Bho Sale 2018

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Available online at www.sciencedirect.

com

ScienceDirect
Materials Today: Proceedings 5 (2018) 7160–7167 www.materialstoday.com/proceedings

IMME17

Material Flow Optimisation of Flexible Manufacturing System


using Real Coded Genetic Algorithm (RCGA)
Bhosale K.C*, Pawar P.J.
Department of Production Engineering, K.K. Wagh Institute of Engineering Education and Research,
Nashik, Savitribai Phule Pune University, Pune, Maharashtra, India

Abstract

In flexible jobshop manufacturing system ‘n’ number of jobs are manufactured on ‘m’ machines in the‘s’ stages. Each
product has its predefined sequence. Material processing is scheduled by following alternate process plans. Material flow consists
of the total completion time which comprises of travel time, waiting time and processing time, etc. Material flow optimisation is
the part of production planning stage. So, it is a crucial part in the flexible manufacturing system of decision making to optimise
a material flow in the system.In this paper, a mathematical model is modified to optimise material flow optimisation of a flexible
manufacturing system. Waiting time due to unavailability of machines is considered for the modification of this mathematical
model. A real coded Genetic Algorithm (RCGA) has implemented on flexible jobshop manufacturing system problem to
optimise material flow of the system. The results shows that RCGA can be applied to solve problems of flexible jobshop
manufacturing system and better as compared to Genetic Algorithm (GA).
© 2017 Elsevier Ltd. All rights reserved.
Selection and/or Peer-review under responsibility of International Conference on Emerging Trends in Materials and
Manufacturing Engineering (IMME17).

Keywords: Flexible Manufacturing system, Material flow, waiting time, Real Coded Genetic Algorithm (RCGA)

1.Introduction:
Advancements in technology, fluctuations in demand from customers and competition in the market has increased
the need to use flexible manufacturing systems which can cater the need of customers from large quantities of a
single product to a mixed variety of products within a short span of time. Thus, FMS is used because of advantages
such as high quality, low inventory cost, low labour cost, rapid changeovers etc. To achieve these advantages,
material flow of a product in the system must be minimum. Material flow depends on material handling i.e.

* Corresponding author. Tel.: 9422319829, 9623960142;


E-mail address: bhosale_kailash@yahoo.co.in

2214-7853 © 2017 Elsevier Ltd. All rights reserved.


Selection and/or Peer-review under responsibility of International Conference on Emerging Trends in Materials and Manufacturing Engineering
(IMME17).
Bhosale et al., / Materials Today: Proceedings 5 (2018) 7160–7167 7161

transportation of a product from one point to another point. Material handling (MH) is responsible for 30–75% of
the total cost of a product and efficient material handling is responsible for reducing the manufacturing system
operations cost by 15–30% [8, 13].
The issues observed in the FMS are such as to maximize system/machine utilizations, meeting due dates,
maximizing production rates, minimizing in-process inventories, minimizing setup and tool changes times,
minimizing mean flow times and balancing machine utilizations.
So, the following objectives are of most important in the context of an FMS [2]:
(a) Maximizing system/machine utilizations
(b) Meeting due dates
(c) Maximizing production rates
(d) Minimizing in-process inventories
(e) Minimizing setup and tool changes times
(f) Minimizing mean flow times
(g) Balancing machine utilizations
Researchers have made the attempts to handle above issues. [9] has considered single objective to minimise
vehicle travel minutes[14] has considered the objective is to minimise the total cost and aisle space requirements, [4]
considered single objective as minimizing total material flow times distance, [12] considered objective function as to
minimise the total flow distance[13] considered two objectives such as to minimize the costs related to the
manufacturing operations, machine set ups, and MH operations. [3,10] considered the objective to minimise
material flow. But, they have not considered a waiting time for the AGV if processing at the station is not
completed. Ramalingam et al. [10] has considered objective of minimization of material flow time and completion
time. But, total completion time is composed of processing time, transfer time and waiting time. So, if the total
completion time is minimised then the objective will be satisfied.
Here, in this paper, a flexible job shop manufacturing system is considered in which automated guided
vehicles (AGVs) are used to carry raw materials, semi-finished and final products in batch sizes. The objective is to
minimise total completion time which comprises of travel time, waiting time and processing time. So, to consider
waiting time of AGV if the process is not completed on the respective machine, mathematical model of [3] has been
modified. An attempt has made to implement Real Coded Genetic Algorithm (RCGA) to optimise material flow of
the flexible manufacturing system. The results have compared with Genetic Algorithm (GA) proposed by
Ramalingam et al. [10] and found that Real coded Genetic Algorithm (RCGA) is giving good results as compared to
Genetic Algorithm (GA).
Material flow optimisation depends on material handling and buffer storage system. So, in this section literature
review based on material handling system, material flow in the manufacturing system, material flow design and
vehicle routing problem is carried out.
Malmborg [9] has considered material transportation and handling cost with buffer storage to formulate a
mathematical model with single objective to minimise vehicle travel minutes and constraints such as to minimise
waiting time, minimise output overflow, minimise number of vehicles needed and to minimise average total space
required. This model was used to find the number of vehicles required for material flow, average space utilization
and average percentage loss due to queuing time. This model used a discrete handling system in which fixed number
of vehicles are used for transferring unit loads between WorkCentre. Welgama and Gibson [14] considered a hybrid
knowledge based optimisation system for the selection of material handling system. The parameters used for
selection are path, flow volume, unit size, distance, palletization and accumulation. The objective is to minimise the
total cost and aisle space requirements. The constraints used are feasibility constraint and utilization constraint. A
knowledge based system is developed to obtain a feasible set of material handling equipment for each move then an
optimisation algorithm is used to determine the optimum material handling equipment for all moves using a system
approach. However, the knowledge base built is very simple. Hermann et al. [4] proposed a mathematical model for
design of material handling flow paths based on single objective as minimizing total material flow times distance
and 5 constraints such as continuity of flow path between origin-destination pair, traffic congestion, capacity of
material handling system on the arc, vehicle collision avoidance and prohibition of flow through non selected arcs.
This problem was optimised by two heuristic methods. The first method, Fixed charge adjustment heuristic (FCAH)
7162 Bhosale et al., / Materials Today: Proceedings 5 (2018) 7160–7167

