International Journal of Advances in Applied Sciences (IJAAS)

Vol. 12, No. 3, September 2023, pp. 195~204

ISSN: 2252-8814, DOI: 10.11591/ijaas.v12.i3.pp195-204  195

Low-carbon no-idle permutation flow shop schedulling

problem: giant trevally optimizer vs African vultures
optimization algorithm

Dana Marsetiya Utama, Cantika Febrita

Department of Industrial Engineering, Faculty of Engineering, Muhammadiyah Malang University, Malang, Indonesia

Article Info ABSTRACT

Article history: Greenhouse gas emissions continue to increase due to increased energy
consumption. One of the largest emission-contributing sectors is the
Received Apr 27, 2023 manufacturing industry. Therefore, the manufacturing industry is required to
Revised Jun 26, 2023 minimize carbon emissions. One of the efforts to solve the emission problem
Accepted Jul 15, 2023 is to minimize machine downtime throughout the production procedure,
which stands for no-idle permutation flowshop scheduling (NIPFSP). This
article uses two metaheuristic algorithms, giant trevally optimizer (GTO)
Keywords: and African vultures optimization algorithm (AVOA), to solve the carbon
emission problem. Both algorithms are tested on 3 cases with 30 runs for
African vultures optimization every population and iteration. To compare the outcomes of each algorithm,
algorithm an independent sample t-test was employed. The results show that the GTO
Flow shop algorithm has better results than the AVOA algorithm on small and large
Giant trevally optimizer case data. The findings indicate that both the GTO and AVOA algorithms
Low carbon yield comparable results when applied to medium-sized research datasets,
No-idle suggesting their effectiveness in such scenarios.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Cantika Febrita
Department of Industrial Engineering, Faculty of Engineering, Muhammadiyah Malang University
Tlogomas highway no 246 Malang, East Java, Indonesia
Email: [email protected]

1. INTRODUCTION
The high use of carbon emissions has become a concern in the world, and one of the sectors that
contribute the highest carbon emissions is the manufacturing industry [1]. According to research [2], high
carbon emissions can be caused by high energy consumption from machine operation [3], and the energy-
efficiency problems in the manufacturing sector are closely related to scheduling [4]. There are two practical
techniques to reduce carbon emissions in manufacturing: using energy-efficient equipment [5] and
implementing an optimal low-carbon production strategy [6]. In the carbon-efficient scheduling model
proposed by Ding et al. [7], carbon emissions are only related to energy consumption but ignore carbon
emissions caused by preparation time and idle time on machines [8]. Furthermore, Asmatulu et al. [9]
showed that when idle machines can consume a large quantity of energy, the manufacturing process can
reduce carbon emissions by minimizing idle machines [10]. One exciting solution is to use no-idle machine
scheduling to minimize carbon emissions. This approach is much more effective in reducing carbon
emissions in the manufacturing process [7].
No-idle permutation flow shop problem (NIPFSP) is different from other permutation flow shop
problems (PFSP) because, in NIPFSP, each machine must process jobs without idle time from the start of the
job until the last job is completed [11]. NIPFSP is categorized as a constraint problem [12]. This is because
the NIPFSP involves constraints that must be satisfied to find a feasible solution. For example, in the

Journal homepage: http://ijaas.iaescore.com

196  ISSN: 2252-8814

NIPFSP, there is a constraint that no-idle time is allowed between the processing of jobs on different
machines. This constraint must be satisfied in order to find a feasible solution to the problem. In real life,
NIPFSP has been applied in the production process of glass fiber processing [13]. Then various methods
were introduced by several researchers such as heuristic algorithms [14], to solve NIPFSP. There is also
research on Flow Shop Dependent Sequence Setup scheduling using the ant-lion optimizer algorithm to
minimize carbon emissions [15].
Some research on NIPFSP uses several algorithms. However, NIPFSP research aimed at minimizing
carbon emissions has not been carried out, so NIPFSP research is needed to minimize carbon emissions.
Thus, the metaheuristic algorithm can be applied in solving NIPFSP to minimize carbon emissions. Due to
the absence of such research, this research is exciting to be researched further by presenting new algorithms,
namely giant trevally optimizer (GTO) and African vultures optimization algorithm (AVOA), to solve the
NIPFSP problem to minimize carbon emissions. GTO is a metaheuristic algorithm that mimics giant trevally
fish's behavior when hunting seabirds [16]. Furthermore, AVOA mimics the behavior of African vultures
when hunting prey [17]. GTO is proposed because it is proven to perform problem-solving for global
optimization, capable of solving optimization problems and sustainable global engineering design [16].
However, GTO has never been used to solve the NIPFSP problem. Whereas AVOA has also solved various
problems such as shell and tube heat exchanger optimization [18], membrane fuel cell optimization [19], and
multi-objective flexible job shop scheduling problems [20]. However, AVOA has never been used to
minimize carbon emissions.
From the aforementioned explanation, it can be inferred that the NIPFSP issue in minimizing carbon
emissions needs additional research using the most recent algorithms. This research proposes the GTO and
AVOA algorithms because they are novel and have never been applied to solve NIPFSP problems focusing
on minimizing carbon emissions. Thus, this research aims to implement two of the newly developed
algorithms, GTO and AVOA, to solve the NIPFSP problem by minimizing carbon emissions and comparing
the results of the GTO and AVOA algorithms on the NIPFSP problem to minimize carbon emissions.

2. DESCRIPTION OF NIPFSP
2.1. Assumptions and mathematical formula
The terminology and mathematical model formulation utilized in the NIPFSP problem are described
in this section. Machines can be idle after finishing work in a typical PFSP scenario [21]. However, in
NIPFSP, it is presumed that the machine must remain active and not idle once the tasks are accomplished.
Completion time on the machine: job identification: machine identification: speed rate identification: The job
processing time on the machine completes one job [22]. Researchers use several assumptions that will be
used in this study. The assumptions used for NIPFSP scheduling include: i) the entire set of n jobs must be
processed on the set of m machines in the same process order, ii) all processes arrive and are ready for
processing at time 0, iii) to fulfill the no-idle requirement, the processing start time for the first job on each
machine must be delayed, iv) every machine can only process one job and can only be processed on one
machine at a certain time, v) when the first job starts processing, it cannot be interrupted until the completion
time of the last job, vi) job processing time includes setup time, and vii) no-idle machines are allowed in
between job processing. The notations used in the NIPFSP problem are shown in Table 1.

