Available online at www.sciencedirect.

com
Available online at www.sciencedirect.com

ScienceDirect
ScienceDirect
Procedia
Available Manufacturing00
online (2018) 65–72
atatwww.sciencedirect.com
Available online
Procedia www.sciencedirect.com
Manufacturing00 (2018) 65–72
www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
ScienceDirect
ScienceDirect 
Procedia Manufacturing 22 (2018) 57–64
Procedia Manufacturing 00 (2017) 000–000
www.elsevier.com/locate/procedia
11th International Conference Interdisciplinarity in Engineering, INTER-ENG 2017, 5-6 October
11th International Conference Interdisciplinarity in Engineering,
2017, Tirgu-Mures, Romania INTER-ENG 2017, 5-6 October
2017, Tirgu-Mures, Romania
Production
Production scheduling optimization in foundry using hybrid Particle
Manufacturingscheduling optimization
Engineering Society International in foundry
Conference using
2017, hybrid
MESIC Particle
2017, 28-30 June
Swarm Optimization
Swarm Optimization algorithm
2017, Vigo algorithm
(Pontevedra), Spain
a, a b
Laxmi A.
Costing models forBewoor
capacity a,*, V.optimization
Chandra Prakash a, Sagar U. Sapkal
Laxmi A. Bewoor *, V. Chandra Prakashin Industry
, Sagar 4.0:b Trade-off
U. Sapkal
P P0F P P P

P P0F P P P

between usedScience
Computer
ComputerEng.
Mechanical
capacity
Science
& Eng. Dept.,and
& Eng.
Dept., W. C.Dept.,
b
o. E., K.
a
operational
K. L. University,
P