iteratively adjusts the arc fixed costs and selectively included arcs in the flow network until a feasible solution is
obtained. The second method state space search heuristic (SSH) partially explores the problem state space.
Hermann et al. [5], considered the same single objective function with 4 constraints such as waiting times due to
congestion or on-line scheduling decisions, each loaded move should have exactly one transporter, continuity on the
path consisting of loaded moves performed by each transporter and sub tour elimination. Two heuristics are
proposed to optimise the problem. First GREEDY heuristic and another an assignment/ bin packing composite
heuristic. Sinriech and Samakh [12] considered a segmented flow based material handling system optimisation
problem of pick up/ delivery location. They considered objective function as to minimise the total flow distance.
Genetic algorithm is used for optimisation.
Schaller et al. [11] considered scheduling problem of sequence dependent set up time of flowline
manufacturing cell. Objective is to minimise makespan of part processing. First, lower bounds are found out by
branch and bound method. Then these bounds are used for optimisation by descent heuristic which systematically
explores solutions in the neighborhood of current solution. However, they have not considered the precedence
constraints and resource restrictions. Kaspi et al. [7] proposed a mathematical model for flow path layout design of
balanced unidirectional AGV system with the objective to minimize the total distance that both empty and loaded
vehicles have to traverse to satisfy the total transportation requirements with the constraints of job allocation to each
point, and balancing inflow and outflow. They proposed branch and bound method for optimisation. The unknown
flow of unloaded vehicles and given flow of loaded vehicles are discussed. However, problems due to buffer storage
and congestion has not addressed.
Fazlollahtabar et al.[3], Ramalingam et al.[10] considered material flow optimisation of a flexible flowshop of
automated manufacturing systems in which automated guided vehicles (AGVs) are used to carry raw materials,
semi-finished and final products in batch sizes. A mathematical model is proposed for the material flow
optimisation with the constraints such as completion time, job allocation, cycle time and AGV capability.
However, all the researchers have not considered the waiting time due to unavailability of the machines. If the
machine is not available for processing the new job, then we have wait till the previous operation on the machine is
completed to assign new job on it. So, here in this paper, an attempt has made to consider the waiting time of the
machines due to unavailability to process the new job and optimised the material flow of the system.

2. Mathematical model for flow optimization:

In this section, mathematical model of Fazlollahtabar et al.[3], Ramalingam et al.[10] which was used
for the flexible job shop manufacturing system problem is modified to consider waiting time of the machines due to
unavailability for processing new job. The indices, parameters, and decision variables are as follows:
Indices:
m Index for shops, m=11,12,…, M.
i Index for products, i=1,2,…,I.
p Index for job position, p=1,2,…,P.
Parameters:
Cipm Completion time of product ith in position pth in mth shop,
P.Timp Processing time of shop mth for product ith in the pth position,
T.Tim(m-1) Transferring time from shop mth to shop (m-1)th of position p,
VAGV Velocity of AGV
fim(m-1) Flow (distance) for product ith between shop mth and shop (m-1)th of product p,
Wimp Waiting time for product ith in shop mth of position p,
C.Ti Cycle time for product ith,
T Total working time in each day,
W.Timp Waiting time for product ith to (i-1)th in shop mth of position p,
0 Otherwise
Zipm   th th th
1 if shopm is allocatedfor producti in positionp
Bhosale et al., / Materials Today: Proceedings 5 (2018) 7160–7167 7163