Table 1. The notation used in the NIPFSP problem

Notation NIPFSP problem
𝑚 Number of machines
𝑛 Number of jobs
𝑗 Index of the machine
𝑖 Index of the job
𝑆𝑗 Start time on machine j
𝐶𝑖,𝑗 Time required for completing the job I on machine j
𝐹𝑗 Completion time of the j machine
𝑃𝑖,𝑗 Processing time of job i
𝐶𝑖,𝑗 Completion time on machine j
𝜑𝑗 Consumption of energy during the machine is idle
𝜏𝑗 Energy consumption during machine i is operating
𝐶𝑚𝑎𝑥 Total time or completion time
𝑇𝐸𝐶 Cumulative energy consumption
𝐸𝐶 Total carbon emission
𝐸𝑓 Factor emission

Int J Adv Appl Sci, Vol. 12, No. 3, September 2023: 195-204
Int J Adv Appl Sci ISSN: 2252-8814  197

The following is a mathematical formulation to minimize carbon emission in NIPFSP problems

presented in the mixed integer programming (MIP) model:
Decision Variable

1, 𝑖𝑓 𝑗𝑜𝑏 𝑖 𝑖𝑠 𝑡ℎ𝑒 𝑝𝑟𝑒𝑑𝑒𝑐𝑒𝑠𝑠𝑜𝑟 𝑜𝑓 𝑗𝑜𝑏 𝑘

𝑋𝑖,𝑗,𝑟 = { }
0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

1, 𝐼𝑓 𝑗𝑜𝑏 𝑖 𝑖𝑠 𝑐𝑜𝑚𝑝𝑙𝑒𝑡𝑒𝑑 𝑎𝑡 𝑠𝑝𝑒𝑒𝑑 𝑟 𝑜𝑛 𝑚𝑎𝑐ℎ𝑖𝑛𝑒 𝑗

𝑌𝑖,𝑗,𝑟 = { }
0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

Objective purpose

Min EC = TEC ∗ Ef (1)

Constraint:

𝐶𝑖,1 ≥ ∑𝑙𝑟=1 𝑃𝑖1∗ 𝑌𝑖𝑙𝑟 ∀𝑖 = {1, . . , n} (2)

𝐶𝑖𝑗− 𝐶𝑖,𝑗−1 ≥ ∑𝑙𝑟=1 𝑃𝑖1∗ 𝑌𝑖𝑙𝑟 ∀𝑗 = {2, . . , m}, 𝑖 = {2, . . , n} (3)

𝐶𝑖𝑗− 𝐶𝑘𝑗 + 𝐷𝑋𝑖𝑘 ≥ ∑𝑙𝑟=1 𝑃𝑖1∗ 𝑌𝑖𝑙𝑟 ∀𝑗 = {1, . . , m}, 𝑖 = {1, . . , n}, 𝑘 = {1, . . , n} (4)

𝐶𝑖𝑗− 𝐶𝑘𝑗 + 𝐷𝑋𝑖𝑘 ≤ 𝐷 − ∑𝑙𝑟=1 𝑃𝑖1∗ 𝑌𝑖𝑙𝑟 ∀𝑗 = {1, . . , m}, 𝑖 = {1, . . , n}, 𝑘 = {1, . . , n} (5)

𝐶𝑚𝑎𝑥 ≥ 𝐶𝑖𝑚 ∀𝑖 = {1, . . , n} (6)

∑𝑙𝑟=1 𝑌𝑖𝑗𝑟 = 1 ∀𝑖 = {1, . . , n}, j = {1, . . , m}c (7)

𝑌𝑖𝑗𝑟 = 𝑌𝑖,𝑗+1,𝑟 ∀𝑖 = {1, . . , n}, j = {1, . . , m}, r = {1, . . , l} (8)

𝑃𝑖1 𝑌𝑖𝑙𝑟
𝜃𝑗 = 𝐶𝑚𝑎𝑥 − ∑𝑛𝑖= 1 ∑𝑙𝑟=1 ∀𝑗 = {1, . . , m} (9)
𝜇𝑟

𝑃𝑖𝑗 𝜏𝑗 𝜆𝑟 𝜑𝑗 𝜃𝑗 𝜏𝑗
𝑇𝐸𝐶 = ∑𝑛𝑖=1 ∑𝑚 𝑙
𝑗=1 ∑𝑟=1 𝑌𝑖𝑗𝑟 + ∑𝑚
𝑗=1 (10)
60𝜇𝑟 60

𝑃𝑖1 𝑌𝑖𝑙𝑟
𝑆𝑗 ≤ 𝐶𝑖𝑗 − ∑𝑙𝑟=1 ∀𝑖 = {1, . . , n}, 𝑗 = {1, . . , m} (11)
𝜇𝑟

𝐹𝑗 ≥ 𝐶𝑖𝑗 ∀𝑖 = {1, . . , n}, 𝑗 = {1, . . , m} (12)

𝑃𝑖1 𝑌𝑖𝑙𝑟
𝐹𝑗 ≥ 𝑆𝑗 + ∑𝑛𝑖= 1 ∑𝑙𝑟=1 ∀𝑖 = {1, . . , n}, 𝑗 = {1, . . , m} (13)
𝜇𝑟

Equation (1) represents the objective function for minimizing carbon emissions. The amount of time
needed to finish each task on the first machine is depicted in Constraint (2). Constraint (3) guarantees that the
subsequent operation will commence only if the preceding one has been finalized. The sequencing of every
job is outlined in Constraints (4), (5). The makespan, or finishing time, is calculated in Constraint (6).
Constraints (7) and (8) ensure that all machines process every job at the same machining.

