretensioner used for the seat belt of a passenger
vehicle is developed. The operating device of the pretensioner is optimised to improve the length of
the web retraction. The three-dimensional rigid dynamic model of the pretensioner is created using the
commercial multi-body programme, ADAMS, and the nonlinear spring element is used to represent
the seat belt. The reliability of the dynamic model is validated by comparing the simulation results
with the actual test results. The pretensioner presented in this paper is operated by the rack and pinion
gear mechanism. The profile of the rack and pinion is optimised to improve the retracting performance
of the pretensioner using the design of experiments. The optimised rack and pinion gear is applied to
the dynamic model, and the optimisation results are verified by comparing the simulation results of
the original and the optimised model.

Keywords: high-power pretensioner; seat belt; dynamic model; web retraction; rack and pinion gear;
design of experiments

1. Introduction

The pretensioner is used to retract the belt webbing and tighten up any slack in the event of a
crash. The retracting force of the pretensioner helps to move the passenger into the optimum
crash position in his or her seat. The pretensioners are wired to the same control processor that
activates the car’s air bags. When an impact is detected, the processor activates the pretensioner
and then the airbag. The pretensioner presented in this paper uses the gas explosion pressure
of the propellant as a power source. So, it is called the pyro-typed high-power pretensioner or
high-power pretensioner. That is, a gas blast pressure is generated by exploding the propellant
right after a crash, and the gas blast is transferred to the retractor through a gear mechanism,
and then the retractor winds the webbing.
There are some restrictions on the performance of the pretensioner. The entire pretensioning
procedure must be completed in less than or equal to 12 ms, and the amount of web retraction

*Corresponding author. Email: park@ajou.ac.kr

ISSN 0042-3114 print/ISSN 1744-5159 online

© 2010 Taylor & Francis
DOI: 10.1080/00423110903259519
http://www.informaworld.com
 66 S.P. Jung et al.

should be greater than or equal to 120 mm. Although the pretensioner is developed for the
safety of the passenger, excessive web retraction may seriously injure the passenger. Thus,
most automobile companies restrict the maximum belt load to be less than 2.5–3 kN.
Since the main function of the pretensioner is to retract the webbing, the amount of the web
retraction is the most important index that represents the performance of the pretensioner. But
based on Breed’s evaluation tests of several compact seat belt pretensioners, the requirement
of 120 mm web retraction is very difficult to achieve [1]. Therefore, it is essential to evaluate
the performance and efficiency of the pretensioner, starting from the development stage. Many
studies have been done to estimate the performance of the pretensioner. Simon derived the
numerical relationship between the belt load and web retraction as a simple nonlinear equation
using the experiment data [1]. However, this equation is too simple to include the effects of
several important factors on the web retraction. Lee carried out a dynamic and structural
Downloaded by [Northeastern University] at 06:56 03 December 2014

analysis on the pretensioning system and reflected the results to a design [2]. Jung verified
the structural safety of the retractor using the finite element analysis method [3]. The above
two studies evaluated the performance of the pretensioner using the computer aided analysis
method (CAE). However, these studies neglected the friction between the components and
did not consider the reaction force of the dummy. Thus, there is a limitation to estimate the
amount of the web retraction exactly. Recently, a motorised seat belt pretensioner, a new
type of pretensioners, is developed. Min has developed the motorised seat belt pretensioner
and proved its robustness by performing the crash experiment [4]. Park evaluated the locking
mechanism of the motorised pretensioner based on the crash scenarios [5]. Although many
researches on the motorised pretensioner have been progressed, the motor is used just to assist
in retracting the belt webbing, and the pyrotechnic mechanism is still used as a power source.
So, it is very important to modify the design of the pyrotechnic mechanism and improve the
retracting performance of the pretensioner.
In this study, the multi-body dynamic model of the pretensioner is developed. The three-
dimensional dynamic simulation model of the pretensioner is created using ADAMS, a
commercial multi-body dynamic programme. The nonlinear spring element is applied to the
simulation model to represent the reaction force of the dummy and the belt webbing. The
actual tests are performed to validate the reliability of the simulation model. The pretensioner
presented in this paper uses the rack and pinion gear mechanism to transform the gas blast
energy into the rotational kinetic energy of the retractor. The profile of the rack and pin-
ion is optimised to improve the retracting performance of the pretensioner using the design
of experiments. The optimised rack and pinion gear is applied to the dynamic model, and
the optimisation results are verified by comparing the simulation results of the original and
optimised models.

