Journal of the Korean Society of Marine Engineering, Vol. 38, No. 8 pp.

961~967, 2014 ISSN 2234-7925 (Print)

J. Korean Soc. of Marine Engineering (JKOSME) ISSN 2234-8352 (Online)
http://dx.doi.org/10.5916/jkosme.2014.38.8.961 Original Paper

Optimum seat design for the quadruple offset butterfly valve by analysis of variance with
orthogonal array
Sang-Beom Lee1 ․ Dong-Myung Lee†
(Received September 30, 2014； Revised October 15, 2014；Accepted October 27, 2014)

Abstract: In onshore and offshore plant engineering, a broad use of pipe system have been achieved and accordingly related
technologies has been developed especially in the field of flow control valves. The aim of this study is to suggest the quad-
ruple offset butterfly valve for bi-directional applications which show equivalent operating torque characteristics of the triple
offset butterfly valve. Seat design parameters for the quadruple offset butterfly valve are determined by the proposed method
utilizing both ANOVA (analysis of variance) and the orthogonal array. Through additive model considering the effect of design
parameters on seating torque, mean estimation is performed and thus its optimization results are verified by design of experi-
ment results. The insight obtained from the present study is beneficial for valve design engineers to develop reliable and in-
tegrated design of the quadruple offset butterfly valve.
Keywords: Butterfly Valve, Quadruple Offset, Analysis of Variance (ANOVA), Orthogonal Array, Optimization

1. Introduction pressure, and flow rate, etc. are likely to become more and

With advancement of industrialization, facilities in industries more severe due to the trend of larger plants and production

are likely to diversify and become more complex. Accordingly, facilities. At present, either of a globe valve or gate valve is

offshore and onshore plant industries tend to focus on not only widely spread for control high-pressure. According to the pre-

improving productivities together with reducing production vious researcher [2][3], the number of applications of both

costs but also maintaining and managing the uniformity of butterfly and ball valves are expected to gradually continue to

products through enhancing energy efficiencies and simplifying rise due to the issues such as high cost and low maintain-

processes with automatic control between components. ability of these valves.

Offshore and onshore plants are aimed at manufacturing Until now, numerical studies related to the butterfly valve

products as inter-connected systems with a variety of control were mainly conducted for examining their performance in-

loops. In order to ensure the quality of the products, each con- cluding the loss coefficient of the valve, torque characteristics,

trol loop must work for limiting the main drivers including flow control characteristics and the structural reliabilities

pressure, flow, water level, temperature, etc. within the allow- [4][5]. With respect to the concentered butterfly valve, the

able range of each parameter [1]. Equipment and devices in- characteristics of the valve is numerically assessed based on

stalled in plants should be provided various sources including the structural and flow analysis in order to ensure the struc-

material, energy, and information enabling them efficiently. In tural reliability and the flow stability and therefore an opti-

this regard, pipelines play an important role to deliver material mum design of the valve disc was performed by the orthogo-

and energy from one device to another and the control of their nal array [5][6]. Further, the study of the pressure drop and

function is mainly dependent on the valve system. cavitation characteristics was carried out for the concentered

Furthermore, it is found that they are essential components butterfly-type valve [7]. For the offset butterfly valve, Park et

transporting various fluids in ships although pipelines are rela- al. [8] focused on the effect of the eccentric shaft on the in-

tively small compared with the cooling system built in onshore ternal flow characteristics and Lee et al. [9] analyzed the flow

plants such as desalination plants, chemical plants, nuclear and characteristics with varying shape of the seat.

thermal power plants. For the studies associated with butterfly valves, since they

Recently, the design requirements related to temperature, were designed in accordance with ANSI B 16.34 Class

†Corresponding Author (http://orcid.org/0000-0002-6132-0636): Mechanical Part Technology Center, Korea Testing Laboratory, Sa-dong,
Sangnok-gu, Ansan-si, 426-910, Korea, E-mail: dmlee1031@ktl.re.kr, Tel: 031-500-0407
1 Global Core Research Center, Pusan National University, E-mail: lsb7766@pusan.ac.kr, Tel: 051-510-2385

Copyright ⓒ The Korean Society of Marine Engineering

 Sang-Beom Lee ․ Dong-Myung Lee

150~300 which was designed for low-pressure. It is hard to by line type.