3. RESEARCH METHOD
3.1. Giant trevally optimizer (GTO)
The first used algorithm is giant trevally optimizer (GTO) [16], The GTO algorithm's foundation is
the behavior displayed by diving giant trevallies when hunting seabirds. As a result, the GTO algorithm's
optimization process comprises three distinct steps: a comprehensive search employing Levy flight, the
selection of an area for hunting, and the pursuit and attack of prey by leaping out of the water. The positional
representation of an individual within the GTO population is defined by (14).

Low-carbon no-idle permutation flow shop schedulling problem: giant trevall … (Dana Marsetiya Utama)
198  ISSN: 2252-8814

𝑋1 𝑥1,1 ⋯ 𝑥1,𝑗 … 𝑥1,𝐷𝑖𝑚

⋮ ⋮ ⋱ ⋮ ⋰ ⋮
𝑋 = 𝑋𝑖 = 𝑥1,1 ⋯ 𝑥1,𝑗 … 𝑥1,𝐷𝑖𝑚 (14)
⋮ ⋮ ⋰ ⋮ ⋱ ⋮
[𝑋𝑁 ] [𝑥𝑁,1 ⋯ 𝑥𝑁,𝑗 … 𝑥𝑁,𝐷𝑖𝑚 ]

Step 1: Extensive search

Giant trevally fish are capable of covering considerable distances. The foraging movement pattern
of giant trevally fish is simulated with (15):

X (𝑡 + 1) = 𝐵𝑒𝑠𝑡𝑝 𝑥 𝑅 + ((𝑚𝑎𝑥𝑖𝑚𝑢𝑚 – 𝑚𝑖𝑛𝑖𝑚𝑢𝑚) 𝑥 𝑅 + 𝑚𝑖𝑛𝑖𝑚𝑢𝑚) 𝑥 𝑙𝑒𝑣𝑦 (𝐷𝑖𝑚) (15)

X (𝑡 + 1) represents the vector of position for the subsequent iteration giant gannets, The selection of
currently available search space by the giant gannets is denoted by 𝐵𝑒𝑠𝑡𝑝 , which is based on the best position
discovered during their previous search. The variable R represents a random number ranging from 0 to 1. The
Levy (Dim) term corresponds to a Levy flight, which is a specific type of non-Gaussian stochastic process.
The step size in the Levy flight is determined by the Levy distribution [23].
One of the primary benefits of employing Levy flights is the ability to circumvent local optima,
leading to enhanced convergence speed [24]. The calculation of Levy (Dim) is performed according to (16).
A step size of 0.01, referred to as "step," is utilized. The Levy flight distribution function index, denoted as β,
is set to 1.5 in this particular investigation, adhering to the range of 0 to 2. Random variables u and v follow a
normal distribution within the interval of 0 to 1. The computation of σ is determined by applying in (17).
𝑢−𝜎
𝑙𝑒𝑣𝑦 (𝐷𝑖𝑚) = 𝑆𝑡𝑒𝑝 𝑥 1 (16)
|𝑣|
𝛽

𝜋𝛽
Г (1+𝛽)𝑥 sin 𝑒( 2 )
σ=( 1+𝛽 𝛽−1 ) (17)
Г( )𝑥𝛽𝑥2( )
2 2

Step 2: Selecting an area

During the area selection step, the giant trevally employs a strategy to identify and choose the
optimal region within the selected search space, considering the abundance of food sources (seabirds)
available for hunting prey. This behavior is mathematically simulated by (18).

X (t + 1) = 𝐵𝑒𝑠𝑡𝑝 𝑥 𝐴 𝑥 𝑅 + 𝑀𝑒𝑎𝑛_𝐼𝑛𝑓𝑜 − 𝑋𝑖 (𝑡)𝑥 𝑅 (18)

Equation (19) calculates the 𝑀𝑒𝑎𝑛_𝐼𝑛𝑓𝑜, which represents the mean value obtained from utilizing all
accessible information from prior points. 𝑋𝑖(𝑡) denotes the i-th position of the giant trevally at the latest
iteration, while X (t + 1) signifies the location vector of the giant trevally for the subsequent iteration. 𝐴, a
parameter controlling the position change, ranges between 0.3 and 0.4.
1
𝑀𝑒𝑎𝑛_𝑖𝑛𝑓𝑜 = ∑𝑁
𝑖=1 𝑋𝑖(𝑡) (19)
𝑁

Step 3: Attack
During the GTO algorithm's exploitation (intensification) step, the goby engages in chasing and
attacking its prey, the bird. This pursuit involves the goby getting close to the bird and executing an acrobatic
leap out of the water to catch it. To replicate this behavior, the GTO algorithm incorporates the concept of
visual distortion caused by light refraction, which occurs when light waves cross the boundary between
different media such as glass, air, and water.
Based on [25], at the refractive point, both the incident and refracted rays must form an angle with
the normal line to the surface. The medium through which light rays pass is equally essential. Snell's law
clarifies this relationship using a refractive index, a fixed value for a given medium. Here, if we know the
angle of incidence, we can predict the refractive angle. Likewise, if we are aware of the refractive angle, the
incidence angle is predictable. Snell's law is shown in (20). Where Ƞ1=1.00029 and Ƞ2=1.33 are the precise
refractive indices of water and air, respectively, and 1 and 2 are the incidence and refractive angles. 2 is the
arbitrary range between (0 and 360). The (11) can be determined using (19). Following that, the visual
distortion V can be calculated using (22).