a
P

L. University,
ShivajiP
Guntur-500002,efficiency
Guntur-500002,
University,
P
India
India
Sangali-416415,India
b
P P

Mechanical Eng. Dept., W. C. o. E., Shivaji University, Sangali-416415,India

A. Santanaa, P. Afonsoa,*, A. Zaninb, R. Wernkeb
P P

Abstract
a
University of Minho, 4800-058 Guimarães, Portugal
Abstract b
Unochapecó, 89809-000 Chapecó, SC, Brazil
Scheduling in foundry consist a type of production where, the hot work-in-processes cannot wait between two successive
Schedulingand
operations in foundry consist aas type
can be modeled a flow of shop
production where,
scheduling the hot
problem work-in-processes
with no-wait constraint. cannot
Withwait between two
the objective successive
to reduce total
operations
flow and appropriate
time, the can be modeled as a of
sequence flow
jobsshop
for scheduling
scheduling problem withhence
is essential, no-waitthe constraint.
problem can With the objective
be observed to reduce
as typical total
NP-hard
Abstract
flow time, the optimization
appropriate sequence
combinatorial problem.ofThisjobs paper,
for scheduling
proposesis hybridization
essential, henceofthe problem
Particle can beOptimization
Swarm observed as typical NP-hard
with simulated
combinatorial
annealing optimization
for planning problem. This
and scheduling issuespaper,
whichproposes
are very hybridization
complex because of Particle Swarm
of the ever Optimization
changing needs of with simulated
customers and
Under
annealing theforconcept
planning ofand"Industry
scheduling 4.0", production
issues which are processes
very complex willbecause
be pushedof the to
ever bechanging
increasingly
needs interconnected,
of customers and
existing constraints in foundry. This Proposed Hybrid Particle Swarm Optimization algorithm represents solution by random key
information
representation
based
existing constraints
rule for
onfoundry.
in a real time
convertingThisthebasis
Proposedand,Hybrid
continuous
necessarily,
Particle
position
much more
Swarm
information
efficient.
Optimization In thistorepresents
algorithm
values of particles
context, capacity
a discretesolution byoptimization
random The
job permutation. key
goes beyond
representation
proposed hybridthe
ruletraditional
particle swarm aim
for converting of
thecapacity
continuous
optimization maximization,
position
algorithm contributing
information
initializes populationvaluesalsoof for organization’s
particles
efficiently with to a discrete profitability
Nawaz-Enscore-Ham and value.
job permutation.
heuristic The
and
Indeed,
proposed
uses lean
hybridmanagement
evolutionary particle swarmand
search guided by thecontinuous
optimization
mechanism improvement
algorithm
of PSOinitializes
and localapproaches
population suggestwith
search byefficiently
mechanism capacity optimization
of Nawaz-Enscore-Ham
simulated Annealing by instead
heuristic of
and
balancing
uses evolutionary
maximization. search
The studyguided
of by the
capacity mechanism
optimizationof PSOand and local
costing search
models by mechanism
is an of
important simulated
research
both global exploration and local exploitation. The proposed hybrid particle swarm optimization algorithm try to bridge the gap Annealing
topic thatby balancing
deserves
both global
contributions
between exploration
theory from bothandthe
and practice local exploitation.
bypractical
considering The proposed
andfoundry
theoretical hybrid
perspectives.
environment, particle
which willswarm
This paper
help optimization
presents
planner andalgorithm
to decide discussestry atomathematical
the sequence bridge the gap
of production
between
model
of theory
for
jobs based and practice
capacity
against management
clients' by considering
orders based
and foundry
on
to develop environment,
different costing
efficient which
models
scheduling will
(ABC helpand
procedures planner
for to decide
TDABC).
minimizing A the
total sequence
generic
flow model
time of production
with has been
relatively
of jobs based against clients' orders and to develop efficient
low computational
developed and it was efforts.
usedExtensive
to analyze computational
idle capacity and toscheduling
experiments design procedures
are carried out based
strategies for on
towards minimizing
various total flow
casting’s
the maximization time with
(job’s) relatively
characteristics
of organization’s
low computational
viz. casting efforts.size
Extensive computational experiments
size ofare carried out based as on 10,12,20,50
various casting’s (job’s) characteristics
value. The type, mould
trade-off capacity andmaximization
type of alloy,vs where
operational job (n) considered
efficiency is highlighted and it and 100.
is shown With
that respect
capacity to
viz. casting type,
performance mould
measure, size and
Average type of
Relative alloy,Deviation
Percent where size of job
which is (n) considered
popular in the as 10,12,20,50
scheduling and the
literature, 100.proposed
With respect
method to
optimization
performance might
measure, hide operational
AverageAnnealing inefficiency.
Relative Percent Deviation which is popular in the scheduling literature, the proposed method
performs
© 2017 The better than
Authors. Simulated
Published Annealing
by Elsevierandand
B.V.Particle Swarm Optimization.
performs
© 2018 The better than Simulated
Authors. Published by Elsevier B.V.Particle Swarm Optimization.
Peer-review
© 2018 The under
Authors. responsibility
Published of
by the scientific
Elsevier B.V. committee of the Manufacturing Engineering Society International Conference
Peer-review
© 2018
2017. The under responsibility
Authors. Published ofthe
by scientific
Elsevier B.V. committee of the 11th International Conference Interdisciplinarity in
Peer-review under responsibility ofthe scientific committee of the 11th International Conference Interdisciplinarity in
Engineering.
Peer-review under responsibility of the scientific committee of the 11th International Conference Interdisciplinarity in Engineering.
Engineering.
Keywords: Cost Models; ABC; TDABC; Capacity Management; Idle Capacity; Operational Efficiency

1. Introduction
* Corresponding author. Tel.:+91-976-653-1977; fax:+91-202-693-2500.
* E-mail
Corresponding author. Tel.:+91-976-653-1977; fax:+91-202-693-2500.
address:[email protected]
The cost
E-mail of idle capacity is a fundamental information for companies and their management of extreme importance
address:[email protected]
in modern production
2351-9789© systems.
2018 The Authors. In general,
Published it is
by Elsevier defined as unused capacity or production potential and can be measured
B.V.
in several under
2351-9789©
Peer-review ways:Thetons of production,
2018 responsibility
Authors. Published available
by Elsevier
ofthe scientific hours
B.V.of
committee of International
the 11th manufacturing, etc. Interdisciplinarity
Conference The management of the idle capacity
in Engineering.
Peer-review underTel.:
* Paulo Afonso. responsibility
+351 253 ofthe scientific
510 761; committee
fax: +351 253 604of 741
the 11th International Conference Interdisciplinarity in Engineering.
E-mail address: [email protected]

2351-9789 © 2017 The Authors. Published by Elsevier B.V.

Peer-review under responsibility of the scientific committee of the Manufacturing Engineering Society International Conference 2017.
2351-9789 © 2018 The Authors. Published by Elsevier B.V.
Peer-review under responsibility of the scientific committee of the 11th International Conference Interdisciplinarity in Engineering.
10.1016/j.promfg.2018.03.010
58 Laxmi A. Bewoor et al. / Procedia Manufacturing 22 (2018) 57–64
66 Laxmi A. Bewoor et al. / Procedia Manufacturing00 (2018) 65–72

Keywords: No wait scheduling; total flow time; NP-hard; Combinatorial optimization; particle swarm optimization.

