Sie sind auf Seite 1von 11

IJE TRANSACTIONS B: Applications Vol. 25, No.

3, (August 2012) 211-221

International Journal of Engineering


Journal Homepage: www.ije.ir

Performance Optimization for Skim Milk Powder Unit of a Dairy Plant Using
Genetic Algorithm
S. Kajal a, *, P. C. Tewari b
a Department of Mechanical Engineering, University Institute of Engineering and Technology, Kurukshetra University, Kurukshetra-136119,

Haryana, India
b Department of Mechanical Engineering, National Institute of Technology, Kurukshetra-136119, Haryana, India

ARTICLE INFO A B S T R A C T

Article history: The paper deals with the performance optimization for skim milk powder unit of a dairy plant at
Received 23 December 2011 National Dairy Research Institute (N. D. R. I.), Karnal using Genetic Algorithm (G. A.). The skim milk
Received in revised form 14 Febuary 2012 powder unit comprises of six subsystems working in series. The failure and repair rates of the
Accepted 17 May 2012 subsystems are taken from maintenance history sheets, which follow the exponential distribution. The
mathematical formulation is carried out using probabilistic approach and the Markov birth – death
process is used to develop the difference differential equations. The steady state availability expression
Keywords:
has been derived using normalizing conditions. The optimal values of failure/repair rates of each
Availability
Markov Process subsystem of the skim milk powder unit have been evaluated using Matlab 7.1 G. A. tool. The steady
Genetic Algorithm state availability obtained from Markov analysis is also compared with the optimal availability
Skim Milk Powder Unit calculated through G. A. tool. So, the findings of the present paper will be highly useful to the plant
management for developing proper maintenance strategies which can be implemented in order to
enhance system performance.

doi: 10.5829/idosi.ije.2012.25.03b.08

1. INTRODUCTION1 complete revolution is made and then scrapped by a


blade. This material is collected and fed into pulverized
NDRI was established in 1996 and is commercially machine which converts it in the fine powder form. This
active research institute at Karnal (India). The plant has fine powder is then packed and stored. Skim milk
the capacity to handle 60,000 liters of milk per day. The powder contains calcium chloride, curic acid and
supply of raw milk is from the model dairy plant and sodium citrate not exceeding 0.3% by weight of the end
institute dairy farm. The fat content in the raw milk is product. The products made in the plant are tested in
weighed and assessed for the SNF (Solid Not Fat) and quality control section by different methods. These
fat contents. Thereafter, the milk undergoes products are then finally packed in hygienic conditions
pasteurization and then cooled to refrigeration in packaging section by means of computerized
temperature (4-5 °C) before it is sent to storage tank. machine.
Further processing involves separation of cream by
cream separator and its standardization for various 1. 1. History Availability of a system can be
purposes. Continuous churning of cream separates improved by several methods like using large safety
butter in a butter making machine. Lassi and Flavored factors, structural redundancy, reducing system
milk is made in this section. In the second section, complexities, practicing a planned maintenance
production of condensed and dried milk products are schedule and repair schedule. Availability optimization
carried out. Fine powder is made with the help of roller problems concentrate on optimal allocation of
drier machine. The milk is fed into the rotating steam- redundancy components and optimal selection of
heating drums or rollers. As the drums revolve, a thin alternative design to meet system requirement. Efforts
film of milk is adhered to it and dried by the time, a were made by some researchers in the direction of
availability optimization and cost analysis. Goldberg [1]
*Corresponding Author Email: Sanjaykajal74@gmail.com (S. Kajal) made a systematic study of G. A. mechanism and
S. Kajal and P. C. Tewari / IJE TRANSACTIONS B: Applications Vol. 25, No. 3, (August 2012) 211-221 212