Int J Adv Appl Sci, Vol. 12, No. 3, September 2023: 195-204
Int J Adv Appl Sci ISSN: 2252-8814  199

Ƞ1 Sin θ1 = Ƞ2 Sin θ2 (20)

Ƞ1
Sin θ1 = sin θ2 (21)
Ƞ2

V = 𝑠𝑖𝑛 (𝜃1○ ) x 𝒟 (22)

The calculation of sin involves taking the sine of the variable in degrees. The distance between the
prey and the attacker denoted as 𝓓, is determined using (23). In this equation, 𝐵𝑒𝑠𝑡𝑝 represents the best
solution obtained, which denotes the quarry location. The action of the giant trevally during the pursuit and
leaping out of the water is a mathematical simulation by (24). The solution for the next iteration, X(t+1), is
generated through the attack step. The launch velocity, L, used to mimic the bird chase, can be calculated
using (25). In this equation, 𝐹_𝑜𝑏𝑗(𝑋𝑖 (𝑡)) denotes X’s fitness value as the latest iteration, (𝑡). The jump
slope function, 𝐻, in (23) facilitates the adaptive transition from the exploration phase to the exploitation
phase, and its calculation is determined by (26).

𝒟 = |(𝐵𝑒𝑠𝑡𝑝 – Xi(t))| (23)

𝑋 (𝑡 + 1) = ℒ x 𝒱 x ℋ (24)

ℒ = Xi (t) x 𝑠𝑖𝑛 (𝜃1○ ) x 𝐹_𝑜𝑏𝑗(𝑋𝑖(𝑡)) (25)

2
ℋ = 𝑅 x (2 – t x ) (26)
𝑇

In the aforementioned context, the variables 𝑡 and 𝑇 respectively denote the latest iteration and the
highest value of iterations. 𝑅 represents a random value that signifies the giant trevally fish's specific motion
sensing during the exploitation process. It is important to note that the value of 𝐻 decreases gradually from 2
to 0 as the iterations progress. During the exploitation step, the algorithm endeavors to exploit the solution
environment, utilizing this decreasing trend of H. Algorithm 1 shows the developed pseudocode of the GTO
algorithm.

Algorithm 1: Pseudocode GTO

1. Begin
2. Enter a number for 𝓐 parameter
3. Indicate the number of giant trevallies; N
4. Define the termination condition, maximum iterations (T)
5. Create a Giant Trevally population (X) at random using (21)
6. for t = 1: T
7. apply LRV to change the best position to job sequences and calculate objective
function f (X)for each search agent
8. Classify the populace
9. Identify the greatest global solution (𝐵𝑒𝑠𝑡𝐺 )
10. Identify 𝐵𝑒𝑠𝑡𝑝 as the prey’s position (best location)
11. for i = 1: N
12. Comprehensive Investigation Step
13. Using (19) and (20), compute the levy flight distribution function levy.
14. Determine new
15. If f (𝐵𝑒𝑠𝑡𝑁𝑃 ) < f (X(I,: ))
16. X (I, : ) = 𝐵𝑒𝑠𝑡𝑁𝑃
17. 𝐼𝑓 𝑓 (𝐵𝑒𝑠𝑡𝑁𝑃 ) < 𝑓(𝐵𝑒𝑠𝑡𝑃 )
18. 𝐵𝑒𝑠𝑡𝐺 = 𝐵𝑒𝑠𝑡𝑁𝑃
19. 𝐸𝑛𝑑 𝑖𝑓
20. 𝐸𝑛𝑑 𝑖𝑓
21. Selecting area step
22. Using (19), compute the mean of X
23. Using (18), Calculate 𝐵𝑒𝑠𝑡𝑁𝑃
24. Steps 15 – 20 should be repeated
25. Attacking move
26. Using (22), Determine distortion of visual 𝓥
27. Determine the launch speed 𝓛 using (25)
28. Using (24), Determine 𝐵𝑒𝑠𝑡𝑁𝑃
29. The transition from exploration to exploitation using (26)
30. Steps 15 - 20 should be repeated
31. 𝐸𝑛𝑑 𝑓𝑜𝑟 𝑡
32. Best solution and visualization after the process

Low-carbon no-idle permutation flow shop schedulling problem: giant trevall … (Dana Marsetiya Utama)
200  ISSN: 2252-8814

3.2. African vultures optimization algorithm (AVOA)

AVOA is a metaheuristic method developed by modeling and simulating African vulture foraging
behavior [17]. Based on the four requirements outlined above, the AVOA technique can be divided into five
stages in the foraging stage to emulate the behavior of different vultures.
Phase 1: Clustering
After the formation of the original population, the appropriateness of all solutions is selected in this
step. The best vulture is acknowledged as the finest option. Using (27), the first, and second solution is
regarded as the second-best vulture, while the third group is given the other vultures.

𝐵𝑒𝑠𝑡𝑉𝑢𝑙𝑡𝑢𝑟𝑒1 𝑖𝑓 𝑃𝑖 = 𝐿1
𝑅(𝑖) = { (27)
𝐵𝑒𝑠𝑡𝑉𝑢𝑙𝑡𝑢𝑟𝑒2 𝑖𝑓 𝑃𝑖 = 𝐿2

𝐵𝑒𝑠𝑡𝑉𝑢𝑙𝑡𝑢𝑟𝑒1 represents the best vulture, 𝐵𝑒𝑠𝑡𝑉𝑢𝑙𝑡𝑢𝑟𝑒2 represents the second-best, 𝐿1 and 𝐿2 are two
random numbers between (0.1) and the sum is 1. 𝑃𝑖 , is used in (28) through the roulette-wheel approach.
Here, Fi represents the fitness of the first and second groups of vultures, and n is the total number of the two
vultures.
𝐹𝑖
𝑃𝑖 = ∑𝑛
(28)
𝑖=1 𝐹𝑖

Phase 2: Starvation rate of vultures

Assume a bunch of vultures is not hungry. In that situation, they have enough energy to travel
further distances to find food. However, if they are hungry, they will not be able to preserve their long-
distance travel. As a result, hungry vultures will exhibit hostile behavior. This behavior can be calculated
using (29), which indicates the vultures' shift from exploration to exploitation. F indicates that the vultures
are full, 𝑟𝑎𝑛𝑑𝑖 is a variable with a number that fluctuates between (0.1), 𝑡 is computed by (30), and z is a
random number in the range (1.1) that changes with every repetition.

