Sunteți pe pagina 1din 5

Vol 03, Issue 04; August 2012

International Journal of Mechanical Engineering applications Research IJMEAR


http://technicaljournals.org ISSN: 2249- 6564


2010-2012 - TECHNICALJOURNALS

, Peer Reviewed International Journals-IJCEA, IJESR, RJCSE, PAPER, ERL, IRJMWC,


IRJSP, IJEEAR, IJCEAR, IJMEAR, ICEAR, IJVES, IJGET, IJBEST TJ-PBPC, India;
Indexing in Process - EMBASE, EmCARE, Electronics & Communication Abstracts, SCIRUS, SPARC, GOOGLE Database, EBSCO, NewJour, Worldcat,
DOAJ, and other major databases etc.,
185
Analysis and Optimization of Pump Performance Variables using
Genetic Algorithm and Taguchi Quality Concept: A Case Study
R.RAGOTH SINGH
1,2
, M. NATARAJ
3

1
Assistant Professor, Kathir College of Engineering, Coimbatore, INDIA

2
PhD. Scholar, Anna University, INDIA
3
Associate Professor, Government College of Technology, Coimbatore, INDIA
1
ragothsingh@rediffmail.com

ABSTRACT
As the diffusion of flow process is highly complex in pump operation, it is necessary to design / develop the
geometry of impeller and casing to reduce the flow losses in pump significantly. This paper describe a research
methodology to find the near optimum combination of pump operating variables for performance enhancement
with consideration of multiple quality characteristics using GA and Taguchi off-line quality design concept. The
experimental results are justified by ANOVA and confirmed by conformation experiments. Taguchi OA based
DoE techniques determines the required experimental trials. The genetic algorithm is developed for finding
parameter values in a continuous solution space to optimize a performance measure on denier and tenacity
qualities. The performance is evaluated by the technique for order preference. The parameters chosen for
design optimization are Impeller eye diameter, Vane outlet angle, number of vanes and volute area. The levels
for the parametric specification are chosen from the ranges where the pump will get the best efficiency point
(BEP). Mathematical model is formulated from the results of experimental analysis using multiple regression
analysis method. Coding is developed to obtain the design variables and responses for the above model. The
formulated problem is solved using Taguchi quality concept and non-traditional search optimization algorithm
called Genetic Algorithm.
Keywords: Centrifugal pump, Performance variable, OA, DoE, BEP, ANOVA, Genetic algorithm

1. INTRODUCTION
Mankind had sensed the importance of pumps in hydraulic field over centuries (Tambe, 1992). In spite of its
age, even today the design of pumps has not attained saturation stage. In fact, pump design compatible to
production environment had been a major challenge to researchers (Li et al., 2002; Li, 2004; Oh & Chung,
1999; Oh & Kim, 2001; Yoon et al., 1998). Particularly, researchers along with engineers had been striving to
develop pump with the specific attention directed to increase the overall efficiency for performance
enhancement (Kumarasamy, 2001; Nataraj & Arunachalam, 2004; Nataraj et al., 2005). Performance prediction
procedures for centrifugal and axial flow turbo machines are not well established in the open literature owing to
the industrial sources being classified as competition-sensitive by every industry. With the emergence of
powerful quality design tools (Park, 1996; Taguchi, 1986) for novelty, an attempt has been made in this research
study to propose a modified design parameter combination which describes the sub-macro-characteristics, such
as total head, discharge and power based on geometry of the pump impeller and casing at Best Efficiency Point
(BEP), since the practical pump design process involves a number of compromises between overall efficiency,
discharge, total head, suppression of cavitations, reduction in product price and many other vendees
requirements. Extensive research work to improve the performance of centrifugal pump over the last few
decades reveals that the pump performance depends on impeller and casing geometry. Of relevance to the work
reported in this study is an extensive analysis for optimizing impeller and casing dimensions of a centrifugal
pump for the requirements through basic design procedure (Nataraj & Arunachalam 2004; Nataraj et al, 2005).
The parameters chosen for design optimization of impeller are eye diameter, vane outlet angle, number of vanes
and the volute area in casing. Dr. Genichi Taguchi has introduced more cost effective engineering methodology
namely robust design to deliver high quality products at low cost through research and development. It can
greatly improve an organizations ability to meet market windows, keep development and manufacturing costs
as low as possible. Genetic algorithms are computerized search procedures based on the mechanics of natural
genetics and natural selection that can be used to obtain global and robust solution for optimization problems. It
concentrates on simple genetic algorithm with reproduction, crossover and mutation operators. Robust design
uses many ideas from statistical experiment design and adds a new dimension to it by explicitly addressing two
major concerns faced by all products and process designers.