identified three basic operators: reproduction, crossover reliable models and methods in this area were discussed
and mutation for obtaining near optimal solutions. and future prospects were investigated. Khanduja et
Srinivas [2] discussed various optimization techniques al.[25] discussed the steady state behavior and
and their implementation in the engineering problems. maintenance planning of the bleaching system in a
Ishii et al. [3] applied the reliability optimization as the paper plant. The development of mathematical
last step in their six step heuristic procedure. Sung et al. modeling was carried out using Markov birth-death
[4,5] applied branch and bound approach for reliability process. The literature review discussed here reflects
optimization. Mettas [6] derived a general model to that G. A. has been applied successfully in various
estimate the minimum reliability requirement for complex industries consisting of series and series-
multiple components within a system that would yield parallel system. The approach of performance modeling
the goal reliability value for the system. Kuo et al. [7] and development of decision support system has also
discussed the reliability redundancy allocation problem been discussed by many researchers. Therefore, in order
for finding simultaneously optimal redundancy levels to bridge the gap, G. A. is being proposed in the present
and optimal component reliabilities that maximize paper for performance optimization in dairy industry.
system reliability subject to resource constraints. Coit et The analysis will definitely help the plant management
al. [8] proposed a multiple objective formulation for to adopt the strategic maintenance planning and
maximizing the system availability. Nahas et al. [9] scheduling.
studied the reliability optimization problem for a series
system with multiple choice constraints, to maximize
the system reliability. Gen et al. [10] studied the G. A. 2. SYSTEM CONFIGURATION
approach for various reliability optimization problems.
Castro et al. [11] proposed a maintenance optimization The skim milk powder unit of the dairy plant at N.D.R.I
model for an Engineering system assembled in a series consists of the following six subsystems:
configuration. Dai et al. [12] developed an optimization 1. Pumping subsystem: It is used for unloading the
model for the grid service allocation using G. A.. milk brought from milk collection centers. The
Taboada et al. [13] studied the reliability optimization system consists of two pumps, one in standby
of the complex systems. Juang et al. [14] proposed a G. having perfect switch-over device. This system
A. based optimization model to optimize the availability never fails.
of a series parallel system. Gupta et al. [15,16,17] dealt 2. Preheating subsystem (B1): The subsystem is used
with the development of decision support system and to heat the roller. The steam is allowed to move
performance modeling of feed water unit,coal handling onto the rollers. After drum’s surface achieve a
unit and feed water system of a thermal power plant. temperature high enough to dry the milk film, the
They derived the availability expression for feed water flow is allowed over the entire drum length. This
and coal handling unit with elements exhibiting subsystem is subjected to major failure only.
independent failures and repairs. Amir Azaron et al. 3. Agitator (B2): It works on the centrifugal force
[18] developed a new methodology for reliability principle. Chilled milk from the chiller is taken to
evaluation and optimization of non repairable dissimilar the agitator where fat is separated from the milk in
component with cold standby redundant systems. the form of cream containing 40 to 50% fats. The
Nakagawa et al. [19] analyzed the reliability skimmed milk is stored in milk silos for preparing
optimization problem by using the concept of the milk powder. It consists of three components
reversed failure rate. Garg et al. [20] developed a working in series – motor, bearings and high speed
reliability model of a block-board manufacturing system gear box. Failure of any component causes
in the plywood industry using time dependent and complete failure of the subsystem.
steady state availability under idealized and faulty 4. Pasteurizer (B3): It pasteurizes the milk coming
Preventive Maintenance (PM). Singh et al. [21] from pumping system. Here the cream is heated up
analyzed the mathematical model with the application to 80 to 82˚C with no holding time. The purpose is
of cupola. The transition state probabilities, reliability, to destroy pathogenic and undesirable organisms,
availability and M.T.T.F. were determined using inactivate the enzymes present, make possible
supplementary variable technique. Khanduja et al. [22] removal of volatile flavors and removal of tanning
developed a mathematical model and did performance substances. Failure of machine causes the complete
optimization for the digesting system of a paper plant failure of the unit.
using G. A.. Asha and Nair [23] developed the 5. Drum drier and scraping machine (B4): The drum
relationship between mean time to failure in an age revolves and a thin film of milk adhered to it which
replacement model with hazard rate and mean residual is dried by the time of a complete revolution of
life functions. Vasili et al. [24] presented a brief review drum. The film is made and then is scrapped by a
of existing maintenance optimization models. Several blade. Partial failure of machine can set the system
213 S. Kajal and P. C. Tewari / IJE TRANSACTIONS B: Applications Vol. 25, No. 3, (August 2012) 211-221

to reduced working capacity while major failure IV. The standby units are of the same nature and
causes complete failure of the unit. capacity as the active units.
6. Pulverized machine (B5): The above material is The notations associated with the transition diagram are
collected and fed into pulverized machine which as summarized in Table 1.
converts it to fine powder form. This fine powder is
then packed and stored. Partial failure of machine
can set the unit to reduced working capacity while 4. MATHEMATICAL ANALYSIS OF THE UNIT
major failure causes complete failure of the unit.
The pumping system, as mentioned above, never fails. The transition diagram (Figure 2) depicts a simulation
So, working of the unit is affected by rest of the five model showing all the possible states for skim milk
subsystems. All the above mentioned subsystems, powder unit. The model helps to develop the Chapman-
except subsystem B4 and B5 are subjected to major Kolmogorov differential equations using mnemonic rule
failures while B4 and B5 can work in reduced state also [22]. This rule states that the derivative of the
(Figure 1). probability of every state is equal to the sum of all
probability flows which comes from other states to the
given state minus the sum of all probability flows which
goes out from the given state to the other states.
3. ASSUMTIONS AND NOTATIONS
Probability considerations give the following
differential equations associated with the transition
I. Failure and repair rates for each subsystem are
diagram:
constant and statistically independent [26, 27].
II. There are simultaneous failures occurring at a  11  9
time. P0'  t   
 i  P t  
 0  i Pi  3  t    1 0 P1 0  t 
(1)
III. Performance wise a repaired unit as good as new  i7  i7