𝑖𝑡𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑖
𝐹 = (2 x 𝑟𝑎𝑛𝑑𝑖 + 1)x z x ( 1 − ) + 𝑡) (29)
𝑚𝑎𝑥𝑖𝑡𝑒𝑟𝑎𝑡𝑖𝑜𝑛

𝜋 𝑖𝑡𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑖 𝜋 𝑖𝑡𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑖
𝑡 = ℎ x (𝑠𝑖𝑛𝑤 ( 𝑥 ) + (𝑐𝑜𝑠 𝑤 ( 𝑥 )−1 (30)
2 𝑚𝑎𝑥𝑖𝑡𝑒𝑟𝑎𝑡𝑖𝑜𝑛 2 𝑚𝑎𝑥𝑖𝑡𝑒𝑟𝑎𝑡𝑖𝑜𝑛

The predefined parameter w determines the probability of a vulture performing the exploitation stage. In
addition, the latest value 𝑖𝑡𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑖 is the number of total iterations, ℎ is a random value between denoted as
iteration, and h is a random value between values (-2.2). Based on (29), 𝐹 will steadily decrease as the
number of iterations increases. When the number of |𝐹𝑖 | exceeds one, vultures enter the stage of exploration
and search for new food in diverse locales. Otherwise, vultures reach the exploitation stage, where they look
for better food in the surrounding area.
Phase 3: Exploration phase
In AVOA, two alternative techniques are employed to inspect various random places, and a
parameter designated P1 in the range of (0.1) is utilized to select one of the strategies. To determine one of
the strategies during the phase, a random strategy 𝑟𝑎𝑛𝑑 𝑃1 between (0.1) is used. If the value is better or
equal to the parameter P1 using in (31) in (32) is used if the value of 𝑟𝑎𝑛𝑑𝑖 is smaller than the parameter P1.

𝑃(𝑖 + 1) = 𝑅(𝑖) – 𝐷(𝑖) 𝑥 𝐹𝑖 (31)

𝑃(𝑖 + 1) = 𝑅(𝑖) − 𝐹𝑖 + 𝑟𝑎𝑛𝑑2 x ((𝑢𝑏 – 𝑙𝑏) x 𝑟𝑎𝑛𝑑3 + 𝑙𝑏) (32)

𝑅(𝑖) was selected as one of the finest vultures in the current iteration based on (27), 𝐹𝑖 is the current
iteration's vulture saturation rate determined in the most recent iteration based on (29), 𝑟𝑎𝑛𝑑2 is a random
value between (0.1), and 𝑙𝑏 and 𝑢𝑏 are the variable’s upper and lower bounds, respectively. 𝑟𝑎𝑛𝑑3 is
accustomed to offering a high random coefficient on the scale of the search neighborhood to boost the variety
and look for different sections of the search space. Equation (33) computes D(i), which shows the distance
between the current and optimal vultures. P(i) denotes the i-th vulture's position, while X is a random number
between (0.2).

𝐷(𝑖) = |𝑋 𝑥 𝑅(𝑖) – 𝑃(𝑖)| (33)

Int J Adv Appl Sci, Vol. 12, No. 3, September 2023: 195-204
Int J Adv Appl Sci ISSN: 2252-8814  201

Phase 4: Exploitation (first stage)

At the moment, the efficiency stage of AVOA is being explored. If |Fi| is less than one, AVOA
starts the first stage of exploitation. 𝑃2 is a parameter with a value between (0.1) that determines the chosen
strategy. At the beginning of the first stage of exploitation, and 𝑃2 a random value between (0.1) is
generated. In (34) illustrates this procedure.

𝐷(𝑖) 𝑥 (𝐹𝑖 + 𝑟𝑎𝑛𝑑4 ) − 𝑑(𝑡) 𝑖𝑓 𝑃2 ≥ 𝑟𝑎𝑛𝑑𝑝2

P(i + 1) = { (34)
𝑅(𝑖) − ( 𝑆1 + 𝑆2 ) 𝑖𝑓 𝑃2 < 𝑟𝑎𝑛𝑑𝑝2

𝑟𝑎𝑛𝑑4 is a random number between (0.1), and 𝑑(𝑡) is the separation between the vultures and one of the two
groups' best vultures, as calculated by (35). 𝑆1 and 𝑆2 are determined using (36) and (37), respectively.
Where, 𝑟𝑎𝑛𝑑5 and 𝑟𝑎𝑛𝑑6 are random values between (0.1), respectively.

𝑑(𝑖) = 𝑅(𝑖) – 𝑃(𝑖) (35)

𝑟𝑎𝑛𝑑5 𝑥 𝑃(𝑖)
𝑆1 = R(i) x ( ) x cos (𝑃(𝑖)) (36)
2𝜋

𝑟𝑎𝑛𝑑6 𝑥 𝑃(𝑖)
𝑆2 = R(i) x ( ) x sin (𝑃(𝑖)) (37)
2𝜋

Phase 5: Exploitation (second stage)

This phase of the algorithm is conducted if |𝐹𝑖| is smaller than 0.5. 𝑟𝑎𝑛𝑑3 is produced in the range
of (0.1) at the start of this phase. Thus, if 𝑃3 is higher than or equal to 𝑟𝑎𝑛𝑑3 , the approach is to attract
different vultures to the food source, resulting in competitive behavior. As a result, (38) can be used to update
the vultures' position. Equations (39) and (40) are used to calculate 𝐴1 and 𝐴2 .
𝐴1 +𝐴2
P(i + 1) = (38)
2