2. PROBLEM FORMULATION
A pump is one of the simplest equipment in any process plant. Figure1 shows the cross section of the casing
with impeller. The energy changes occur by virtue of its two main parts impeller and casing. Impeller is a
Vol 03, Issue 04; August 2012

International Journal of Mechanical Engineering applications Research IJMEAR
http://technicaljournals.org ISSN: 2249- 6564


2010-2012 - TECHNICALJOURNALS

, Peer Reviewed International Journals-IJCEA, IJESR, RJCSE, PAPER, ERL, IRJMWC,


IRJSP, IJEEAR, IJCEAR, IJMEAR, ICEAR, IJVES, IJGET, IJBEST TJ-PBPC, India;
Indexing in Process - EMBASE, EmCARE, Electronics & Communication Abstracts, SCIRUS, SPARC, GOOGLE Database, EBSCO, NewJour, Worldcat,
DOAJ, and other major databases etc.,
186
rotating part that converts driver energy into kinetic energy. Volute casing is the stationary part that converts
kinetic energy into pressure energy.
2.1 Velocity Distribution in Impeller and Volute.
Velocity distribution around the casing (1) and the impeller (2) is shown in the figure 2. The discharge portion
(3) carries liquid from the impeller so as to keep the losses to a minimum and maintain the flow through the
impeller as axially symmetric as possible. The radial velocity component is obtained from the continuity
equation.

Fig.1. Casing and impeller of centrifugal pump

2.2 Parameter Specification and Range
The pump is designed for 30m Head (H), 0.667 lps Discharge (Q) and 2800 rpm speed. Input data for impeller
and casing are provided by sponsoring industry in Coimbatore.

Fig.2. Velocity distribution in volute
Brainstorming session comprising of technical crew and field experts on impeller and casing for the centrifugal
pump concluded that the most influencing factors for improving the overall efficiency which increase the
discharge, maximize the head level and minimize the power consumption were
- Vane outlet angle (
2
)
- Impeller eye diameter (D
e
)
- Number of vanes (Z)
- Volute area (Diameter of volute) (D
v
)
Table 1 shows the control factors (A, B, C and D) and their corresponding levels. (Among the above control
factors impeller Vane outlet angle (A) is given four levels, remaining factors Impeller eye diameter(B), No. of
vanes(C) and Volute area(D) are considered with two levels), and noise factors (speed, current and frequency)
do not assigned levels as they cannot be maintained steadily.
2.3 Experimental setup
The problem is illustrated with a real case study. Impellers and casing are fabricated for the different parameters
combination (Table 2). Trial runs are conducted from the fabricated model and the results are obtained. Figure 3
shows the experimental setup in R&D laboratory.
Table 1. Parameters and Levels
Control Factors
Levels
1 2 3 4
A Vane outlet angle (deg) 35 37 39 41
B Impeller eye Diameter (mm) 27 32 - -
C No. of vanes 6 12 - -
D Volute Dia. (mm) 12 14 - -
Noise Factors Range
Speed (rpm) 2640-2814
Current (amps) 3.54-5.60
Frequency (Hz) 49.2-49.6
The impeller is designed by single arc method; volute area is estimated by constant mean velocity method.
Prototype models of pump have been fabricated for the Taguchi L
8
(4
1
X2
3
) mixed orthogonal array based
experiments design, these pumps performance is evaluated for increased discharge, raise in head level with
Vol 03, Issue 04; August 2012

International Journal of Mechanical Engineering applications Research IJMEAR
http://technicaljournals.org ISSN: 2249- 6564