1. Introduction

Planning and scheduling of jobs is prime factor for deciding the competitive market position for any type of
production environment. In this paper, we present a case study taken from foundry for finding efficient production
plan for scheduling jobs on specified set of machines to minimize total flow time. The production process of
foundry enterprise is assorted and complex with various physical and chemical changes happening at the same time
[1]. So, in order to ensure complete production target a reasonable production planning along with stable operation
of production is the major element for foundry enterprise. Foundry environment of any metal industry has a
multistage process in which melted metal is converted into products sequentially by the processes of converting
furnace, heating furnace, and rolling mill [2]. Distinctly, this represents a typical model of the flow shop where
every job is processed by every machine in fixed order. Additionally foundry is a flow shop with no wait constraint
which has a type of production where, the hot work-in-processes cannot wait between two successive operations..
Furthermore, as the size of the work-in-process of foundry is large, the storage capacity of the buffer between two
successive machines is limited so in-process-inventory should be reduced. Thus, in order to reduce the in-process
inventory, total flow time of jobs should be reduced. This is also generally treated as no wait process demand with
resource constraint [3]. To respond to this call, in this paper, we are aiming at minimizing the total flow time (TFT)
as an optimality criterion, as it determines total processing time of individual job on every machine. The research of
no-wait flow shop with TFT objective is reasonably effective for any metal manufacturing industry.
The remainder of this paper is organized as follows. Section 2 describes production process of foundry. Section
3 formally defines and formulates NWFSSP. Section 4 describes metaheuristics PSO and SA along with detailed
procedure for implementing the proposed metaheuristics. Section 5 evaluates empirically the proposed algorithm for
foundry environment, and then compares the performance of the proposed metaheuristics with that of the best-so-far
algorithms. Finally, concluding remarks are given in section 6.

2. Production Process in Foundry

The considered factory produces various casting products with variety of material combinations. Cast products
produced based on the order of clients are of several sizes and with different material (alloy) combinations. The type
of alloy decides Mechanical property like Tensile strength, Compressive Strength, Shear Modulus of Rupture,
Tensile Modulus of Elasticity, Torsion Modulus of Elasticity, Endurance Limit, hardness, price etc. Castings are
produced using moulds of various sizes, however, neither the number nor the type of moulds are restrictive. Each
mould is filled with a single type of molten metal (ferrous/non ferrous). Client orders determine a set of production
jobs that have to be processed in this system. Each job corresponds to a single alloy and mould type, which has to
pass through consecutive stages with typical sequence. A foundry is a flow line production system in which the
sequence of operations are fixed and the workflow is in a single direction. First step is to prepare moulds/cores using
patterns as per the planned order. In second stage, molten metal from furnace is poured into ladder and then to
mould, which is allowed to cool. Later, castings are knocked out from moulds, which are shot blasted to process
further for finishing. Then, the finished castings are grouped, stored and undergo for heat treatment. Heat treated
castings are fettled and undergo finishing process, and then they are inspected and dispatched to clients. A typical
technological order of operations in a foundry is given in figure 1.
In the said process in the foundry, each stage is considered as a single machine. All castings (jobs) are passing
through all nine stages in the same technological order [3] with defined job processing times for each stage. Because
of technology requirements, a casting cannot wait between melting and pouring. In order to have appropriate
utilization of furnace and work in process inventory (WIP) should be minimum. The time to prepare molten metal
and the time for heat treatment strongly depends on a type of alloy used to produce a typical casting. The processing
times assumed in the second and third stages cannot be changed because to maintain temperature of either heating or
cooling as the prepared casting cannot be either too soft or too hard to be cut in the fourth stage of the technological
process [4].
 Laxmi A. Bewoor et al. / Procedia Manufacturing 22 (2018) 57–64 59
Laxmi A. Bewoor et al. / Procedia Manufacturing 00 (2018) 65–72 67

Client orders and raw material

1 Pouring molten metal from Furnace to Ladder

Pouring molten metal from Ladder to Moulds kept for cooling

2
No-wait

3 Hardening of molten metal in the Moulds

4 Knocking castings from Moulds

5 Shot blasting of castings

No-wait for Shot blasted castings are finished

resource 6
constraint
Heat treatment for finished castings
7

Fettling and finishing process of castings

8

9 Inspection of finished castings for Quality Control

10 Dispatching of inspected castings to clients

Fig. 1. Technological order of operations in a foundry.

The basic problem is to determine such a production schedule that allows minimizing the maximal job
completion time, one of the most frequently considered optimization criteria.