𝐵𝑒𝑠𝑡𝑉𝑢𝑙𝑡𝑢𝑟𝑒1 (𝑖)𝑥𝑃(𝑖)
𝐴1 = 𝐵𝑒𝑠𝑡𝑉𝑢𝑙𝑡𝑢𝑟𝑒1 − 2 𝑥 𝐹𝑖 (39)
𝐵𝑒𝑠𝑡𝑉𝑢𝑙𝑡𝑢𝑟𝑒1 (𝑖)−(𝑃(𝑖))

𝐵𝑒𝑠𝑡𝑉𝑢𝑙𝑡𝑢𝑟𝑒2 (𝑖)𝑥𝑃(𝑖)
𝐴2 = 𝐵𝑒𝑠𝑡𝑉𝑢𝑙𝑡𝑢𝑟𝑒1 − 2 𝑥 𝐹𝑖 (40)
𝐵𝑒𝑠𝑡𝑉𝑢𝑙𝑡𝑢𝑟𝑒2 (𝑖)−(𝑃(𝑖))

As AVOA progresses to the second stage, to scavenge for leftovers, vultures congregate around the
best vulture. As a consequence, (41) can be used to update the vultures' position. In this case, D represents
the problem dimension, and the AVOA efficacy is improved by employing 𝑡ℎ𝑒 𝐿é𝑣𝑦 𝑓𝑙𝑖𝑔ℎ𝑡 (𝐿𝐹) pattern,
which is calculated using in (42). Where, 𝑣 and 𝑢 are each random values between (0.1), and β is a fixed
number of 1.5. Algorithm 2 shows the development of the AVOA algorithm.

𝑃(𝑖 + 1) = 𝑅(𝑖) − |𝑑(𝑡)| 𝑥 𝐹𝑖 𝑥 𝑙𝑒𝑣𝑦(𝑑) (41)

𝑢𝑥 𝜎
𝐿𝐹 (𝑥) = 0,001 x 1 (42)
|𝑣|𝜌

1
𝜋𝛽 𝛽
Г (1+ 𝛽)𝑥 sin( 2 )
𝜎 = ( 𝛽−1 ) (43)
Г (1+ 𝛽2) 𝑥 𝛽 𝑥 2 𝑥 ( )
2

Algorithm 2: AVOA
1. The population size N and the maximum number of iterations T are inputs.
2. Outputs: The location of the vulture and its fitness value
3. Start the random population 𝑃𝑖 (i = 1,2,…,N)
4. do while (no halting condition satisfied)
5. Use LRV to alter the best position in job sequences.
6. Vulture carbon emission values
7. Make 𝑃𝐵𝑒𝑠𝑡𝑉𝑢𝑙𝑡𝑢𝑟𝑒1 , the Vulture's location (Best Vulture Category 1 first best
location).
8. Make 𝑃𝐵𝑒𝑠𝑡𝑉𝑢𝑙𝑡𝑢𝑟𝑒2 , the Vulture's position (the second best location in Best Vulture
Category 2).
9. do for (each vulture (𝑷𝒊 ))

Low-carbon no-idle permutation flow shop schedulling problem: giant trevall … (Dana Marsetiya Utama)
202  ISSN: 2252-8814

10. Using Equation (27) determine R(i)

11. Use Equation (34) to update the F
12. if (|𝑭| ≥ 1) then
13. if (𝑷𝟏 ≥ 𝒓𝒂𝒏𝒅𝑷𝟏) then
14. Update the location of the vulture using Equation (31)
15. else
16. Using Equation (32) update the location vulture
17. if (|𝑭| < 1) then
18. if (|𝑭| >0.5) then
19. if (𝑷𝟐 ≥ 𝒓𝒂𝒏𝒅𝑷𝟐 ) then
20. Using Equation (34) Update the location vulture
21. else
22. Using Equation (35) Update the location vulture
23. else
24. if (𝑷𝟑 ≥ 𝒓𝒂𝒏𝒅𝑷𝟑) then
25. Using Equation (39) Update the location vulture
26. else
27. Using Equation (42) Update the location vulture
28. Return 𝑃𝐵𝑒𝑠𝑡𝑉𝑢𝑙𝑡𝑢𝑟𝑒1

3.3. Experimental data and procedures

This research uses 3 different case studies and energy emissions per kWh of 0.998 kg in all 3 case
studies. In the first case study [26] with a problem of 10 jobs 6 machines, the second case study [27] with a
problem of 50 jobs 6 machines, and the third case study [28] with a problem of 100 jobs 10 machines. The
experimental procedure in both algorithms is run 30 times to find the minimum energy consumption. Each
case study consists of two variations of iterations (100 and 200) and population (100 and 200) in each
algorithm. To determine which method is superior and more successful between AVOA and GTO, an
independent test was used to examine the sig value (2-tailed). When the value is > 0.05, then GTO and
AVOA algorithms perform similarly. When the value is < 0.05, the GTO and AVOA algorithms behave
differently. The results of the experiment will also be presented in a boxplot diagram. This experiment was
carried out using a Windows 10 device with an Intel Core-i5 processor.

4. RESULT AND DISCUSSION

This section describes and explains the study's results by considering the effect of different
population numbers and the number of iterations on carbon emission consumption. In addition, this section
will compare the effectiveness of the GTO and AVOA algorithms with the independent sample t-test. The
following are the results and discussion of the AVOA and GTO algorithms.
Table 2 shows the comparison of populations and iteration experiment results using the GTO and
AVOA algorithms. In Case 1, the best experimental result is at iteration 100, population 200. The most
effective result in Case 2 is at iteration 200 population 200. The most effective result in Case 3 is at iteration
100 population 100. According to the experimental results, the larger the population and iteration, the lower
the carbon emission consumption. And, conversely, as the population and iterations decrease, so do the
carbon emissions produced. Table 3 displays the outcomes of running EC with the GTO and AVOA
algorithms. The results reveal that the EC in Cases 1, 2, and 3 have different values. Case 1 demonstrates that
the GTO and AVOA algorithms have the same variation in data distribution. Case 2 demonstrates that the
GTO data distribution is more diversified than the AVOA data distribution. In example 3, the GTO data
distribution is smaller than the AVOA data distribution.