2010-2012 - TECHNICALJOURNALS

, Peer Reviewed International Journals-IJCEA, IJESR, RJCSE, PAPER, ERL, IRJMWC,


IRJSP, IJEEAR, IJCEAR, IJMEAR, ICEAR, IJVES, IJGET, IJBEST TJ-PBPC, India;
Indexing in Process - EMBASE, EmCARE, Electronics & Communication Abstracts, SCIRUS, SPARC, GOOGLE Database, EBSCO, NewJour, Worldcat,
DOAJ, and other major databases etc.,
187
minimized power consumption at the BEP. Table 2 gives the responses of the real model analysis arrived from
taguchi quality design concept.


Fig. 3. Experimental setup
Table 2. Experimental Result













3. ANALYSIS USING GENETIC ALGORITHM & TAGUCHI QUALITY CONCEPT
3.1 Genetic Algorithm
Genetic Algorithm (GA) proposed by goldberg based on natural genetics has been made use of in this work for
an unconstrained problem. Formulation of fitness function as applicable to the present study has also been
presented. Mathematical modeling was created from the result obtained from experimental analysis of the
centrifugal pump parameters using multiple regression analysis method. In this case there are four independent
variables and three dependent variables. The relationships between these variables are of the form as follows
f
1
(x) = o
0
+ o
1

2
+ o
2
De + o
3
Z + o
4
Dv ------- (1)
f
2
(x)= o

0
+ o

1

2
+ o

2
De + o

3
Z + o

4
Dv ------- (2)
f
3
(x) = o

0
+ o

1

2
+ o

2
De + o

3
Z + o

4
Dv ------- (3)
The model is satisfying ANOVA results, the mathematical equations 1, 2 and 3 obtained among centrifugal
pump parameters are valid and the modeled equations are as follows
f
1
(x) = 0.3512 + 0.00071
2
+ 0.004998 D
e
+0.003021 Z + 0.00823 D
v
------- (4)
f
2
(x) = 28.646 0.0842
2
0.1143 D
e
0.0596 Z + 0.3083 D
v
------- (5)
f
3
(x) = 0.9504 + 0.000378
2
+ 0.1187 D
e
0.001852 Z + 0.00423 D
v
------- (6)
3.1.1 Objective Function
The objective function of the problem is given in the equation 7. w
1
, w
2
, w
3
are constants whose values indicate
the relative importance of one objective function with another objective function.
f (x)= [w
1
*1/(1+ f
1
(x))] + [w
2
* 1/(1+ f
2
(x))] - [w
3
* f
3
(x))]. ------- (7)
3.1.2 Effect of Genetic Parameters
The effect of genetic parameters in the centrifugal pump optimization problem using genetic algorithm are
Population size (250 to 550), Cross over probability (0.6 to 0.95), Mutation probability (0.01 to 0.08) and
Number of iterations (100 to 400). Table 3 shows the effect of genetic parameters. The population size, cross
over probability, mutation probability and iterations are evaluated for the maximum discharge, maximum head
level and minimum power consumption.
3.1.3 Optimum Parameter of Centrifugal Pump
The following are the near optimum parameter combination of pump operating variables obtained from Genetic
Algorithm approach for the population size of 400, the cross over probability 0.65, mutation probability 0.04
and iteration of 150.
3.2. Factorial Experimental Design
Designs of Experiments (DoE) are frequently used to measure the effects of one or more variables and response.
The factorial design, based on the principle of orthogonal array, is very useful in product development process.
Trial
Inner Array L
8
RESULTS
Parameters Discharg
Q(lps)
Head
H(m)
Power
P(kw)
A B C D
1 1 1 1 1 0.6057 25.512 1.056
2 1 2 2 2 0.7292 24.868 1.008
3 2 1 1 2 0.6476 27.278 1.021
4 2 2 2 1 0.6104 26.063 1.059
5 3 1 2 1 0.6640 25.786 1.067
6 3 2 1 2 0.7183 24.624 1.024
7 4 1 2 2 0.6718 24.750 0.994
8 4 2 1 1 0.6312 25.484 1.072
Vol 03, Issue 04; August 2012