3. Problem Definition

Customers randomly submit orders with specific item type, quantity, and alloy which lead towards high diversity
of products in any foundry. So, it is difficult to schedule the task with exact methods, with reasonable computational
time prior to the production start. So, in order to solve such problem metaheuristic algorithms are proposed [5].
Here, we begin with the mathematical model built using classic notions of no wait flow shop scheduling (NWFSS).
The production task is given by set of jobs {J 1 ,J 2 ,….J n } which represent alloy requirement to be poured into mould
for producing casting of specified size. The production system consists of sequence of machines {M 1 ,M 2 ,…M m }
with additional constraint of no wait as job cannot wait in between in any of the stage. Let σ = {σ1, σ2,. . . σ n }
represent the sequence of ‘n’ jobs to be processed on ‘m’ machines, and δ(i,s) represent the minimum delay on the
first machine between the start of job i and the start of job s. Also, let p(σ i , σ j ) represent the processing time on
machine j of the job at the ith position of a given sequence, and let δ(σ i-1 ,σ i ) denote the minimum delay on the first
machine between the start of two consecutive jobs found in the (i-1)th and ith position of the sequence. Let C(σ i ) be
the completion time of the job in the ith position of a given sequence.
For i = 1,2,. . . ,n and j = 1,2,. . . ,m

C(σ1) = ∑𝑚𝑚
𝑗𝑗=1 𝑝𝑝(𝜎𝜎1 , 𝑗𝑗) (1)

C(σ2) = δ(𝜎𝜎1 , 𝜎𝜎2 ) + ∑𝑚𝑚

𝑗𝑗=1 𝑝𝑝(𝜎𝜎2 , 𝑗𝑗) (2)

C(σi) = ∑𝑖𝑖𝑘𝑘=2 δ(𝜎𝜎𝑘𝑘−1 , 𝜎𝜎𝑘𝑘 ) + ∑𝑚𝑚

𝑗𝑗=1 𝑝𝑝(𝜎𝜎𝑖𝑖 , 𝑗𝑗) (3)

The formula for total flow time (TFT) is given as:

C max (𝜎𝜎)=∑𝑛𝑛𝑖𝑖=1 𝐶𝐶(𝜎𝜎𝑖𝑖 ) (4)

60 Laxmi A. Bewoor et al. / Procedia Manufacturing 22 (2018) 57–64
68 Laxmi A. Bewoor et al. / Procedia Manufacturing00 (2018) 65–72

= ∑𝑛𝑛𝑖𝑖=2{ ∑𝑖𝑖𝑘𝑘=2 δ(𝜎𝜎𝑘𝑘−1 , σ𝑘𝑘) + ∑𝑚𝑚 𝑚𝑚

𝑗𝑗=1 𝑝𝑝(𝜎𝜎𝑖𝑖 , 𝑗𝑗) } + ∑𝑗𝑗=1 𝑝𝑝(𝜎𝜎1 , 𝑗𝑗) (5)

= ∑𝑛𝑛𝑖𝑖=2(𝑛𝑛 + 1 − 𝑖𝑖)δ(𝜎𝜎𝑖𝑖−1 , 𝜎𝜎𝑖𝑖 ) + ∑𝑛𝑛𝑖𝑖=1 ∑𝑚𝑚

𝑗𝑗=1 𝑝𝑝(𝑖𝑖, 𝑗𝑗)) (6)

The delay matrix of n x n provides all the δ(i,k) values between the start of any two consecutive jobs i and k (i ≠
k) in a given sequence of n jobs to determine the objective function value. The delay matrix of the δ(i,k) values
obtained from the (7)

𝑑𝑑(𝑖𝑖, 𝑘𝑘) = 𝑝𝑝(𝑖𝑖, 1) + 𝑚𝑚𝑚𝑚𝑚𝑚(∑𝑟𝑟ℎ=2 𝑝𝑝(𝑖𝑖, ℎ) − ∑𝑟𝑟−1

ℎ=1 𝑝𝑝(𝑘𝑘, ℎ), 0) 2 ≤ 𝑟𝑟 ≤𝑚𝑚 (7)

For the given matrix of size n jobs) x m (machines) with processing time 𝐩𝐩[i,j] generates (n!) number of feasible
sequence of solutions(σ) from which optimal sequence (σ*)is to be chosen and can be stated as per (8)

F (σ*) ≤ F (σ); ∀σ ∈ σ (8)

The problem is to determine a sequence of n jobs which gives minimum total flow time.