Table 2. Results of iteration and population comparison experiments utilizing the GTO and AVOA
algorithms
Research Number of machines and jobs Iteration Population GTO EC consumption AVOA EC consumption
100 16833.58 16868.96
100
200 16833.58 16867.14
Case 1 10 jobs and 6 machines
100 16833.58 16867.14
200
200 16833.58 16868.96
100 11988.12 11996.24
100
200 11995.99 11995.29
Case 2 50 jobs and 10 machines
100 11993.37 11995.34
200
200 11990.34 11995.73
100 2068.451 2226.769
100
200 2069.28 2228.185
Case 3 100 jobs and 10 machines
100 2069.014 2235.167
200
200 2069.521 2222.163

Int J Adv Appl Sci, Vol. 12, No. 3, September 2023: 195-204
Int J Adv Appl Sci ISSN: 2252-8814  203

Table 3. Test result carbon emission consumption of GTO and AVOA algorithms
Cases Result GTO AVOA
Average 16833.585 16883.517
Standard Deviation 0.000 19.543
Case 1
Minimum 16833.590 16867.136
Maximum 16833.590 16930.437
Average 11995.077 11995.533
Standard Deviation 4.207 1.505
Case 2
Minimum 11990.245 11992.212
Maximum 12002.951 11998.288
Average 2069.239 2237.446
Standard Deviation 0.245 21.524
Case 3
Minimum 2068.635 2198.696
Maximum 2069.656 2278.357

Table 4. Results of the independent sample t-test

Case 1 Case 2 Case 3
AVOA GTO AVOA GTO AVOA GTO
Mean 16883.517 16833.59 11995.53 11995.08 2237.446 2069.239
Standard Deviasi 19.54305 0 1.505 4.207 21.524 0.245
t -13.676 0.559 42.799
Sig.(2-tailed) 0 0.578 0

5. CONCLUSION
This study proposes GTO and AVOA to minimize carbon emissions. To answer the NIPFSP
problem, these techniques are compared. The experimental results reveal that in case 1, with small data, the
GTO method is more effective at minimizing carbon emissions. In Case 2 which is medium data, AVOA and
GTO algorithms are effectively used to minimize carbon emissions, and In Case 3, the GTO method is
utilized more efficiently to minimize carbon emissions, which is a large amount of data in the NIPFSP
problem. The parameter testing of the GTO and AVOA algorithms reveals that the higher the iteration and
population, the less carbon is consumed. Suggestions for future research are to compare with other
algorithms with different objectives and approaches.

ACKNOWLEDGEMENTS
The authors would like to thank the Department of Industrial Engineering's Optimisation Laboratory
for allowing us to use their facilities.

REFERENCES
[1] D. M. Utama and M. D. Primayesti, “A novel hybrid Aquila optimizer for energy-efficient hybrid flow shop scheduling,” Results
in Control and Optimization, vol. 9, p. 100177, Dec. 2022, doi: 10.1016/j.rico.2022.100177.
[2] C. Li, Y. Tang, L. Cui, and P. Li, “A quantitative approach to analyze carbon emissions of CNC-based machining systems,”
Journal of Intelligent Manufacturing, vol. 26, no. 5, pp. 911–922, Oct. 2015, doi: 10.1007/s10845-013-0812-4.
[3] D. M. Utama, A. A. P. Salim, and D. S. Widodo, “A novel hybrid archimedes optimization algorithm for energy-efficient hybrid
flow shop scheduling,” International Journal of Advances in Intelligent Informatics, vol. 8, no. 2, pp. 237–250, Jul. 2022, doi:
10.26555/ijain.v8i2.724.
[4] D. M. Utama, M. D. Primayesti, S. Z. Umamy, B. M. N. Kholifa, and A. D. Yasa, “A systematic literature review on energy-
efficient hybrid flow shop scheduling,” Cogent Engineering, vol. 10, no. 1, Dec. 2023, doi: 10.1080/23311916.2023.2206074.
[5] S. Rahimifard, Y. Seow, and T. Childs, “Minimising Embodied Product Energy to support energy efficient manufacturing,” CIRP
Annals, vol. 59, no. 1, pp. 25–28, 2010, doi: 10.1016/j.cirp.2010.03.048.
[6] C. Zhang, P. Gu, and P. Jiang, “Low-carbon scheduling and estimating for a flexible job shop based on carbon footprint and
carbon efficiency of multi-job processing,” Proceedings of the Institution of Mechanical Engineers, Part B: Journal of
Engineering Manufacture, vol. 229, no. 2, pp. 328–342, Feb. 2015, doi: 10.1177/0954405414527959.
[7] J. Y. Ding, S. Song, and C. Wu, “Carbon-efficient scheduling of flow shops by multi-objective optimization,” European Journal
of Operational Research, vol. 248, no. 3, pp. 758–771, Feb. 2016, doi: 10.1016/j.ejor.2015.05.019.
[8] W. Lin, G. Tian, Z. Li, Y. Zhang, and C. Zhang, “Flow shop scheduling with low carbon emission and variable machining
parameters,” Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, vol. 233, no.
5, pp. 1561–1572, Apr. 2019, doi: 10.1177/0954405418782290.
[9] R. Asmatulu, W. Khan, and M. B. Yildirim, “Acoustical properties of electrospun nanofibers for aircraft interior noise reduction,”
ASME International Mechanical Engineering Congress and Exposition, Proceedings, 2010.
[10] G. Mouzon, M. B. Yildirim, and J. Twomey, “Operational methods for minimization of energy consumption of manufacturing
equipment,” International Journal of Production Research, vol. 45, no. 18–19, pp. 4247–4271, Sep. 2007, doi:
10.1080/00207540701450013.
[11] F. Zhao, L. Zhang, Y. Zhang, W. Ma, C. Zhang, and H. Song, “A hybrid discrete water wave optimization algorithm for the no-
idle flowshop scheduling problem with total tardiness criterion,” Expert Systems with Applications, vol. 146, p. 113166, May

Low-carbon no-idle permutation flow shop schedulling problem: giant trevall … (Dana Marsetiya Utama)
204  ISSN: 2252-8814

2020, doi: 10.1016/j.eswa.2019.113166.