find track records in the high-pressure which causes severe Figure 2 shows schematic representation of the triple offset
structural deformation to the disc, detachment of the seat and butterfly valve. Using the triple offset butterfly valve in re-
non-applicability of the reverse direction. In this regards, Lee verse direction, when the engagement angle is large, the leak-
and Kim [10] proposed the quadruple offset butterfly valve age is more likely to happen, and on the other hand, when the
which is keeping advantages of triple offset butterfly valve angle is relatively small, the durability of seat could drop due
with improving reversal performance and identified sensitivities to the significant increment in torque. To tackle these issues,
of each parameter based on the parametric studies in order to Lee and Kim [10] proposed the new quadruple offset butterfly
ensure the applicability under high-pressure. valve applying offset for the seat as shown in Figure 3.
The purpose of this study is to define design range of the According to their results, it was expected to improve the
main parameters for the seat and to optimize the shape of the valve performance in reverse direction with keeping the bene-
seat in association with seating torque of the quadruple offset fits of the triple offset valve.
butterfly valve which is capable of applying bi-directional
applications. By applying ANOVA and orthogonal array for
the seat optimization, sensitivities for design parameter are
identified and compared with the results of mean estimation.
The mean estimation of seating torque is evaluated by the ad-
ditive model considering the main effects of each design pa-
rameter and validated by comparison with the analysis results.

2. Quadruple Offset Butterfly Valve

The butterfly valve can be divided into two types in aspect
of operating axis. Concentric type and offset type whether the Figure 1: Seating torque characteristics of butterfly valve
disc aligns with rotating axis depending on the structure of the
seat. There are two offset valves are mostly used which named
as double and triple eccentric offset butterfly valve. According
to the structure of offset in the valve, the operating torque
characteristic is significantly varies according to the flow in-
duced pressure differences between upstream and downstream
of the pipeline. However, for valves under operating con-
ditions, seating torque required to open the closed valve disc
separating from the seat is the most important factors affecting
the valve performance and its effect on the disc becomes more
significant when it is fully closed. Therefore, the most im-
portant factor in design of a driving part of valves is to dimin- Figure 2: Triple offset butterfly valve
ish seating torque acting on the seat itself. As a result, a vari-
ety of seat configurations has been proposed in terms of re-
ducing seating torque.
Figure 1 describes a comparison example of the seating tor-
que ratio for double and triple offset butterfly valves in terms
of disc opening angles. It is observed that seating torque of
triple offset valves in low opening angles varies more dramati-
cally than the double offset valves due to the effect of the
changing contact type from a surface to a line [11]. However,
it has been recognized that the triple offset butterfly valve
shows weak performance for the reverse direction because of
the applied contact type which is only able to seal the leakage Figure 3: Quadruple offset butterfly valve

Journal of the Korean Society of Marine Engineering, Vol. 38, No. 8, 2014. 10 962
 Optimum seat design for the quadruple offset butterfly valve by analysis of variance with orthogonal array

3. Optimization Algorithm
In the present study, the optimization algorithm is developed
based on the structural analysis, the orthogonal array and
ANOVA, and the detail of each step is as follows:
Step 1. Definition and selection of object values
Step 2. Determine design parameters, the number of level
and their values
Step 3. Selection of orthogonal array and allocation of de-
sign parameters considering overall degrees
Figure 4: Valve operating direction
Step 4. Executing design of experiment and calculating ob-
ject values according to the combinations of design
parameters determined by the orthogonal array
Step 5. Calculating mean of object values for each level of
design parameters
Step 6. Selection of significant factors influencing most on
the object values based on the results of ANOVA
Step 7. Determining optimum design conditions resulting
from the mean estimation
Step 8. Estimation of the object values for significant factors
under optimum design conditions by additive model
Step 9. Implementing confirmation experiment and estimat-
Figure 5: Boundary condition for experiment
ing real object values under optimum design con-
ditions
Step 10. Determining optimum design conditions as an opti- 4.1 Determining design parameters
mum solution based on the validation of additivity In the present study, 5 design parameters for the quadruple
Step 11. If the additivity of the additive model is not met offset butterfly valve are defined as shown in Figure 6. The
for the criteria, back to the Step 4 in consideration detail is as follows.
of developing new orthogonal array of selection of - Parameter 1: Offset between flange and shaft center lines
the design parameters, adjusting the level range and (fixed at 10.0mm)
including the interaction term and perform all the - Parameter 2: Offset between seat and shaft center lines
steps accordingly. (fixed at 47.5 mm)
- Parameter 3: Offset between flange center line and seat
4. Optimization Seat Configuration angle
The aim of this chapter is to analyze the relation between - Parameter 4: Angle offset between shaft center line and
each design parameter and seating torque for the quadruple seat center line
offset butterfly valve. By using finite element method compu- - Parameter 5: Engagement angle of seat
tation, torque of the shaft is numerically calculated in terms of The configuration optimization is carried out for parameters 3,
rotational directions, here preferred and reverse directions. In 4 and 5.
the present study, the fluid flow from left to right is defined Park [12] noted that in order to identify linear and nonlinear
as the preferred direction as shown in Figure 4. Applied boun- effect, level 2 and 3 of the design parameter are proper,
dary conditions of the finite element method computation are respectively. In the present study, level 3 is applied. Table 1
shown in Figure 5. It is assumed that both ends of valves are represents the level of each design parameter based on the de-
fixed and the load, 16.85 MPa is applied for the inside of termined results along with the maximum and minimum rang-
Class 900. To apply the contact condition between the disc es at which no physical interference. To analyze interactions
and the body of the seat, “No separation” contact condition is among parameters, 3 design parameters with 3-level by using
applied. The “Bond” contact condition is applied for the valve the orthogonal array of full factorial experiment are defined
body and the disc with their stems. and summarized in Table 1. Table 2 shows the numerical