4. Proposed Hybrid-PSO (PHPSO)

Though, the field of metaheuristics for the application of combinatorial optimization problems of scientific and
industrial world is rapidly growing [7]; but the use of metaheuristics for solving combinatorial optimization
problems had begun late. The comparative analysis of various metaheuristics by Bewoor et.al [8], [9] advocated the
effectiveness of PSO. Recent past witnessed notable shift towards the hybridization of metaheuristics for
optimization [10]. The work reported in this paper also exhibits the use of hybridization of metaheuristic techniques
for solving real life scheduling problem observed in foundry environment.

4.1 Particle Swarm Optimization (PSO) algorithm

PSO is an optimization algorithm simulating its behaviour from the biological example of flock of birds
searching for food in a defined area [6]. Birds follow nearest to food strategy for searching food which is exactly
adapted by PSO for searching the best solution vector in the given search space. A single solution is called particle
which has fitness value that is evaluated by function to be minimized and each particle has velocity that directs the
flying of the particles. The algorithm is initialized with particles at random positions and then it explores the search
space to find better solution. At every iteration, each particle adjusts its velocity to follow two best solutions. The
first is cognitive part where particle follow its own best solution found so far called pbest and other is current best
solution of swarm called gbest. In NWFSS context, each particle learns from the best particle in its neighbourhood
thus, it adapts to a changing environment more easily. The new velocity denoted by Vnew and new position denoted
by X new given in (9) and (10)

Vnew = w*Vcurr + c1*r1*(pbest-Xcurr) + c2*r2*(gbest-Xcurr) (9)

Xnew = Xcurr+Vnew (10)

Where, w is inertia weight which provides balance between local and global search capabilities. The acceleration
constants c1 and c2 in (9) are cognitive parameter which develops the bird’s own confidence (cognitive behavior)
and confidence in swarm (social behavior) respectively. Low values allow particles to roam far from target regions
before being tugged back, while high values result in abrupt movement from target regions. In this paper, we set the
acceleration constants c1 and c2 each equal to1.65 and 1.75 respectively. r1 and r2 are even distributed random
number between[0,1]. In this paper, r1 and r2 are set to the value 0.5 each. X new and V new is new position and
velocity of particle respectively, X curr and V curr is current position and velocity of particle respectively. In the
 Laxmi A. Bewoor et al. / Procedia Manufacturing 22 (2018) 57–64 61
Laxmi A. Bewoor et al. / Procedia Manufacturing 00 (2018) 65–72 69

standard PSO, new velocity of the particle is found by (9) considering its previous velocity and the distance of its
current position from both its own best historical position and its neighbors’ best position. Generally, the value of
each component in velocity is set to the range (V max ,-V max ) to control the excessive roaming of particles outside the
search space. With this new velocity, the particle moves towards a new position according to (10). This process is
repeated until a user-defined stopping criterion is reached.

4.2 Solution Representation

Solution representation is one of the most important issues while designing PSO algorithm. Usually, a job-
permutation-based encoding scheme has been widely used in many papers for the no-wait FSSP. However, because
of the continuous characters of the position of particles in PSO, a standard encoding scheme of PSO cannot be
directly adopted for the no-wait FSSP. In order to construct a direct relationship between the problem domain and
the PSO particles for the NWFSS, ‘n’ numbers of dimensions are presented for ‘n’ number of jobs. Each dimension
represents a typical job and the particle X i = {x 1 , x 2 , x 3 ,...x n ) corresponds to the continuous position values for n
number of jobs in the NWFSS. The particle itself does not present a permutation. So it is necessary to find suitable
mapping between the job sequence and the position of particles in PSO so, in this paper the ranked-order-value
(ROV) rule based on random key value [11] is used to determine the permutation implied by the position values x ij
of particle Xi. The ROV rule converts the continuous position values to a discrete job permutation which is the key
to enable the continuous PSO algorithms to be applied to sequencing problems and thus, the performance of a
particle can be evaluated. In particular, the position information of the particle represented as Xi={x 1 ,x 2 ,…..x n }
itself does not represent a sequence, whereas the rank of each position value of a particle represents a job index,
which can be used to construct a permutation of jobs. In our ROV rule, the particle with smallest position value is
firstly handled and assigned a smallest rank value 1. If there are two or more particles with same position values in
PSO-based searching process, then the position value with the smaller or smallest dimension number is assigned a
rank value 1. And the rank values of the rest of the position values are incremented by 1 as per the increase of
dimension number. Then, the second smallest position value will be handled in the same way. Note that, if there is k
same smallest position values, the rank value of the second smallest position value with the smallest dimension
should be k+1. In the same way, all the position values will be dealt with so as to convert the position information of
a particle to a job permutation π ij =[j 1 ,j 2 ,j 3 …j n ].