one.   1 1 P1 4  t 

 9  9


P1'  t   

 i7
 i  11  12  P1  t  



i7
 i Pi  t   1 0 P0  t 
(2)
 1 1 P18  t 

 10  9
P2'  t   


 i7
 i  13  P2  t  



i7
 i P4  i  t   10 P1 9  t 
(3)
 1 1 P0  t 

 9  9
Figure 1. Schematic flow diagram of skim milk powder unit 
P3'  t   

 i7
 i  12  13  P3  t  



i7
 i P8  i  t   10 P2  t 
(4)
 11 P1  t 
TABLE 1. Notations of skim milk powder unit
Transition diagram Figure 2 Pi '  t    7 Pi  t    7 P j  t  (5)
‘(t) d/dt f o r i = 4 , j = 0 ; i = 7 , j = 1; i = 1 1 , j = 2 ; i =
15, j=3;
Full capacity states (without
B1, B2, B3, B4, B5
standby) Pi '  t    8 Pi  t    8 P j  t  (6)
Failed states b1, b2, b3, b4 ,b5
f o r i = 5 , j = 0 ; i = 8 , j = 1; i = 1 2 , j = 2 ; i =
Reduced capacity B'٤ B and'٥ 16, j=3;
λi, i = 7 to 13
Failure rates Pi '  t    9 Pi  t    9 P j  t  (7)
from B1, B2, B3, B4, B5, B'4,B'5
to b1,b2,b3, B'4,B'5,b4,b5 for i = 6, j=0; i = 9, j=1; i = 13, j=2; i = 17, j=3;
µi, i = 7 to 11
Repair rates
Pi '  t   10 Pi  t    j Pk  t  (8)
from b1,b2,b3, b4,b5 to B1,
B2, B3, B4, B5, for i = 10, j =6,k=1; i =19,j=6, k =3;
Probability of full capacity
working states (without standby P0 Pi '  t   11 Pi  t    j Pk  t  (9)
unit) for i = 14, j =7,k=2; i =18,j=7, k =3;
Probability of reduced capacity
working states
P١ , P٢ P and٣ Initial conditions at time t = 0 are:
Probability of failed states P4 to P19 Pi (t )  1 for i = 0 Pi (t )  0 for i ≠ 0
S. Kajal and P. C. Tewari / IJE TRANSACTIONS B: Applications Vol. 25, No. 3, (August 2012) 211-221 214

Figure 2. Transition diagram of skim milk powder system

4. 1. Steady State Availability The skim milk λ10 λ11 (C1 + C 2 )


powder unit is required to be available for long duration L3 =
C1C 2 C 3 - μ11 H 5 (C1 λ10 + C 2 λ11 )
of time. So, the long run or steady state availability of
the skim milk powder unit is obtained by putting C1  T2  7 H1   8 H 2  9 H 3
d = 0 as t   into differential Equations (1) to (9)
dt C2  T3  7 H1   8 H 2  9 H 3
and solving these equations recursively, we get:
C3  T4  7 H1   8 H 2  9 H 3
P1  L1 P0 P2  L2 P0 P3  L3 P0
11 9
P4  H1 P0 P5  H 2 P0 P6  H 3 P0
T1   i T2     i 11  12
P7  H1L1 P0 P8  H 2 L1 P0 P9  H 3 L1 P0 i 7 i 7

P10  H 4 L1 P0 P11  H1L2 P0 P12  H 2 L2 P0 10 9

P13  H 3L2 P0 P14  H 5 L2 P0 P15  H1L3 P0


T3   i  13 T4    
i 7
i 12  13
i 7

P16  H 2 L3 P0 P17  H 3 L3 P0 P18  H 5 L3 P0 Now, using normalizing conditions i.e. sum of all
the probabilities is equal to one, we get:
P19  H 4 L3 P0 19
where P  1
i 0
i
λj
Hi = ; i = 1, 2, 3; j = 7, 8, 9
μj 1
1  L1  L2  L3  H 1  H 2  H 3  H 1 L1  H 2 L1  
 