2. Operating mechanism and experiment of the pretensioner

2.1. Operation mechanism of the pretensioner

Figure 1 shows the operating mechanism of the pretensioner. When an electric signal of a
crash is produced, the gas generator is activated, and then the propellant is exploded. The gas
blast is transferred to the rack through the manifold. The rack moves straightly and rotates
the pinion. Three clutches are inserted in the pinion. When the angular velocity of the pinion
is high enough, the clutches are projected and combine the pinion with the spool. Then, the
belt webbing rolled around the spool is retracted. If the pretensioner is deactivated, the pinion
has no effect on the movement of the spool. On the other hand, a torsional spring is equipped
at the other side of the spool. Thus, the webbing is retracted automatically by the torsional
 Vehicle System Dynamics 67
Downloaded by [Northeastern University] at 06:56 03 December 2014

Figure 1. Operating mechanism of the pretensioner.

spring even though the pretensioner is not operated. When the pretensioner is activated, the
torsional spring helps the pretensioner to retract the webbing; however, the effect is very small
about 1%.

2.2. Experiment of a pretensioner

Figure 2 shows the test configuration to check the performance of the pretensioner. Figure 3
shows the gas pressure curve with respect to the time measured in 10 cc volume. The maxi-
mum pressure is 65 MPa. Figure 4 shows the test results obtained when the propellant whose
maximum gas blast pressure is 65 MPa is exploded. In Figure 4, the solid line represents the

Figure 2. Test configuration of the pretensioner.

 68 S.P. Jung et al.
Downloaded by [Northeastern University] at 06:56 03 December 2014

Figure 3. 65 MPa gas pressure curve.

Figure 4. Experiment data at 65 MPa.

web retraction and the dotted line represents the belt load. In the figure, the maximum web
retraction and belt load are about 85 mm and 1.8 kN, respectively.

3. Dynamic modelling and verification

3.1. The clutch model

The pretensioner presented in this paper is operated by the rack and pinion gear mechanism.
As prescribed in the preceding section, the clutches in the pinion are projected and connect the
pinion to the spool only when the angular velocity of the pinion is greater than a certain velocity.
Figure 5 shows how the pinion is installed inside of the pinion. As shown in the figure, there
are three clutches inside of the pinion and constrained by the O-ring. The projection velocity
of the clutch is controlled by the stiffness of the O-ring. In other words, if the stiffness of
the O-ring is too low, the clutch is not fixed on the pinion even if the pinion is not rotated.
So, the pinion may be connected with the spool and interrupt the spool to rotate freely. On
the other hand, if the stiffness of the pinion is too high, the clutch may be fully fixed on the
pinion and not projected. So, the pinion is not connected with the spool, and the belt webbing
is not retracted. Therefore, the stiffness of the O-ring is an important factor to determine the
retracting performance of the pretensioner. In order to find the projection time of the clutches
precisely, the angular velocity of the pinion is measured when the clutches are projected.
 Vehicle System Dynamics 69
Downloaded by [Northeastern University] at 06:56 03 December 2014

Figure 5. Pinion and clutch assembly.

Figure 6. The measurement system of the moment of the projection of the clutches.

Figure 6 shows the experiment device to measure the angular velocity of the pinion. The
pinion including clutches is connected to the DC motor, and the DC motor is driven by the
motor driver. The RPM metre shows the angular velocity of the motor, and the motor driver is
connected to the computer through the Data Acquisition (DAQ) board. The high-speed camera
is used to check whether the clutch is projected or not. As a result of several experiments, the
clutches are projected when the angular velocity of the motor is 190 rpm.
Figure 7 shows the dynamic model of the pinion and clutch assembly. The clutches are
connected to the pinion by the revolute joints. To describe the effect of the O-ring, the spring
elements are used between the clutches and the pinion. The projection moment of the clutches
is controlled by adjusting the stiffness of the spring element. To verify the dynamic model,
the spool is added to the model, and the contact condition is applied between the clutches
 70 S.P. Jung et al.
Downloaded by [Northeastern University] at 06:56 03 December 2014

Figure 7. Dynamic model of the pinion and clutch assembly.

Figure 8. Angular velocity of the pinion in the simulation model.

and spool. Torque is applied to the rotational centre of the pinion in the form of the second-
order polynomial function about time. To check whether the angular acceleration of the pinion
affects on the projection moment of the clutches, several simulations were done with varying
the coefficient of the second-order term of the torque. Figure 8 shows the angular velocity
profiles of the pinion obtained from dynamic simulations. In the figure, the angular velocity
of the pinion reaches its peak at 190 rpm regardless of the magnitude of torque and then
decreases sharply. At this moment, the clutches are projected and the pinion is combined
with the spool. Therefore, the dynamic model of the pinion and clutch assembly is created
successfully.