4.3 Population Initialization

The initial swarm generation is random often in the standard PSO. In order to provide efficient solution it is
necessary to generate an initial population with a certain quality and diversity. In this paper we propose an efficient
population initialization procedure based on the NEH heuristic [12]. NEH heuristic has two phases. In phase I, jobs
are ordered in ascending sums of their total flow times. In phase II, a job sequence is established by evaluating the
partial schedules based on the initial order of the first phase. Suppose a current sequence is already determined for
the first k jobs, k+1 partial sequences are constructed by inserting job k+1 in k+1 possible slots of the current
sequence. Among this k+1 sequence, the one generating the minimum total flowtime is kept as the current sequence
for the next iteration. Then job k + 2 from phase I is considered and so on until all jobs have been sequenced.

4.4 Simulated Annealing (SA) based local search

Annealing is a process in metallurgy where metals are slowly cooled to make them reach to a state of low energy
where they are very strong [13]. When applied to engineering design, an analogy is made between energy and the
objective function. The design is started at a high “temperature”, where it has a high objective. Random
perturbations are then made to the design. If the objective is lower, the new design is made the current design; if it is
higher, it may still be accepted according the probability given by the Boltzmann factor given in (11).

−∆𝐸𝐸�
𝑃𝑃 = 𝑒𝑒 𝛼𝛼𝑇𝑇 (11)
62 Laxmi A. Bewoor et al. / Procedia Manufacturing 22 (2018) 57–64
70 Laxmi A. Bewoor et al. / Procedia Manufacturing00 (2018) 65–72

where, α is the Boltzmann constant and T is the current temperature ∆ E is change in energy.The Boltzmann
probability is compared to a random number drawn from a uniform distribution between 0 and 1; if the random
number is smaller than the Boltzmann probability, the configuration is accepted. This allows the algorithm to escape
local minima. A proper initial temperature should be high enough so that all states of the system have an equal
probability of being visited. In this paper, we set an initial temperature as 3.0 and final temperature as 0.9 and α as
0.99.

4.5 Computational Procedure for Hybrid PSO

Based on the solution representation by ROV rule, population initialization by NEH-based local search and
neighbourhood searching through SA-based local search, the complete computational procedure of PHPSO
proposed Hybrid PSO framework for the NWFSS can be summarized as follows:
Algorithm: PHPSO algorithm applied in Foundry
Step 1: For foundry environment, alloy production data as per customer’s need is considered as’ job’ (n) for
production
Step 2: Technological order followed for Production process of foundry requires ‘m’ machines. Here, m=8.
Step 3: Processing time required by each job on each machine is inputted for finding sequence of jobs for
minimizing total completion time (makespan).
Step 4: Initial Population was generated with NEH algorithm. Ranked order value (ROV) algorithm was used for
converting continuous particle position to discrete job permutation.
Step 5: For i :=0 to n -1 do
Particle’s best (pbest) and swarm’s best (gbest) was calculated for makespan objective. Evolutionary search
was performed by updating position and velocity of particle.
Step 6: For the current value of gbest find the neighbourhood with local search initiated with SA. Initialize initial
temperature and final temperature.
Step 7: while initial_temperature <=Final_temperature
Find the value of makespan for given job sequence and reduce temperature with Boltzman’s factor
Step 8: Output gbest as optimal makespan result and corresponding job sequence as plan for scheduling jobs.
End Procedure

5. Computation Results

The proposed method considers a real time problem observed in foundry. Following are the important
considerations for the stated problem [14][15][16].

• For scheduling, all jobs arrive at the beginning of every fixed interval of time under consideration.
• Scheduling planning period is one week. At the beginning of the planning period, the number of castings
corresponding to first day of the planning period is in WIP. Furthermore, the entire jobs corresponding to the
planning period have to be scheduled first before considering the jobs associated with the next planning period
and no re-scheduling is allowed.
• Once processing of a job is initiated, it cannot be interrupted and other jobs cannot be introduced until the
current processing is completed.
• The set-up times of the jobs are included in the processing time.
• Machine breakdowns are ignored and manpower of uniform skill is continuously available.