λ12 λ13 P0   H 3 L1  H 4 L1  H 1 L2  H 2 L2  H 3 L2 
H4 = H5 =
μ10 μ11   H 5 L2  H 1 L3  H 2 L3  H 3 L3  H 5 L3  H 4 L3 
μ H L +λ μ H L +λ
L1 = 11 5 3 10 L2 = 10 4 3 11 The maximum steady state availability of skim milk
C1 C2
powder unit ( Av1 ) may be obtained as summation of
215 S. Kajal and P. C. Tewari / IJE TRANSACTIONS B: Applications Vol. 25, No. 3, (August 2012) 211-221

probabilities of all working and reduced capacity states TABLE 2. G. A. operators for skim milk powder unit
i.e.: Population type: Double
Reproduction (elite count):2
vector
Av1  P0  P1  P2  P3  1  L1  L2  L3  P0 (10)
Population size: 20 to 160 in Reproduction (crossover
Av1  92.82 % step size of 20 fraction):0.80
Creation function: Uniform Mutation probability: 0.01
Taking λ7=0.00007, µ7=0.009, λ8=0.008, µ8=0.55,
Fitness scaling: Rank Crossover probabilities:
λ9=0.008, µ9=0.05, λ10=0.005, µ10=0.08, λ11=0.00009, Heuristic-0.80
µ11=0.008, λ12=0.005, λ13=0.0002 Selection: Stochastic Migration: Direction- forward
Stopping criteria:
(Data taken from maintenance history sheets) Generation-50 to 450 in step Number of variables:12
The availability ( Av1 ) as given by Equation (10) includes size 50
all possible failure events ( i ) and the identification of Variable Range (Min. to Max.)
all the courses of action i.e. repair priorities ( i ). This λ7 0.00007 to 0.00011
model can be used to adopt strategic maintenance µ7 0.009 to 0.045
policies for skim milk powder unit. Similarly, the λ8 0.008 to 0.016
various availability levels may be computed for µ8 0.11 to 0.55
different combinations of failure and repair rates. The λ9 0.0008 to 0.0012
optimal availability value for skim milk powder unit is µ9 0.05 to 0.25
further evaluated using G. A. technique.
λ10 0.005 to 0.009
µ10 0.08 to 0.40
4. 2. Performance Optimization Using Genetic
λ11 0.00009 to 0.00013
Algorithm G. A. is a powerful computerized tool for
search and optimization problems. The G. A. optimizes µ11 0.008 to 0.04
the results using the following steps: λ12 0.005 to 0.009
1) Identify all the constraints of the G. A. tool. λ13 0.0002 to 0.001

2) Randomly generate the initial population and


prepare the coded strings.
3) Compute the fitness of each individual and form the 5. RESULTS AND DISCUSSION
mating pool from the old population.
The maximum steady state availability as calculated by
4) Divide the population repeatedly into random Markov analysis comes out to be 92.82% by using the
tournaments. failure and repair rates available through maintenance
5) Determine the new generation pool and select two history sheets. A G. A. is a search strategy that employs
parents from the mating pool randomly. random selection to guide a highly exploitative search,
6) Perform the crossover of the parents to produce two striking a balance between explorating of feasible
off springs. domain and exploting of good solutions. The
performance optimization is carried out by varying the
7) Mutate, if required.
number of generations from 50 to 450 in step size 50 as
8) Place the child strings to new population and shown in Table 3. It is observed that the optimum
compute the fitness of each individual. availability of skim milk powder unit of an N.D.R.I.
9) Replace old population with new population. plant comes out to be 97.52 % at 350 number of
10) Repeat the steps 3 to 6 until the best individuals in generation. The corresponding values of failure and
new population represent the optimum value of the repair rates are λ7 = 0.00007 (once in 14286 h), µ7 =
performance function. 0.0449 (once in 22.2 h), λ8 = 0.008 (once in 125 h), µ8 =
0.5461 (once in 1.83 h), λ9 = 0.0008 (once in 1250 h),
The performance optimization of the skim milk µ9 = 0.2448 (once in 4.08 h), λ10 = 0.005 (once in 200
powder unit is influenced by the failure and repair rates h), µ10 = 0. 3989 (once in 2.5 h), λ11= 0.00012 (once in
of each subsystem. The designed range of failure and 8333.3 h), µ11=0.029 (once in 34.48 h),λ12 = 0.005(once
repair rates are taken from the maintenance history in 200 h), λ13 = 0.0002 (once in 5000 h), which is the
sheets. In order to optimize the availability, G. A. tool in best possible combination of failure and repair rates of
Matlab 7.1 is being used in the present paper. The different subsystems. The effect of number of
following G. A. operators along with permissible range generations on the availability of skim milk powder unit
of failure and repair rates are used for the performance is shown in Figure 3.
optimization. Now the performance optimization is carried out by
S. Kajal and P. C. Tewari / IJE TRANSACTIONS B: Applications Vol. 25, No. 3, (August 2012) 211-221 216