Decision variable:

0 Otherwise
Z ipm   th th th
1 if shop m is chosen for product i in position p
Objective Function:
M P I
Minimise  Z
m 1 p 1 i 1
ipm .Z i ( p 1) m .T .Timp (1)

Subjected to:
I P M M P I
Cipm   (Z ipm ( P.Tipm  Wimp ))   Z ipm .Z i ( p1) m .T .Timp 
i 1 p 1 m1 m1 p 1 i 1
M P I (2).. i, m, p
 Z
m1 p 1 i 1
ipm .Z i ( p1) m .W.Timp

Cipm  Ci ( p 1)( m1)  P.Tipm  Wim p  Tim ( m1) (3) i , m , p


Z ipm .Cipm  C.Ti i , m , p (4)
C ipm  T (5)
f im(m-1)
T.Tm(m-1)  VAGV= 4……… ..(6)
VAGV
Equation (1) is the objective function of the proposed problem which minimizes the material flow. The
output of the objective function is to select the types of the shops which minimize the total material flow. Equation
(2) indicates the computation of completion time for each product. Equation (3) certifies that the differences
between completion time of product ith in position pth in shop nth of type mth is larger than or equal to the addition
of the processing time and waiting time of shop nth of type mth and the transferring time between shop nth of type
mth and shop hth of type kth. Equation (4) guarantees that if a shop is chosen then the corresponding completion
time for each product is lower than or equal to the cycle time. Equation (5) certifies that the completion time for any
product in any position in any shop is lower than or equal to the total working time in each day. Equation (6)
indicates the transferring time between two shops is related directly to the flow of each product.

3. Problem descriptions and assumptions


Here, a flexible job shop manufacturing system as shown in figure 1, which employs an AGV for material
handling is considered Fazlollahtabar et al.[3], Ramalingam et al.[10] .The AGV is used to carry raw material,
semi produced and final products in batch sizes. In this process, three products are processed in four types of shops
with their predefined sequence the three jobs are independent.. The sequence of Product one is 2-4-3-1, for product
two is 1-2-3-4 and for product three is 3-1-4-2. The distances between two shops is given in table no. 1, process time
of each product is given in table no.2 and The waiting times of each product in the shops is given in table no. 3. The
objective is minimise the material flow i.e. to find the minimum completion time for all the three products in the
system. The completion time which depends on the distances between two shops from which total travel time can be
computed, processing time and waiting time for the AGV if operation is not completed in the shop.

4. Real coded genetic algorithm (RCGA)


Genetic algorithms have been developed by John Holland which is based on mechanics of natural selection
and natural genetics. It depends on the principle of survival of fittest. First, a few set of real coded strings are
generated randomly on which objective function depends. In each next generation, new set of strings are created
from old set which are stronger than old sets of strings.
7164 Bhosale et al., / Materials Today: Proceedings 5 (2018) 7160–7167

Real coded genetic algorithm (RCGA) composed of three operators:


1. Reproduction
2. Crossover
3. Mutation

4.1 Reproduction
Reproduction is a process in which individual strings are copied according to their objective function values.
This selection is based on Darwinian’s survival of fittest principle. For the maximization function, the objection
functions having higher values are having higher probability to be selected in the next generation. The selection of
offspring is carried away by roulette wheel method. Selection is based on roulette wheel selection method. The
probability selected is 10%.
Table 1. Distance between shops Fazlollahtabar et al.[3]
from↓, To→ 11 12 13 21 22 31 32 33 41 42
11 1,000 1,000 1,000 6 20 12 30 14 18 26
12 1,000 1,000 1,000 18 24 12 6 18 6 30
13 1,000 1,000 1,000 6 12 12 6 6 18 18
21 6 18 23 1,000 1,000 6 24 14 12 23
22 20 24 12 1,000 1,000 26 18 6 30 6
31 12 12 14 6 26 1,000 18 1,000 6 18
32 30 6 6 24 18 1,000 1,000 1,000 12 24
33 20 18 6 20 6 1,000 1,000 1,000 24 12
41 18 6 18 12 30 6 12 24 1,000 1,000
42 26 33 18 26 6 32 24 12 1,000 1,000