International Journal of Mechanical Engineering applications Research IJMEAR
http://technicaljournals.org ISSN: 2249- 6564


2010-2012 - TECHNICALJOURNALS

, Peer Reviewed International Journals-IJCEA, IJESR, RJCSE, PAPER, ERL, IRJMWC,


IRJSP, IJEEAR, IJCEAR, IJMEAR, ICEAR, IJVES, IJGET, IJBEST TJ-PBPC, India;
Indexing in Process - EMBASE, EmCARE, Electronics & Communication Abstracts, SCIRUS, SPARC, GOOGLE Database, EBSCO, NewJour, Worldcat,
DOAJ, and other major databases etc.,
188
These designs are economic and easy to use and can provide a great deal of valuable information. Taguchi
Orthogonal Array (OA) concept is versatile recipes that apply to several experimental conditions.
Table 3. Optimized result
Sl.no Descriptions
Optimum
values
Design parameters
1 Vane outlet angle (|
2
) 35
2 Eye diameter (De) 32
3 No. of vanes (Z) 12
4 Volute area (Dv) 12
Objective functions
1 Discharge (lps)

0.671
2 Head (m) 25.026
3 Power (Kw) 1.0301
3.2.1. Data Analysis and Evaluation
Table 4 gives the Signal-to-Noise (SN) ratios calculated using taguchi off-line quality design equations (1) and
(2). There are several S/N ratios available depending on the type of characteristic - Smaller The Better (STB);
Larger The Better (LTB).
SN
LTB
= -10 log
(

=
i
i
y n
1
2
1 1
------- (8)
SN
STB
= -10 log
(

=
i
i
y
n
1
2
1
-------- (9)
Table 4. Summary Table for SN Ratio Calculation
Trial
Control
Factors
SN Ratio dB
A B C D Head Discharge Power
1 1 1 1 1 23.26 -4.61 -0.35
2 1 2 2 2 23.26 -2.34 -0.15
3 2 1 1 2 23.29 -2.32 -0.45
4 2 2 2 1 23.03 -4.11 -0.61
5 3 1 2 1 23.17 -4.17 -0.60
6 3 2 1 2 23.30 -2.85 -0.41
7 4 1 2 2 23.16 -3.37 0.06
8 4 2 1 1 23.20 -3.76 -0.54

3.2.2 Predicted SN Ratio For Overall Optimum Condition
The optimal head, discharge and power are predicted at the determined optimal parameters. The significant
parameters with optimal levels are selected as: A
1
B
2
C
1
D
2
. The Predicted SN ratio is A
1
+ B
2
+ C
1
+ D
2
3T.
3.2.3 Over All Predicted Levels
SN ratio for head =
8
67 . 187
3
4
01 . 93
4
05 . 93
4
79 . 92
2
52 . 46
+ + +
= 23.34 dB
SN ratio for discharge=
8
53 . 27
3
4
88 . 10
4
54 . 13
4
06 . 13
2
95 . 6
+

= -2.52 dB
SN ratio for power =
8
78 . 2
3
4
68 . 0
4
48 . 1
4
44 . 1
2
5 . 0

= -0.1075 dB

3.2.4 Confirmation Test
Confirmation experiment was conducted at the optimal setting (A
1
= 35 deg, B
2
= 32 mm, C
1
= 6 and D
2
= 14
mm) of pump parameters.

4. CONCLUDING REMARKS
Optimum design specifications obtained from Taguchi off-line quality design and Genetic Algorithm approach
are given in table 5. This comprehensive analysis indicates that the parameter level for Vane outlet angle and
Impeller eye diameter do not vary; However Number of vanes(C) and Volute diameter (D) have no influence in
Vol 03, Issue 04; August 2012

International Journal of Mechanical Engineering applications Research IJMEAR
http://technicaljournals.org ISSN: 2249- 6564