Journal of the Korean Society of Marine Engineering, Vol. 38, No. 8, 2014. 10 963
 Sang-Beom Lee ․ Dong-Myung Lee

4.2 Analysis of variance (ANOVA)

ANOVA results for torque in terms of the level average are
presented in Table 3. It is found that the maximum difference
(∆) between average levels in mean estimation shows C, A
and B order.
Figure 7 summarizes the mean response in terms of parameter
levels. It is observed that the slope according to the level
change as depicted in Table 3 is relatively large in design pa-
rameter C which shows the most influential design parameter on
Figure 6: Quadruple offset butterfly valve design parameters the seating torque. In order to analyze the significant factors in
seating torque, ANOVA is implemented and the results are pre-
Table 1: Design parameters and levels sented in Table 4 and Table 5 in terms of design parameters. As
Design Parameter Leval a result, design parameter C shows the most significant factor
Symbol Description Unit 1 2 3 from the  value and contribution ratio of parameters. Figure 8
A Offset 3 deg. 12 15 18
displays the level of contribution among design parameters.
B Offset 4 deg. 1 4 7
C Engagement angle deg. 8 14 20 Figure 9 shows the results of interaction effect between design
parameters and it shows that effect is insignificant.

Table 2: Orthogonal array design and the experimental results

Table 3: Mean effect response table
Response (kN·m)
Run A B C Preferred Direction Reverse Direction
Preferred Reverse Level
A B C A B C
Direction Direction
1 10.178 9.953 10.638 9.824 9.717 10.305
1 1 11.06 10.76
2 9.860 9.839 9.783 9.572 9.554 9.468
2 1 2 10.22 9.91
3 9.504 9.750 9.121 9.273 9.398 8.896
3 3 9.55 9.32
△ 0.673 0.202 1.518 0.551 0.319 1.409
4 1 11.04 10.65
Rank 2 3 1 2 3 1
5 1 2 2 10.1 9.72
6 3 9.37 9.09
7 1 11.04 10.56
8 3 2 10 9.52
9 3 9.21 8.88
10 1 10.69 10.43
11 1 2 9.91 9.66
12 3 9.3 9.11
13 1 10.64 10.32
14 2 2 2 9.79 9.49
15 3 9.13 8.91
16 1 10.62 10.21
17 3 2 9.68 9.31
18 3 8.98 8.72 Figure 7: Mean effect response graph
19 1 10.27 10.05
20 1 2 9.56 9.36 Table 4: ANOVA results table for preferred direction
21 3 9 8.86 Percentage of
22 1 10.21 9.94 Source DOF SS MS F value
contribution (%)
23 3 2 2 9.44 9.2 A 2 2.0425 1.0212 4000.35 15.99
24 3 8.84 8.67
B 2 0.1851 0.0925 362.51 1.45
25 1 10.17 9.83
C 2 10.4246 5.2123 20417.55 81.64
26 3 2 9.34 9.04
27 3 8.7 8.51
A×B 4 0.0003 0.0001 0.26 0
A×C 4 0.0670 0.0168 65.64 0.52
B×C 4 0.0501 0.0125 49.04 0.39
results of in accordance with the design parameters determined Error 8 0.0020 0.0003
by the orthogonal array of full factorial experiment. Total 26 12.7695 100.0

Journal of the Korean Society of Marine Engineering, Vol. 38, No. 8, 2014. 10 964
 Optimum seat design for the quadruple offset butterfly valve by analysis of variance with orthogonal array