The industry produces castings of sizes denoted casting type a, b, c, d, e, f, g using moulds of various sizes
ranging from 10 to 1000 Kg with variety of material combination specified with Alloy1 to Alloy10 as specified in
Table 1 which imply demands by the customers. Next, all orders are transformed into a set of jobs, where each job
corresponds to a specific mould filled by specified type of alloy to provide castings of specified weight/dimensions
(see column job characteristics in Table 2). According to the technological process, there are defined job processing
times, Table 2. The total flow time is 581 with optimal sequence as 4,5,1,2,7,3,6.
 Laxmi A. Bewoor et al. / Procedia Manufacturing 22 (2018) 57–64 63
Laxmi A. Bewoor et al. / Procedia Manufacturing 00 (2018) 65–72 71

Table 1. Data for alloy production system as per demands

Alloy type Casting Type (By size) Number of blocksper mould
Alloy1 a 100
Alloy 2 a 50
Alloy 3 a 25
Alloy 4 b 80
Alloy 5 b 40
Alloy 6 c 20
Alloy7 d 20
Alloy 8 e 20
Alloy 9 f 15
Alloy10 g 10

Table 2. Data for instance.

Processing Time
Job Job Characteristics
Pj,1 Pj,2 Pj,3 Pj,4 Pj,5 Pj,6 Pj,7 Pj,8
J1 a-100-Alloy1 25.0 76.0 95.0 26.0 8.0 8.0 69.0 83.0
J2 a-50-Alloy2 38.0 46.0 62.0 26.0 72.0 28.0 11.0 72.0
J3 a-25-Alloy3 53.0 32.0 21.0 98.0 48.0 18.0 1.0 89.0
J4 b -80-Alloy 4 86.0 87.0 95.0 12.0 27.0 86.0 79.0 56.0
J5 b – 40-Alloy 5 65.0 21.0 42.0 74.0 55.0 11.0 56.0 76.0
J6 c -20-Alloy 6 4.0 67.0 52.0 11.0 23.0 5.0 80.0 37.0
J7 d -20-Alloy7 95.0 3.0 72.0 67.0 96.0 4.0 16.0 28.0
J8 e- 20-Alloy 8 77.0 33.0 65.0 16.0 45.0 74.0 84.0 33.0
J9 f -15-Alloy 9 76.0 71.0 83.0 79.0 25.0 39.0 58.0 91.0
J10 g-10- Alloy10 61.0 53.0 49.0 55.0 90.0 46.0 94.0 93.0

Table 3.Average value of TFT obtained from 30 instances for the given size.
ARPD
nxm
PSO SA PHPSO Exact
10x8 0.66 0.48 0 0
12x8 0.75 0.60 0 0
20x8 0.70 0.61 0 *--
50x8 0.72 0.70 0 *--
100x8 0.78 0.69 0 *--
[‘*--’ indicates not able to find out exact values for these problem instances.]

The proposed PSO based SA is coded in java and run on Intel Core i5, 8 GB RAM, 2.20 GHz PC. To test the
performance of the proposed algorithm, a computational simulation is carried out with random instances. Each test
problem comprises with n jobs and m machines (here m=8).The processing time matrix (P ij ) was generated with
seed index 1 to 30 and uniform distribution on [1,100]. For each n x m : 10x8; 20x8; 50x8;100x8; 200x8, a sample
of 30 instances were provided. Initial solution for the algorithm PHPSO was found by NEH method. Since there are
no solution algorithms for our scheduling problem in the literature, we made a comparison of PHPSO with exact
methods for n=10, 12 along with PSO and SA. To compare the proposed PHPSO with the existing heuristics, we
carried out the experimentation by running each instance independently 10 times, in which in each replication, we
used ‘average relative percentage deviation’ (ARPD) as a performance measure, which is popular in the scheduling
literature [17][18][19]. ARPD is given by,

100
𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 = ∑𝑘𝑘𝑖𝑖=1(𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝑐𝑐𝑖𝑖 − 𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐻𝐻𝑖𝑖 )/𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐻𝐻𝑖𝑖 (12)
𝑘𝑘
64 Laxmi A. Bewoor et al. / Procedia Manufacturing 22 (2018) 57–64
72 Laxmi A. Bewoor et al. / Procedia Manufacturing00 (2018) 65–72

where, Heuristic i is the total flowtime obtained by any of four algorithms and the BestH i was the lowest total
flowtime obtained in any treatment for that specific instance. Table 3 displays comparative evaluation of the
proposed metaheuristic PHPSO, SA and PSO based on ARPD.

6. Conclusion

In this paper, we have proposed hybridization of PSO with simulated annealing for the flow shop scheduling with
no-wait constraint for real time example of foundry. The developed PHPSO algorithm apply evolutionary search
guided by the mechanism of PSO and local search guided by the NEH based initial population and mechanism of
SA. The proposed hybrid algorithm ensures balance between both global exploration and local exploitation. In order
to demonstrate the superiority of the PHPSO the simulations were carried out and the results are compared with
another metaheuristic viz.SA and PSO. The effectiveness of the proposed method was measured by using ARPD,
which is widely used performance measure. The comparative analysis proves that PHPSO outperforms the existing
metaheuristics. The proposed algorithm can help production planner for efficient scheduling of jobs in foundry to
meet the requirements of scheduling procedures and satisfaction of fulfilling the clients’ order by minimizing TFT
with relatively low computational efforts.