Table No.2. Process time Fazlollahtabar et al.[3],

11 12 13 21 22 31 32 33 41 42

Product 1 5 4 6 3 5 8 7 5 12 15
Product 2 6 7 2 4 5 8 11 4 10 15
Product 3 5 6 6 3 2 10 7 7 9 15

Table No.3. Waiting time Fazlollahtabar et al.[3],

11 12 13 21 22 31 32 33 41 42

Product 1 6 3 4 5 8 10 9 6 12 14
Product 2 7 7 4 8 6 11 13 4 12 7
Product 3 6 8 6 5 9 8 12 8 7 11

4.2 Crossover
After reproduction, simple crossover is carried out in two steps. First, two members are selected randomly from
the reproduced offspring. The crossover is performed as shown in Table 4. . A random number is generated between
1 and 11. Here we get random number as 4. So, new strings called as child strings are created by swapping the real
numbers between 5-12.
Bhosale et al., / Materials Today: Proceedings 5 (2018) 7160–7167 7165

Table no 4. Crossover of the selected strings


Parent 1
21 41 33 13 12 22 32 41 33 12 42 21
Parent 2
21 42 32 13 13 22 31 41 32 12 41 22

Child 1
21 41 33 13 13 22 31 41 32 12 41 22

Child 2
21 42 32 13 12 22 32 41 33 12 42 21

4.3 Mutation
After crossover, mutation is performed on the string. A random site is selected for the mutation. Mutation
probability is taken as 0.01.mutation is performed on the site number 4 as given in Table 5.
Table no 5. .Mutation of the selected strings
21 42 32 13 12 22 32 41 33 12 42 21

21 42 32 12 12 22 32 41 33 12 42 21

5. Result and Discussion

In this problem, Real coded Genetic Algorithm is used to optimise the material flow problem. The
optimised path for product number one is 21-41-32-12, for product number two is 13-22-33-42 and for product
number three is 33-13-41-21. The total completion time is 179 as shown in table no. 4 which is less as compared to
180 of Ramalingam et al.[10] and 186 of Fazlollahataber et al. [3].

Table no.6. Optimum value of the material flow [3]


Product1 Product2 Product3 Total
21 41 31 11 13 22 33 42 32 12 41 21
Processing time 8 24 18 11 6 11 8 22 19 14 16 8
Distance 0 12 6 12 0 12 6 12 0 6 6 12
travel time 0 3 1.5 3 0 3 1.5 3 0 1.5 1.5 3 21
waiting time 0 0 0 0 0 0 0 0 0 0 0 0
Start 0 11 36.5 57.5 0 9 21.5 32.5 0 20.5 36 55
Start+waiting 0 11 36.5 57.5 0 9 21.5 32.5 0 20.5 36 55
Finish 8 35 54.5 68.5 6 20 29.5 54.5 19 34.5 52 63 186

A new developed model has been applied to the optimal solution of Fazlollahtabar et al.,[3] as shown in table 6, and
found that there is no waiting time in the optimal solution. Similarly, the optimal solution obtained by Ramalingam
et al. [10] as shown in table 7 also do not have waiting time. But, authors have considered waiting time and it is
observed in the product 3 as shown in table 8 that for shop 41 there is a waiting of 2 minutes. It means that though
the parts are available at 33 minutes, but the machine is still busy in processing previous part up to 35 minute, so,
there will be waiting time of 2 minutes for the part. After considering the waiting time, the optimal completion time
is 179 minutes.
7166 Bhosale et al., / Materials Today: Proceedings 5 (2018) 7160–7167

Table no.7. Optimum value of the material flow Ramalingam et al. 3]


Product1 Product2 Product3 Total
21 41 32 12 13 22 33 42 32 12 41 21
Processing time 8 24 16 7 6 11 8 22 19 14 16 8
Distance 0 12 12 6 0 12 6 12 0 6 6 12
travel time 0 3 3 1.5 0 3 1.5 3 0 0 1.5 1.5 3 21
waiting time 0 0 0 0 0 0 0 0 0 0 0 0
Start 0 11 38 55.5 0 9 21.5 32.5 0 20.5 36 55
Start+waiting 0 11 38 55.5 0 9 21.5 32.5 0 20.5 36 55
Finish 8 35 54 62.5 6 20 29.5 54.5 19 34.5 52 63 180