[12] M. Fatih Tasgetiren, H. Öztop, L. Gao, Q. K. Pan, and X. Li, “A variable iterated local search algorithm for energy-efficient no-
idle flowshop scheduling problem,” Procedia Manufacturing, vol. 39, pp. 1185–1193, 2019, doi: 10.1016/j.promfg.2020.01.351.
[13] P. J. Kalczynski and J. Kamburowski, “A heuristic for minimizing the makespan in no-idle permutation flow shops,” Computers
and Industrial Engineering, vol. 49, no. 1, pp. 146–154, Aug. 2005, doi: 10.1016/j.cie.2005.05.002.
[14] R. K. Chakrabortty, A. Abbasi, and M. J. Ryan, “Multi-mode resource-constrained project scheduling using modified variable
neighborhood search heuristic,” International Transactions in Operational Research, vol. 27, no. 1, pp. 138–167, Jan. 2020, doi:
10.1111/itor.12644.
[15] D. S. Widodo and D. M. Utama, “The hybrid ant lion optimization flow shop scheduling problem for minimizing completion
time,” Journal of Physics: Conference Series, vol. 1569, no. 2, p. 022097, Jul. 2020, doi: 10.1088/1742-6596/1569/2/022097.
[16] H. T. Sadeeq and A. M. Abdulazeez, “Giant trevally optimizer (GTO): A novel metaheuristic algorithm for global optimization
and challenging engineering problems,” IEEE Access, vol. 10, pp. 121615–121640, 2022, doi: 10.1109/ACCESS.2022.3223388.
[17] B. Abdollahzadeh, F. S. Gharehchopogh, and S. Mirjalili, “African vultures optimization algorithm: A new nature-inspired
metaheuristic algorithm for global optimization problems,” Computers and Industrial Engineering, vol. 158, p. 107408, Aug.
2021, doi: 10.1016/j.cie.2021.107408.
[18] D. Gürses, P. Mehta, S. M. Sait, and A. R. Yildiz, “African vultures optimization algorithm for optimization of shell and tube heat
exchangers,” Materialpruefung/Materials Testing, vol. 64, no. 8, pp. 1234–1241, Aug. 2022, doi: 10.1515/mt-2022-0050.
[19] Y. Chen and G. Zhang, “New parameters identification of Proton exchange membrane fuel cell stacks based on an improved
version of African vulture optimization algorithm,” Energy Reports, vol. 8, pp. 3030–3040, Nov. 2022, doi:
10.1016/j.egyr.2022.02.066.
[20] Z. He, B. Tang, and F. Luan, “An improved African vulture optimization algorithm for dual-resource constrained multi-objective
flexible job shop scheduling problems,” Sensors, vol. 23, no. 1, p. 90, Dec. 2023, doi: 10.3390/s23010090.
[21] C. Y. Cheng, S. W. Lin, P. Pourhejazy, K. C. Ying, and Y. Z. Lin, “No-idle flowshop scheduling for energy-efficient production:
An improved optimization framework,” Mathematics, vol. 9, no. 12, p. 1335, Jun. 2021, doi: 10.3390/math9121335.
[22] X. S. Yang, T. O. Ting, and M. Karamanoglu, “Random walks, Levy flights, Markov chains and metaheuristic optimization,” in
Lecture Notes in Electrical Engineering, vol. 235 LNEE, 2013, pp. 1055–1064. doi: 10.1007/978-94-007-6516-0_116.
[23] D. M. Utama, A. K. Garside, and W. Wicaksono, “Development of a Three-Stage Hybrid Flowshop Algorithm Considering Setup
Time in Indonesian,” Jurnal Ilmiah Teknik Industri, vol. 18, no. 1, pp. 72–78, Jul. 2019, doi: 10.23917/jiti.v18i1.7683.
[24] M. Chawla and M. Duhan, “Levy flights in metaheuristics optimization algorithms–a review,” Applied Artificial Intelligence, vol.
32, no. 9–10, pp. 802–821, Nov. 2018, doi: 10.1080/08839514.2018.1508807.
[25] C. A. Bennett, Principles of Physical Optics. usa: Hoboken, NJ, 2022.
[26] J. Carlier, “Scheduling with disjunctive constraints.,” RAIRO Recherche Operationnelle, vol. 12, no. 4, pp. 333–350, Mar. 1978,
doi: 10.1051/ro/1978120403331.
[27] C. R. Reeves, “A genetic algorithm for flowshop sequencing,” Computers and Operations Research, vol. 22, no. 1, pp. 5–13, Jan.
1995, doi: 10.1016/0305-0548(93)E0014-K.
[28] J. Heller, “Some numerical experiments for an M × J flow shop and its decision-theoretical aspects,” Operations Research, vol. 8,
no. 2, pp. 178–184, Apr. 1960, doi: 10.1287/opre.8.2.178.

BIOGRAPHIES OF AUTHORS

Dana Marsetiya Utama is currently a Lecturer and a Researcher in the Industrial

Engineering Department, at the University of Muhammadiyah Malang Indonesia. His research
interests include optimization engineering, scheduling, and modeling. He received a bachelor's
degree in industrial engineering from Trunojoyo University, in 2008, and a master's degree in
industrial engineering from the University of Brawijaya, in 2011. He can be contacted at
email: [email protected]

Cantika Febrita is currently a Bachelor's Student in the Industrial Engineering

Department, at the University of Muhammadiyah Malang Indonesia. Her research interests
include optimization engineering, scheduling, and modeling. She can be contacted at email:
[email protected]

Int J Adv Appl Sci, Vol. 12, No. 3, September 2023: 195-204