varying the population size from 20 to 160 in step size 0.005, µ10 = 0. 3913, λ11= 0.0001, µ11=0.039, λ12 =
of 20, as shown in Table 4. 0.005, λ13 = 0.0002, at which optimum value of system
It is evident that the best possible combination of failure availability is 97.49% with population size 100. The
and repair parameters are λ7 = 0.00007, µ7 =0.0401, λ8 effect of population size on system availability is shown
= 0.008, µ8 = 0.5409, λ9 = 0.0008, µ9 =0.2426, λ10 = in Figure 4.

TABLE 3. Effect of number of generations on the availability of skim milk powder unit

TABLE 4. Effect of population size on availability of skim milk powder unit

Figure 3. Effect of number of generation on the availability of Figure 4. Effect of population size on the availability of skim
skim milk powder system milk powder system
217 S. Kajal and P. C. Tewari / IJE TRANSACTIONS B: Applications Vol. 25, No. 3, (August 2012) 211-221

TABLE 5. Comparison in steady state availability and  The comparision between steady state availability
optimized availability using genetic algorithm and optimized availability using G. A. is shown in
System Steady state Optimized availability Table 5.
availability using genetic algorithm
The major implication lies in the fact that the failure
Milk skim 92.82% 97.52% and repair rates are assumed to be constant which
powder unit λ7=0.00007, λ7 = 0.00007, follow the exponential distribution in order to simplify
µ7=0.009, λ8=0.008, µ7 = 0.0449,
µ8=0.55, λ9=0.008,
the problems.
λ8 = 0.008,
µ9=0.05, λ10=0.005,
µ10=0.08, µ8 = 0.5461,
λ11=0.00009, λ9 = 0.0008,
µ11=0.008, µ9 = 0.2448, 7. ACKNOWLEDGEMENT
λ12=0.005, λ10 = 0.005,
λ13=0.0002
µ10 = 0. 3989, The author is grateful to Dr. A. P. S. Chauhan, Principal
λ11= 0.00012, µ11=0.029, Scientist, N.D.R.I., Karnal regarding the concerned
λ12 = 0.005, research work.
λ13 = 0.0002

8. REFERENCES
6. CONCLUSION
1. Goldberg, D.E., “Genetic Algorithm in Search, Optimization and
The mathematical formulation has been carried out Machine Learning”, Addison-Wesley, Reading, MA, (1989).
using probabilistic approach and Markov birth–death 2. Srinivas, N. and Deb, K., “Multi Objective Optimization Using
process is used to develop the system performance Non Dominated Sorting in Genetic Algorithms”, Evolution of
model in terms of availability expression. Then, the Competition, Vol. 2, (1995), No. 3, 221-248.
system availability of the skim milk powder unit is 3. Ishii, N., Fuchino, T. and Muraki, M., “Life Cycle Orientated
being optimized in the paper. The derived results from Process Synthesis at Conceptual Planning Phase”, Computers
and Chemical Engineering, Vol. 21, (1997), 953-958.
the analysis were discussed with plant management and
found to be useful for the management in the following 4. Sung, C.S. and Lee, H.K., “A Branch and Bound Approach for
Spare Unit Allocation in a Series System”, European Journal
ways: Operation Research, (1994), 217-232.
 The steady state availability can be determined at 5. Sung, C.S. and Cho, Y.K., “Reliability Optimization of a Series
System with Multiple-Choice and Budget Constraints”,
any instant for different combinations of failure and European Journal Operation Research, (2000), 159-171.
repair rates of each subsystem taken from
6. Mettas, A., “Reliability Allocation and Optimization for
maintenance history sheets. Complex Systems”, Proc. Annual Reliability and
Maintainability Symposium, (2000), 216-221.
 The optimized availability value comes out to be
7. Kuo, W., Prassad, V.R., Tillman, F.A., Hwang, C., “Optimal
97.52% by selecting various feasible values of Reliability Design”, Cambridge University Press, (2001).
failure and repair rates. The selected values of these 8. Coit, D.W. and Jin, T., Wattanapongsakorn, N., “System
parameters must be maintained by the plant Optimization Considering Component Reliability Estimation
management. Uncertainty: A Multi-Criteria Approach”, IEEE Transactions on
Reliability, Vol. 53, (2004), No. 3, 369-380.
 The corresponding failure rates can be maintained 9. Nahas, N. and Nourelfath, M., “Artificial Neural Networks for
by using redundancy, scheduled preventive Reliability Maximization Under Budget and Weight
maintenance and good inherent design etc. Constraints”, Journal of Quality in Maintenance Engineering,
Vol. 11, (2005), No. 2, 139 – 151.
 The optimal repair rates can be achieved by 10. Gen, M. and Yun, Y., “Soft Computing Approach for Reliability
enhancing the skill of maintenance personnel Optimization: State-of-the-Art Survey”, Reliability Engineering
through training, employing sufficient workers and and System Safety, Vol. 91, (2006), 1008-1026.
motivating them to attain the target. 11. Castro, H.F.D. and Cavalca, K.L., “Maintenance Resources
Optimization Applied to a Manufacturing System”, Reliability
 The availability has been increased by 10.95% using Engineering and System Safety, Vol. 91, (2006), 413-420.
G. A.. This implies that mean time between failures 12. Dai, Y.S. and Wang, X.L., “Optimal Resource Allocation on
of the unit has been increased from 339 days to 356 Grid Systems for Maximizing Service Reliability Using a
Genetic Algorithm”, Reliability Engineering and System Safety,
days. Vol. 91, (2006), 1071-1082.
 The effect of number of generation and population 13. Taboada, H.A., Espiritu, J.F. and Coit, D.W., “A Multi-
size on unit availability is also discussed in the Objective Multi-State Genetic Algorithm for system reliability
optimization design problems”, IEEE Transactions on
paper. Reliability, Vol. 57, (2008), No. 1, 182-191.
S. Kajal and P. C. Tewari / IJE TRANSACTIONS B: Applications Vol. 25, No. 3, (August 2012) 211-221 218