Table 5: ANOVA results table for reverse direction

4.3 Determining optimum design parameters
Percentage of
Source DOF SS MS F value Since the optimum conditions of significant factors minimiz-
contribution (%)
A 2 1.3699 0.6849 2348.15 12.49 ing seating torque is a combinations of both ends, according to
B 2 0.4574 0.2287 784.05 4.17 the results of mean estimate, the design parameters of A, B
C 2 9.0384 4.5192 15493.26 82.42 and C with 3-level, would be the optimum conditions. To ap-
A×B 4 0.0015 0.0004 1.29 0.01
ply for the additive model, ANOVA is conducted for design
A×C 4 0.0721 0.0180 61.79 0.66
B×C 4 0.0269 0.0067 23.09 0.25 parameters of A, B, and C without interaction effect by the
Error 8 0.0023 0.0003 orthogonal array. The results are presented in Table 6 and
Total 26 10.9662 100.0 Table 7. Mean estimation of seating torque within optimum
design conditions is predicted by additive model using
Equation (1) considering main effects of significant factors.



     
 
 (1)

Where,  is the mean of in Table 2 represents the results

of measured seating torque for experiment simulations in terms
of directions. And  is number of design parameters and   
(a) Preferred direction (b) Reverse direction th
is the mean of torque in the optimum condition for i design
Figure 8: Contribution of factors
parameter. Selected optimum design condition is the same as
"Run 27" of the orthogonal array. The additivity of additive
model is validated by the seating difference between prediction
and experiment. The prediction is estimated based on the 95%
of confidence interval of the population mean. The population
mean is calculated by Equation (2), it is confirmed that the
difference, is within the error limit [13].




 
    
 
(2)

Figure 9: Interaction Plot  

 
  
  

× 

(3)

Table 6: ANOVA results for preferred direction : without

interaction effect Where  is degree of error,  is confidence coefficient (1 -

Source DOF SS MS F value confidence interval),  is variance of error and  is sig-
A 2 2.0425 2.0425 1.0212 nificant number variation. And  is number of experiment
B 2 0.1851 0.1851 0.0925
data in each parameter and  is number of total experiment
C 2 10.4246 10.4246 5.2123
Error 20 0.1194 0.1194 0.006 data.
Total 26 12.7715
4.4 Results of Optimum seat design
For alternative of selected model, it is unable to implement
Table 7: ANOVA results for reverse direction : without for seat interaction in practical cases. In order to settle down
interaction effect
this problem, the engagement angle is adjusted and therefore
Source DOF SS MS F value the interaction effect is eliminated. Table 8 summarizes the se-
A 2 1.3699 1.3699 0.6849 lected optimum design conditions. Under selected conditions,
B 2 0.4574 0.4574 0.2287
mean estimation of the seating torque using additive model are
C 2 9.0384 9.0384 4.5192
developed by a form of prediction equation suggested
Error 20 0.1029 0.1029 0.0051
Total 26 10.9685 Equation (1). Table 2 shows the mean of measured seating

Journal of the Korean Society of Marine Engineering, Vol. 38, No. 8, 2014. 10 965
 Sang-Beom Lee ․ Dong-Myung Lee

torque,  in experiment simulation for each direction. 