2010-2012 - TECHNICALJOURNALS

, Peer Reviewed International Journals-IJCEA, IJESR, RJCSE, PAPER, ERL, IRJMWC,


IRJSP, IJEEAR, IJCEAR, IJMEAR, ICEAR, IJVES, IJGET, IJBEST TJ-PBPC, India;
Indexing in Process - EMBASE, EmCARE, Electronics & Communication Abstracts, SCIRUS, SPARC, GOOGLE Database, EBSCO, NewJour, Worldcat,
DOAJ, and other major databases etc.,
189
the pump performance, since these level vary with optimization technique. The analysis also discloses the
optimized pump using Taguchi off-line quality design offer better performance [0.702 lps discharge, 27.04 m
head and 1.01 kW power] than the optimized pump using Genetic Algorithm approach [0.671 lps discharge,
25.02 m head and 1.03 kW power]. Hence it is concluded from the research work the Taguchi off-line quality
design method is the best choice for the optimizing the pump performance variables with multiple quality
characteristics.
Table 5. Comparison of result
Optimum Design Specifications
Parameters Taguchi GA
A Vane angle (deg) 35 35
B Impeller eye Dia. (mm) 32 32
C No. of vanes 6 12
D Volute Dia. (mm) 14 12
Performance of the pump
Response Response Response
Discharge (lps) 0.702 0.671
Head (m) 27.04 25.02
Power (kw) 1.01 1.03

5. REFERENCES
[1] Li, W.-G., Su, F.-Z. and Xiao, C. (2002) Influence of the number of impeller blades on the performance of
centrifugal oil pumps. World Pumps, 427, 32-35.
[2] Li, W.-G. (2004) A Method for analyzing the performance of centrifugal oil pumps. ASME Journal of
Fluids Engineering, Vol. 126, 482-485.
[3] Nataraj, M., Arunachalm, V.P. (2004) " A Novel Design Approach for Quality Improvement of Centrifugal
Pump Using Taguchi's Robust Design Concept with a Case Study from M/S V-Guard Industries (P)
Limited", 21st All India Manufacturing Technology, Design and Research (AIMTDR) Conference on
Innovating The Future Through Manufacturing, Vellore Institute of Technology Deemed University, 20th-
22nd December 2004, pp. 536-547.
[4] Nataraj, M., Arunachalm, V.P., and Ragoth Singh, R.(2005) " Design and development of volute casing for
centrifugal water pump: A virtual model analysis", National Conference on Product Development with
Mechatronic System for Global Quality (PMGQ 2005), at Thiagarajar college of Engg, Madurai, 2nd-3rd
May 2005, pp.1-5.
[5] Oh, H. W., and Chung, M. -K.(1999) Optimum values of design variables versus specific speed for
centrifugal pumps. Proc. Instn. Mech. Engrs, 1999, Vol. 213 Part A, 219-226.
[6] Oh, H. W., and Kim, K. -Y.(2001) Conceptual design optimization of mixed flow pump impellers using
mean streamline analysis. Proc. Instn. Mech. Engrs, 2001, Vol. 215 Part A, 133-138
[7] Park, S. H.(1996) Robust Design and Analysis for Quality Engineering, London: Chapman and Hall,
1996.
[8] Taguchi, G.(1986) Introduction to Quality Engg, Asian Productivity Organization, New York: UNIPUB,
White Plains, 1986.
[9] Tambe, M L.(1992) A Practical Approach to Centrifugal Pump Design, Pune: Brahmalikhit Prakashan
Publishers 1992.
[10] Yoon, E. S., Oh, H. W., Chung M K and Ha, J. S.(1998) Performance prediction of mixed flow
pumps. Proc. Instn. Mech. Engrs, 1998, Vol. 212, Part A, 109-115.

AUTHOR BIOGRAPHY
R. Ragoth Singh is currently pursuing Ph.D degree program from Anna University of Technology,
Coimbatore, India. He is currently Assistant Professor in Department of Mechanical Engineering,
Kathir College of Engineering, Coimbatore. His research interest is in Design and optimization of
fluid flow machines. E-mail: ragothsingh@rediffmail.com

Dr.M. Nataraj, received PhD from Bharathiyar University, Coimbatore. He is curently Associate
Professor of Mechanical Engineering, Government College of Technology, Coimbatore. India.

S-ar putea să vă placă și