3.2. The seat belt model

To estimate the amount of the web retraction, the characteristic of the seat belt should be
taken into account. The seat belt contacts and slides on the dummy. Thus, the reaction force
between the belt and dummy determines the axial load of the belt. The seat belt, however, is
very difficult to model since the material property is nonlinear and its geometry is irregularly
 Vehicle System Dynamics 71

varied according to the surface shape of the dummy. Thus, in this study, the seat belt is modelled
using the nonlinear spring as Equation (1) on the basis of Zhu’s research [6]. In the actual test,
the axial tension of the belt is measured using the load cell; however, the inertial effects of the
load cell are neglected in this work.

F = K1 δ + K2 δ 3 + K3 δ δ̇, (1)

where δ is the displacement and δ̇ the velocity of the seat belt, respectively. K1 , K2 and K3
are coefficients estimated by the correlation with the test results.
Downloaded by [Northeastern University] at 06:56 03 December 2014

Figure 9. The three-dimensional dynamic model of the pretensioner.

Figure 10. Comparison of the test and simulation results at 65 MPa.

 72 S.P. Jung et al.

3.3. The entire dynamic simulation model

Figure 9 shows the three-dimensional dynamic model of the pretensioner. The gas pressure
curve shown in Figure 3 was used as the input force. The net force applied to the rack is changed
by the rotational reaction force of the pinion and the resist force caused by the belt model. The
thermal effect of the gas blast was not considered since the whole pretensioning process is com-
pleted in very short time, 12 ms. The thermodynamic behaviour is approximated by the pressure
curve measured in a constant volume. The contact conditions [7] are applied between the rack
and pinion, the clutches and spool and the rack and spool. In the real system, a rubber packing
is equipped behind the rack to prevent the gas leak. Thus, the coulomb friction condition is
applied between the rear part of the rack and manifold to consider the friction effect between
the manifold and the rubber packing. In addition, there is a torsional spring at the other side of
Downloaded by [Northeastern University] at 06:56 03 December 2014

the spool in the real system to wind the belt webbing when the pretensioner is not activated.
Thus, the torsional spring is applied at the spool in the simulation model. The nonlinear spring
that represents the seat belt is extended according to the rotation of the spool. The displacement
and force of the spring represent the web retraction length and belt load, respectively.

3.4. Model verification

Figure 10(a) and (b) shows the simulation and test results when the maximum gas blast
pressure is 65 MPa. In the figures, the dotted line is the test result and the solid line is the

Figure 11. Comparison of the test and simulation results at 80 MPa.

 Vehicle System Dynamics 73

simulation result. The simulation results are nearly the same as the test results. To validate the
reliability and robustness of the simulation model, the simulation and test are performed when
the maximum gas blast pressure is increased up to 80 MPa. Figure 11(a) and (b) shows the
results. In the figures, the dotted line is the test result and the solid line is the simulation result.
As shown in the figures, the simulation results follow the experiment result well enough, and
thus the dynamic model is verified.

4. The profile optimisation of the rack and pinion gear

The rack and pinion gear is another important element to be considered to improve the perfor-
mance of the pretensioner. The gas explosion energy is transformed to the rotational kinetic
Downloaded by [Northeastern University] at 06:56 03 December 2014

energy of the spool through the rack and pinion gear. So, the transmission efficiency of the gear
pairs affects significantly the amount of the web retraction. In order to design the optimum pro-
file of the gear pairs, the rack and pinion design programme is created using MATLAB. Then,
the optimum design of the rack and pinion gear is found using the design of experiments [8].

4.1. The rack and pinion gear design programme

Design parameters of the rack and pinion gear are the pressure angle (ϕ), module (m), number
of teeth (N ), profile correction coefficient (G), addendum coefficient (α), dedendum coefficient
(β), tip radius coefficient (r), tooth thickness (t) and distance between centres of gear pairs
(L) [9,10]. Figure 12 shows the rack and pinion gear created using the developed programme.
As shown in the figure, not only the shape of the rack and pinion gear but also the tooth root
circle, tooth tip circle and pitch circle are designed exactly.

4.2. Profile optimisation using design of experiments

The profile of the rack and pinion gear is created using the developed programme and applied
to the simulation model. In order to design the optimum profile of the gear pairs, design
parameters are selected as shown in Table 1. The number of teeth and module are chosen

Figure 12. The rack and pinion gear.

 74 S.P. Jung et al.

Table 1. Design variables.