 deg 5. Conclusion
represents the prediction values at 25 degrees resulting from The aim of this study is to define design range of the main
the 2nd order regression analysis for mean estimation of de- parameters for seat and to optimize the shape of the seat in
sign parameter, C in Table 3. association with seating torque of quadruple offset butterfly
valve capable of bi-directional applications.
Table 8: Optimum values of design parameters Based on the results obtained from the present study, the
Design Parameter Unit Value following conclusions can be drawn.
Offset 1 mm 47.5 1. Design parameters for the quadruple offset butterfly valve
Offset 2 mm 10.0 are identified by orthogonal array and ANOVA and there-
Offset 3 deg. 18.0 fore it is observed that the engagement angle is the most
Offset 4 deg. 7.0 influential parameter. Further, it is noted that the inter-
Engagement Angle deg. 25.0 action effect among design parameters for representing
the shape of the seat are negligible.
2. Using optimization algorithm utilizing orthogonal array
A validation experiment is implemented according to the fi-
and ANOVA, optimization of seat configuration for the
nal optimum design conditions and their results are 8.924
quadruple offset butterfly valve is carried out. As a result,
kN·m for preferred direction and 8.182 kN·m for reverse the design parameter without interference is determined
direction. Table 9 summarizes the experimental results of opti- and the reduction of seating torque is found 4.7 % and
mum seat design. Seating torque shows reduction in 4.7% for 3.9% in preferred and reverse directions, respectively.
preferred direction and in 3.9 % for reverse direction com- 3. At present study, the optimization is only taking into ac-
pared to the optimal design. Furthermore, the additivity of ap- count the seating torque as an objective function.
plied additive model in optimal model prediction is verified by However, it is necessary to consider the hydrodynamic
using not only the difference of seating torque between the es- torque. In near future, to establish the validity of the pro-
timated and actual torque at optimum design conditions but al- posed method, the actual product will be manufactured
so the 95% confidence interval. The difference of seating tor- and tested.
que between the estimated and the actual torque for both di-
rections is proved within the error limit of the population Acknowledgement
mean calculated from the Equation (2) and Equation (3).
This work was supported by the National Research
Foundation of Korea (NRF) grant funded by the Korea gov-
Table 9: Experimental results of optimum seat design
ernment (MSIP) through GCRC-SOP (No. 2011-0030013).
Response (kN·m)
Condition Item
Preferred Reverse
References
Direction Direction [1] Fisher, Control Valve Handbook, 4th ed., Iowa, USA:
Emerson Process Management, 2005.
Experiment 8.704 8.511
[2] D. G. Kim and J. H. Kim, “A study on structural
Estimation 8.680 8.455
  analysis of globe valve for LNG carrier,” Journal of
(Run 27)  the Korean Society of Marine Engineers, vol. 31, no.
0.024 0.056
(       ) 8, pp. 1013-1019, 2007 (in Korean).
Error limit () 0.082 0.076 [3] H. S. Jung, Y. H. Kim, J. R. Cho, J. H. Kim, J. R.
Experiment 8.294 8.182 Kim, and J. H. Park, “A study on structural design of
cryogenic miniature globe valve using finite element
Estimation 8.293 8.192
Optimum method,” Journal of the Korean Society of Marine
values 
0.001 0.010 Engineers, vol. 31, no. 4, pp. 343-349, 2007 (in
(       )
Korean).
Error limit () 0.082 0.076
[4] J. U. Lee, D. H. Lee, and Y. H. Choi, “Numerical
Experiment - 4.7% - 3.9%
Remark analysis of incompressible and compressible flow
Estimation - 4.5% - 3.1% around a butterfly valve,” Journal of Energy

Journal of the Korean Society of Marine Engineering, Vol. 38, No. 8, 2014. 10 966
 Optimum seat design for the quadruple offset butterfly valve by analysis of variance with orthogonal array

Engineering, vol. 11, no. 1, pp. 26-33, 2002 (in

Korean).
[5] Y. C. Park, J. H. Kang, J. M. Lee, and J. Kang,
“Optimization of butterfly valve’s disc using the
DACE model based on CAE,” Journal of Ocean
Engineering and Technology, vol. 20, no. 3, pp.
96-102, 2006 (in Korean).
[6] X. G. Song, L. Wang, S. H. Baek, and Y. C. Park,
“Multidisciplinary optimization of a butterfly valve,”
ISA Transactions, vol. 48, no. 3, pp. 370-377, 2009.
[7] K. Ogawa and T. Kimura, “Hydrodynamic character-
istics of a butterfly valve - Prediction of torque char-
acteristics,” ISA Transactions, vol. 34, no. 4, pp.
327-333, 1995.
[8] S. M. Park, H. K. Choi, and G. J. Yoo, “Study on
flow characteristics for eccentric shaft in the butterfly
valve system,” Proceedings of Korean Society for
Computational Fluid Engineering Spring Conference,
pp. 587-591, 2011 (in Korean).
[9] J. W. Lee, H. K. Choi, and G. J. Yoo, “Characteristic
of butterfly valve flow with different design factors,”
Journal of Computational Fluids Engineering, vol. 15,
no. 1, pp. 64-70, 2010 (in Korean).
[10] D. M. Lee and S. Y. Kim, “Sensitivity analysis of
design parameters for quadruple offset butterfly valve
by operating torque,” Journal of Ocean Engineering
and Technology, vol. 28, no. 2, pp. 160-166, 2014
(in Korean).
[11] Tomoe Valve Co.,Ltd, http://www.tomoevalve.co-
m/english/digital_catalog/pdf/general_catalogue.pdf,
Accessed January 11, 2014.
[12] G. J. Park, “Design of experiments,” Analytic
Methods for Design Practice, Berlin, Germany:
Springer, pp. 325-339, 2007.
[13] M. S. Phadke, Matrix Experiments Using Orthogonal
Arrays, Quality Engineering Using Robust Design,
New Jersey, USA: PTR Prentice Hall, pp. 41-66,
1989.

Journal of the Korean Society of Marine Engineering, Vol. 38, No. 8, 2014. 10 967