References
[1] F. Ballestín, F. Mallor P.M. Mateo, Production scheduling in a market-driven foundry: a mathematical programming approach versus a
project scheduling metaheuristic algorithm, Optimization and Engineering. 13(4) (2012) 663–687.
[2] H. Yuan, Y. Jing, J. Huang, T. Ren, Optimal research and numerical simulation for scheduling no-wait flow shop in steel production,
Journal of Applied Mathematics, vol. 2013, 2013.
[3] M. Mathirajan, V. Chandru, A.I. Sivakumar, Heuristic algorithms for scheduling heat-treatment furnaces of steel casting industries, Sadhana
- Academy Proceedings in Engineering Sciences. 32(5) (2007) 479–500.
[4] S.L. Nonas, K.A. Olsen, Optimal and heuristic solutions for a scheduling problem arising in a foundry, Computers and Operations
Research. 32(9) (2005) 2351–2382.
[5] M. Gravel, W.L. Price, C. Gagné, Scheduling jobs in an Alcan aluminium foundry using a genetic algorithm, International Journal of
Production Research. 38(13) (2000) 3031–3041.
[6] R.C. Eberhard, J. Kennedy, A new optimizer using particle swarm theory, in: Proceedings of the Sixth International Symposium on Micro
Machine and Human Science, Nagoya, Japan. pp.39–43,1995.
[7] C. Blum, A. Roli, Metaheuristics in Combinatorial Optimization : Overview and Conceptual Comparison, vol. 35, no. 3, pp. 268–308, 2003.
[8] L.A. Bewoor, V.C. Prakash, S.U. Sapkal, Comparative Analysis of Metaheuristic Approaches for m-Machine No Wait Flow Shop
Scheduling for minimizing Total Flow Time with Stochastic Input, International Journal of Engineering and Technology. 8(6) (2016) 3021–
3026.
[9] L. Bewoor,V. Chandra, S. Sapkal, Comparative Analysis of Metaheuristic approaches for Makespan Minimization for No Wait Flow Shop
Scheduling Problem, Int.J.Elec.7(1) (2017) 31-37.
[10] C. Blum, “Hybrid metaheuristics in combinatorial optimization: A tutorial,” Lecture Notes in Computer Science (including subseries
Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 7505(6) (2012) 1–10.
[11] J.C. Bean, Genetic algorithms and random keys for sequencing and optimization, ORSA J Comput. 6(2) (1994) 154–160.
[12] M. Nawaz, E.E. Enscore, I. Ham, A Heuristic Algorithm for the m- Machine, n-Job Flow-shop Sequencing Problem, OMEGA, The int. J.
of Mgmt Sci. 11(1) (1983) 91–95.
[13] B. Jarboui, S. Ibrahim, P. Siarry, A. Rebai, A combinatorial particle swarm optimisation for solving permutation flowshop problems,
Computers and Industrial Engineering. 54(3) (2008) 526–538.
[14] G. Matičević, N. Majdandžić, T. Lovrić, Production scheduling model in aluminium foundry, Strojniski Vestnik/Journal of Mechanical
Engineering. 54(1) (2008) 37–48.
[15] J.-M. Yang, Y.K. Park, Scheduling of casting in real foundries using linear programming, Proceedings of the Institution of Mechanical
Engineers, Part B: Journal of Engineering Manufacture. 223(10) (2009) 1351–1360.
[16] S.L. Nonås, K.A. Olsen, Optimal and heuristic solutions for a scheduling problem arising in a foundry, Computers and Operations
Research. 32(9) (2005) 2351–2382.
[17] S.U. Sapkal, D. Laha, A heuristic for no-wait flow shop scheduling,International Journal of Advanced Manufacturing Technology. 68(5–8)
(2013) 1327–1338.
[18] Q, Pan, M. F. Tasgetiren, Y. Liang, A discrete particle swarm optimization algorithm for the no-wait flowshop scheduling problem,
Computers & Operations Research. 35(9) (2008) 2807-2839.
[19] Q.-K. Pan, R. Ruiz, Local search methods for the flowshop scheduling problem with flowtime minimization, European Journal of
Operational Research. 222(1) (2012) 31–43.