14. Juang, Y.S., Lin, S.S. and Kao, H.P., “A Knowledge Repairmen”, IJE Transition A: Basics, Vol. 24, No. 4 (2011),
Management System for Series Parallel Availability 395-401.
Optimization and Design”, Journal of Expert System with 22. Khanduja, R., Tewari, P.C. and Kumar, D., “Mathematical
Application, Vol. 34, (2008), 181-193. Modeling and Performance Optimization for the Digesting
15. Gupta, S., Tewari, P.C. and Sharma, A.K., “Performance System of a Paper Plant”, IJE Transition A: Basics, Vol. 23, No.
Modeling and Decision Support System of Feed Water Unit of a 3 & 4 (2010), 215-225.
Thermal Power Plant”, South African Journal of Industrial 23. Asha, G. and Nair, N., “Reliability Properties of Mean Time to
Engineering, Vol. 19, (2008), No. 2, 125-134. Failure in Age Replacement Models”, International Journal of
16. Gupta, S., Tewari, P.C. and Sharma, A.K., “A Markov Model for Reliability, Quality and Safety Engineering (IJRQSE), Vol.17,
Performance Evaluation of Coal Handling Unit of a Thermal (2010), No. 1, 15-26.
Power Plant”, Journal of Industrial and Systems Engineering 24. Vasili, M., Hong, T.S., Ismail, N. and Vasili, M., “Maintenance
(JISE), Vol. 3, (2009), No. 2, 85-96. optimization models: a review and analysis”, Proceedings of
17. Gupta, S. and Tewari, P.C., “Performance Modeling of Power the International Conference on Industrial Engineering and
Generation System of a Thermal Plant”, IJE Transition A: Operations Management, Kuala Lumpur, Malaysia, January 22 –
Basics, Vol. 24, No. 3 (2011), 239-248. 24, (2011), pp. 1131-1138.
25. Khanduja, R., Tewari, P.C. and Kumar, D., “Steady State
18. Azaron, A., Perkgoz, C., Katagiri, H., Kato, K. and Sakawa, M.,
Behaviour and Maintenance Planning of Bleaching System in a
“Multi-Objective Reliability Optimization for Dissimilar-Unit
Paper Plant”, Journal Industrial Engineering International, Vol.
Cold-Standby Systems Using a Genetic Algorithm”, Journal
7, No. 12 (2011), 39-44.
Computers and Operations research, Vol. 36, (2009), No. 5,
1562-1571. 26. Vora, Y., Patel, M.B. and Tewari, P.C., “Simulation Model for
Stochastic Analysis and Performance Evaluation of Steam
19. Nakagawa, T. and Mizutani, S., “Optimum Problems in
Generation System of a Thermal Power Plant”, International
Backward Times of Models”, IEEE Transactions on Reliability,
Journal of Engineering Science and Technology, Vol. 3, No. 6,
Vol. 41, (2009), 65-71.
(2011), 5141-5149.
20. Garg, S., Singh, J. and Singh, D.V., “Availability and
27. Kumar, R., Sharma, A.K. and Tewari, P.C., “Performance
Maintenance Scheduling of a Repairable Block-Board
Modeling of Furnace Draft Air Cycle in a Thermal Power
Manufacturing System”, International Journal of Reliability and
Plant”, International Journal of Engineering Science and
Safety, Vol. 4, (2010), No.1, 104 - 118.
Technology, Vol. 3, No. 7, (2011), 6792-6798.
21. Singh, S.B., Ram, M. and Chaube, S., “Analysis of the
Reliability of a Three- Component System with Two
219 S. Kajal and P. C. Tewari / IJE TRANSACTIONS B: Applications Vol. 25, No. 3, (August 2012) 211-221