Factor Level Value

Number of teeth and module (N × m) A 0 8 × 1.75

1 8×2
2 9 × 1.75
3 10 × 1.5
Pressure angle (◦ ) B 0 20
1 25
Tip radius coefficient C 0 0.25
1 0.3
Downloaded by [Northeastern University] at 06:56 03 December 2014

to make the tooth tip radius less than or equal to 15 mm since the radius of the pinion
housing where the clutches are installed is 15.1 mm. The generally used values in the gear
industry [11] are chosen for the values of the pressure angle and tip radius coefficient to
consider the machinability. If the gas blast pressure is too high or the power transmission is
not processed smoothly, the tooth of the gear may be deformed. In order to keep the gear
tooth from being deformed or damaged, the maximum bending stress and surface stress are
calculated analytically, and then minimised. The maximum bending stress generated in the

Figure 13. Force relation applied on the gear tooth.

Table 2. Orthogonal array and stress calculation results.

Maximum Maximum
bending surface
Number N m φ r G stress stress

1 9 1.75 20 0.25 0.56 −5.27E5 8.83E4

2 9 1.75 25 0.3 0.25 2.32E6 7.72E4
3 8 1.75 20 0.25 0.62 −1.97E6 9.10E4
4 8 1.75 25 0.3 0.36 −3.22E6 8.06E4
5 10 1.5 20 0.3 0.47 3.07E6 9.14E4
6 10 1.5 25 0.25 0.21 2.59E6 8.22E4
7 8 2 20 0.3 0.59 −2.03E7 8.38E4
8 8 2 25 0.25 0.39 −2.40E6 7.64E4
 Vehicle System Dynamics 75

tooth root is calculated using the Lewis equation [12] and the form is

6Fn cos(φ  )H
σb = , (2)
bS 2 cos(φ)

where Fn is the applied load at the tooth tip, φ  the angle between the tangent line of the tooth
tip circle and the line of Fn , φ the pressure angle, S the smallest thickness of the tooth root
and H the distance from the measurement point of S to the intersection point between the
vertical line of the tooth and the line of Fn (Figure 13). According to the Hertz equation, the
Downloaded by [Northeastern University] at 06:56 03 December 2014

Figure 14. The simulation results.

 76 S.P. Jung et al.

maximum surface stress is written as [12]


F (1/R1 + 1/R2 )
σn = 0.35 , (3)
t (1/E1 + 1/E2 )

where F is the vertical load applied on the tooth surface, t the tooth thickness, R1 and R2 the
radius of the meshed gear pairs and E1 and E2 the elastic coefficient of two gears. As shown
in Table 1, the level of the design parameter A is 4, B and C are 2. Thus, 4 × 2 × 2 orthogonal
array [13,14] shown in Table 2 is created and eight simulations are performed randomly. The
maximum gas blast pressure of 80 MPa is used as an input force. The maximum bending stress
and surface stress shown in Table 2 are calculated automatically in the developed gear profile
design programme.
Downloaded by [Northeastern University] at 06:56 03 December 2014

Figure 14(a) and (b) shows the web retraction and belt load results obtained from the
simulations according to Table 2. In Figure 14(a), the increasing velocity of the web retraction
is the fastest at the eighth case. So, it is considered that the transmission efficiency of the eighth
combination of the gear pair is the best. In Figure 14(a), the discontinuous points exist around
10 ms and the web retraction is decreased sharply. These are because the rack collides with
the spool. The spool is designed to wind the belt webbing by 120 mm when the pinion makes
one revolution. The number of teeth and module of the pinion, however, are changed, and
thus the radius of the tip circle of the pinion is also changed. Therefore, the length of the rack
should be modified for one revolution of the pinion. But the length of the rack is constant in

Figure 15. The variation of mean values of maximum stresses according to the level of design parameters.
 Vehicle System Dynamics 77

the simulation model, so the collision between the rack and spool occurs. Figure 14(b) shows
the axial tension of the belt. In this figure, the increasing velocity of the belt load at the eighth
case is also faster than any other cases.
Figure 15(a) and (b) shows the variation of mean values of the maximum bending stress and
maximum surface stress according to the level of the pressure angle and tip radius coefficient.
In the figures, the maximum bending and surface stresses are the lowest when the levels of
the pressure angle and tip radius coefficient are 1 and 0, respectively [15]. Therefore, it is
concluded that the transmission efficiency of the gear pair is the best and also the generated
stress is the lowest when the number of teeth is 8, module is 2, pressure angle is 25◦ and the
tip radius coefficient is 0.25.
Downloaded by [Northeastern University] at 06:56 03 December 2014