Table no.8. Optimum value of the material flow in the manufacturing system with RCGA
Product1 Product2 Product3 Total
21 41 32 12 13 22 33 42 33 13 41 21
Distance 0 12 12 6 0 12 6 12 0 6 18 12 96
travel time 0 3 3 1.5 0 3 1.5 3 0 1.5 4.5 3 24
Start 0 11 38 55.5 0 9 21.5 32.5 0 16.5 33 54
waiting time 0 0 0 0 0 0 0 0 0 0 2 0 2
Start+waiting 0 11 38 55.5 0 9 21.5 32.5 0 16.5 35 54
Processing 8 24 16 7 6 11 8 22 15 12 16 8 153
time
Finish 8 35 54 62.5 6 20 29.5 54.5 15 28.5 51 62 179

Shop 11 Shop 21 Shop 31 Shop 41

Shop 12

Shop 32

Shop 42 Shop 22 Shop 33 Shop 13

Figure 1. Plant Layout Fazlollahtabar et al[3] Ramalingam et al. [10]

Table No.9. Comparison of results

Methods Used Author Sequence of Operations Optimum Finish Time


Traditional Method Fazlollahtabar et al. [3] 21-41-31-11-13-22-33-42-32-12-41-21 186
Genetic algorithm (GA) Ramalingam et al. [10] 21-41-32-12-13-22-33-42-32-12-41-21 180
Real coded genetic Proposed 21-41-32-12-13-22-33-42-33-13-41-21 179
algorithm (RCGA)

A comparison between the proposed method and other two methods is shown in table 9. From the result is observed
that Real Coded Genetic Algorithm is best suitable for this type of problems and giving better results. The optimal
material flow path obtained by the proposed method is as shown in figure no. 1 and 2.
Bhosale et al., / Materials Today: Proceedings 5 (2018) 7160–7167 7167

Shop 11
Shop 12 Shop 13 Product #2

Shop 31
1
Shop 21 3

Shop 32 Product #3
Product #1 3 2
Shop 22
Shop 33
2
2
1

3
Shop 41

Shop 42

Figure 2. Material flow path of all three products


5. Conclusion:
In this paper, a flexible job shop manufacturing system is considered to minimise material flow i.e.to find
out the minimum total completion time of the three products. The mathematical model of Fazlollahtabar et al.[3], is
modified to accommodate waiting time due to unavailability of machines to process new job. Fazlollahtabar et al.
[3] has used traditional method to optimise the material flow of the system. The optimal completion time was 186
units. Ramalingam et al. [10] has proposed Genetic Algorithm (GA) to optimise the same problem. The optimal
completion time was 180 units. However, both the authors have not considered the waiting time due to
unavailability of the machine to process new job .In the present paper, to optimise material flow of the same system,
Real Coded Genetic Algorithm (RCGA) is proposed. As shown in table no.7, optimum material flow by the
proposed method is 179 units which is less than provided by Fazlollahtabar et al.[3], and Ramalingam et al. [10].

References
[1] Aiello, G., Enea, M. & Galante, G., 2002.. International Journal of Production Research, 40(15), pp.4007–4017.
[2] Burnwal, S. & Deb, S., 2013, International Journal of Advanced Manufacturing Technology, 64(5-8), pp.951–959.
[3] Fazlollahtabar, H., Rezaie, B. & Kalantari, H., 2010, International Journal of Advanced Manufacturing Technology, 51(9-12), pp.1149–1158.
[4] Herrmann, J.J.W.J.J.W. et al., 1994, Journal of Manufacturing Systems, 14(4), pp.277–289.
[5] Herrmann, J.W. et al., 1999, IIE Transactions (Institute of Industrial Engineers), 31(7), pp.679–693.
[6] Ioannou, G., 2007, Computers and Industrial Engineering, 52(4), pp.459–485.
[7] Kaspi, M., Kesselman, U. & Tanchoco, J.M. a., 2002, International Journal of Production Research, 40(2), pp.389–401.
[8] Kulak, O., 2005. Expert Systems with Applications, 29(2), pp.310–319.
[9] Malmborg, C.J., 1990, the International Journal of Production Research, 28(10), pp.1741–1758.
[10] Ramalingam, P.S., Chandrasekar, K. & Raj, M.V., 2015, International Journal of Computer Aided Engineering and Technology, 7(1), p.109.
[11] Schaller, J.E., Gupta, J.N.D. & Vakharia, A.J., 2000, European Journal of Operational Research, 125(2), pp.324–339.
[12] Sinriech, D. & Samakh, E., 1999, Journal of Manufacturing Systems, 18(2), pp.81–99.
[13] Sujono, S. & Lashkari, R.S., 2007, International Journal of Production Economics, 105(1), pp.116–133.
[14] Welgama, P.S. & Gibson, P.R., 1995, Computers and Industrial Engineering, 28(2), pp.205–217.

You might also like