APPENDIXES

Effect of number of generations on the availability of skim milk powder unit


S. Kajal and P. C. Tewari / IJE TRANSACTIONS B: Applications Vol. 25, No. 3, (August 2012) 211-221 220

Effect of population size on the availability of skim milk powder unit

Population Size =40 Population Size =60 Population Size =80

Population Size =100 Population Size =120 Population Size =140

Population Size =160


‫‪221‬‬ ‫‪S. Kajal and P. C. Tewari / IJE TRANSACTIONS B: Applications Vol. 25, No. 3, (August 2012) 211-221‬‬

‫‪Performance Optimization for Skim Milk Powder Unit of a Dairy Plant Using‬‬
‫‪Genetic Algorithm‬‬
‫‪S. Kajal a, P. C. Tewari b‬‬
‫‪a Department of Mechanical Engineering, University Institute of Engineering and Technology, Kurukshetra University, Kurukshetra-136119,‬‬

‫‪Haryana, India‬‬
‫‪b Department of Mechanical Engineering, National Institute of Technology, Kurukshetra-136119, Haryana, India‬‬

‫‪ARTICLE INFO‬‬ ‫ﭼﮑﯿﺪه‬

‫‪Article history:‬‬
‫‪Received 23 December 2011‬‬
‫اﯾﻦ ﻣﻘﺎﻟﻪ‪ ،‬ﺑﻬﯿﻨﻪ ﺳﺎزي ﻋﻤﻠﮑﺮد واﺣﺪ ﺷﯿﺮ ﺧﺸﮏ ﺑﺪون ﭼﺮﺑﯽ از ﮐﺎرﺧﺎﻧﻪ ﻫﺎي ﻟﺒﻨﯽ ﺑﺎ اﺳﺘﻔﺎده از اﻟﮕﻮرﯾﺘﻢ ژﻧﺘﯿﮏ )‪(G. A.‬‬
‫‪Received in revised form 14 Febuary 2012‬‬ ‫در ﻣﻮﺳﺴﻪ ﻣﻠﯽ ﺗﺤﻘﯿﻘﺎت ﻟﺒﻨﯽ )‪ (N.D.R.I‬را ﻣﻮرد ﺑﺮرﺳﯽ ﻗﺮار ﻣﯽ دﻫﺪ‪ .‬ﮐﺎرﺧﺎﻧﻪ ﺑﻪ ﺑﺨﺶ ﻫﺎي زﯾﺎدي از ﺟﻤﻠﻪ واﺣﺪ‬