4.3. Verification of the optimisation results

The performances of the simulation models including original and optimised rack and pinion
gear are compared to verify the optimisation results. As prescribed in the preceding section,
the length of the rack should be changed to let the pinion make one revolution. The length of
the optimised rack is increased by 7 mm since the radius of the optimised pinion is greater
than that of the original pinion. During simulation, the collision between the rack and spool
does not occur. In order to make the amount of the web retraction be up to 120 mm, the
propellant whose maximum explosion pressure is 80 MPa is used. Figure 16(a) and (b) shows
the simulation results of the original and optimised models. In both figures, the solid line is the

Figure 16. Comparison of simulation results of the original and optimised model.
 78 S.P. Jung et al.

original model and the dotted line is the optimised model. In Figure 16(a), the web retraction
of the optimised model reached nearly 120 mm. And also, the maximum belt load is about
2.5 kN as shown in Figure 15(b). Therefore, the optimisation is performed successfully, and
the maximum web retraction may be up to 120 mm if the optimised rack and pinion gear and
the propellant whose maximum explosion pressure is 80 MPa are used simultaneously.

5. Conclusions

In this study, the dynamic model of the seat belt pretensioner is developed, and then verified
by comparing the simulation results with the test results. The dynamic model is created on
the basis of the test results obtained when the maximum gas blast pressure is 65 MPa. The
Downloaded by [Northeastern University] at 06:56 03 December 2014

gas pressure curve of 80 MPa is applied to the dynamic model and the simulation results are
compared with the actual test results. The simulation results follow the test results well enough,
and thus it is concluded that the reliability of the developed dynamic model is not only validated
but also the simulation model can be used to estimate the performance of the pretensioner.
On the other hand, the profile of the rack and pinion gear, the operating mechanism of the
pretensioner, is optimised to improve the amount of the web retraction using the design of
experiments. The simulation results of the original and optimised models are compared, and
the web retraction of the optimised model does nearly reach 120 mm. The results of this
paper are considered to be used to design the driving mechanism and improve the retracting
performance of the pretensioner.

References

[1] X.H. Simon and D.W. Michael, A method to evaluate the energy capability of seat belt pretensioners, SAE Paper
No. 9901008, International Congress and Exposition, 1999.
[2] Y.B. Lee, W.H. Ryu, and I.A. Hyun, Leveraging the development of the high-power pretensioner using CAE and
evaluate the influences to injuries, MSC Korea user’s conference, MSC Korea, 2007.
[3] C.K. Jung, H.D. Ahn, H.Y. Kim, and S.W. Woo, Structural analysis of seatbelt system, Fall Conference
Proceedings Vol. 3, Part 10, Korean Society of Automotive Engineers, 1999, pp.1523–1528.
[4] S.K. Min, J.K. Lee, and B.S. Kim, Development of pre-crash safety system using motorized seat belt, Spring
Conference Proceedings, Korean Society of Automotive Engineers, 2004, pp. 1219–1224.
[5] J.S. Park, M.G. Kuk, D.H. Kim, and T.O. Tak, Evaluation of motorized retractor locking mechanism based on
predetermined collision scenarios, Spring Conference Proceedings, Korean Society of Mechanical Engineers,
pp. 212–217, 2007.
[6] S.J. Zhu, Y.F. Zheng, and Y.M. Fu, Analysis of non-linear dynamics of a two-degree-of-freedom vibration system
with non-linear damping and non-linear spring, J. Sound Vib. 271 (2004), pp. 15–24.
[7] P.E. Nikravesh, Computer Aided Analysis of Mechanical Systems, Prentice-Hall, Inc., Englewood Cliffs, NJ,
1988.
[8] S.P. Jung and T.W. Park, An optimum design of a gas circuit breaker using design of experiments, Mech. Based
Des. Struct. Mach. 36(4) (2008), pp. 346–363.
[9] J.R. Colbourne, The Geometry of Involute Gears, Springer-Verlag, New York, 1987.
[10] H.E. Merritt, Gear Engineering, John Wiley & Sons, Inc., New York, 1971.
[11] D.W. Dudley, Handbook of Practical Gear Design, CRC press, New York, 1984.
[12] G.N. Maitra, Handbook of Gear Design, McGraw-Hill, New Delhi, 1985.
[13] C.R. Hicks, Fundamental Concepts in the Design of Experiments, Holt, Rinehart and Winston Inc., New York,
1973.
[14] D.C. Montgomery, Design and Analysis of Experiments, John Wiley & Sons, New York, 1991.
[15] S.H. Park, Understanding of Design of Experiments, Minyoungsa, Seoul, 2006.