‫‪Accepted 17 May 2012‬‬ ‫ﺑﺎزارﯾﺎﺑﯽ‪ ،‬واﺣﺪ ﺗﻐﻠﯿﻆ و ﺧﺸﮏ ﮐﺮدن‪ ،‬واﺣﺪ ﭘﻨﯿﺮ و ﮐﺮه‪ ،‬واﺣﺪ ﮐﻨﺘﺮل ﮐﯿﻔﯿﺖ‪ ،‬واﺣﺪ ﭘﻮدر ﺷﯿﺮ و واﺣﺪ ﺑﺴﺘﻪ ﺑﻨﺪي ﺗﻘﺴﯿﻢ ﻣﯽ‬
‫ﺷﻮد‪ .‬واﺣﺪ ﭘﻮدر ﺷﯿﺮ ﺑﺪون ﭼﺮﺑﯽ‪ ،‬ﺷﺎﻣﻞ ﺷﺶ زﯾﺮ ﻣﺠﻤﻮﻋﻪ اﺳﺖ‪ .‬ﻧﺮخ ﺷﮑﺴﺖ و ﺗﺮﻣﯿﻢ از اﯾﻦ زﯾﺮ ﻣﺠﻤﻮﻋﻪ ﻫﺎ از ﺗﺎرﯾﺦ‬
‫‪Keywords:‬‬ ‫ﻫﺎي ﺗﻌﻤﯿﺮ و ﻧﮕﻬﺪاري ﮐﻪ ﺑﻪ دﻧﺒﺎل ﺗﻮزﯾﻊ ﻧﻤﺎﯾﯽ ﮔﺮﻓﺘﻪ ﺷﺪه اﺳﺖ ﺗﺒﻌﯿﺖ ﻣﯽ ﮐﻨﺪ‪ .‬ﺑﻪ ﻣﻨﻈﻮر ﺗﻮﺳﻌﻪ ﺗﻔﺎوت ﻣﻌﺎدﻻت‬
‫‪Availability‬‬
‫‪Markov Process‬‬ ‫دﯾﻔﺮاﻧﺴﯿﻞ‪ ،‬ﻓﺮﻣﻮل رﯾﺎﺿﯽ ﺑﺎ اﺳﺘﻔﺎده از روش اﺣﺘﻤﺎﻻﺗﯽ و ﺗﻮﻟﺪ ﻣﺎرﮐﻒ ‪ -‬ﻓﺮآﯾﻨﺪ ﻣﺮگ ﺑﻪ دﺳﺖ ﻣﯽ آﯾﺪ‪ .‬اﯾﻦ ﻣﻌﺎدﻻت ﺑﺎ‬
‫‪Genetic Algorithm‬‬ ‫اﺳﺘﻔﺎده از ﺷﺮاﯾﻂ ﻋﺎدي ﺣﻞ ﺷﺪ ﺗﺎ ﻓﺮﻣﻮﻟﻪ ﮐﺮدن ﺷﺮاﯾﻂ ﭘﺎﯾﺪار ﺗﺪوﯾﻦ ﮔﺮدد‪ .‬ﻧﺘﺎﯾﺞ ﺑﻬﯿﻨﻪ ﻧﺮخ ﺷﮑﺴﺖ ‪ /‬ﺗﻌﻤﯿﺮ ﻫﺮ ﯾﮏ از زﯾﺮ‬
‫‪Skim Milk Powder Unit‬‬
‫ﻣﺠﻤﻮﻋﻪ ﻫﺎي واﺣﺪ ﺷﯿﺮ ﺧﺸﮏ ﺑﺪون ﭼﺮﺑﯽ ﺑﺎ اﺳﺘﻔﺎده از ﻧﺮم اﻓﺰار ‪ MATLAB 7.1‬و اﺑﺰار اﻟﮕﻮرﯾﺘﻢ ژﻧﺘﯿﮏ ﻣﻮرد ﺑﺮرﺳﯽ‬
‫ﻗﺮار ﮔﺮﻓﺘﻪ اﺳﺖ‪ .‬ﺷﺮاﯾﻂ ﭘﺎﯾﺪار ﺑﻪ دﺳﺖ آﻣﺪه از ﺗﺠﺰﯾﻪ و ﺗﺤﻠﯿﻞ ﻣﺎرﮐﻒ ﻧﯿﺰ ﺑﺎ در دﺳﺘﺮس ﺑﻮدن ﺑﻬﯿﻨﻪ ﻣﺤﺎﺳﺒﻪ ﺷﺪه از ﻃﺮﯾﻖ‬
‫اﺑﺰار ‪ G.A.‬ﻣﻘﺎﯾﺴﻪ ﺷﺪه اﺳﺖ‪ .‬ﺑﻨﺎﺑﺮاﯾﻦ‪ ،‬ﯾﺎﻓﺘﻪ ﻫﺎي اﯾﻦ ﻣﻘﺎﻟﻪ ﺑﺮاي ﻣﺪﯾﺮﯾﺖ ﮐﺎرﺧﺎﻧﻪ و ﺗﻮﺳﻌﻪ اﺳﺘﺮاﺗﮋي ﻫﺎي ﻧﮕﻬﺪاري‬
‫ﻣﻨﺎﺳﺐ ﺑﺴﯿﺎر ﻣﻔﯿﺪ ﻣﯽ ﺑﺎﺷﺪ و ﻣﻮﺟﺐ اﻓﺰاﯾﺶ ﻋﻤﻠﮑﺮد ﺳﯿﺴﺘﻢ ﻣﯽ ﮔﺮدد‪.‬‬

‫‪doi: 10.5829/idosi.ije.2012.25.03b.08‬‬