You are on page 1of 11

February 8, 2022

Computers & Industrial Engineering

Dear reviewers,
We are grateful for all the comments, suggestions and shortcomings highlighted by the
reviewers. All the points suggested by the reviewers were very important to improve the
quality of the paper and brought significant changes and more clarity to the purpose of
the paper. In order to improve the quality of the paper to the focus of Computers &
Industrial Engineering (CAIE), we try to adequate the paper to the comments pointed out
by the reviewers. The integral version of the manuscript in .pdf format is attached and
modifications can be seen highlighted in red. The corrections and answers of the
comments can be seen as follows.

Best regards,

The authors.
REVIEWER #1
This paper proposed an iterated greedy algorithm for the distributed blocking flow shop
scheduling with makespan minimization considering the setup times and maintenance
operations.
1. It takes up much space to introduce the machine breakdowns and operation
maintenance, which should be more concise to improve the readability.
Answer: We thank Reviewer #1 for the relevant recommendation and comments about
our manuscript. The part of the manuscript that introduces machine breakdowns and
operation maintenance was modified. We reduce the text to improve the readability of
introduction section. The modification of this part can be seen highlighted in red in the
new version of the manuscript. We also dispose as follows the first version submitted and
the new version.

First version submitted:


In most of conventional scheduling problem models, machines are considered always
available for processing. However, in real world, machines can be unavailable by a period
of time because of breakdowns or maintenance activities reasons (Ruiz et al. (2007), Bock
et al. (2012), Yu & Seif (2016)). Although most of the literature considers separately
maintenance operations and job scheduling, both activities are closely related with each
other. While maintenance actions consume potential production time, postponing them
may increase the occurrence of machine breakdowns Ruiz et al. (2007). Furthermore, the
joint scheduling of jobs and maintenance can optimize the utilization of resources (Ruiz
et al. (2007), Xu et al. (2010), Miyata et al. (2019b), Miyata et al. (2019a)), reduction of
costs (Yu & Seif (2016), Xu et al. (2010), Xiao et al. (2016)), increase of safety and
reliability of system and ensure the availability of machines (Ruiz et al. (2007)). In
literature, the integration between maintenance operations and flow shop scheduling
problem is described under several perspectives. Aggoune (2004) considered the
incorporation of fixed and flexible maintenance activities in a flow shop to minimize
Cmax. Kubzin & Strusevich (2005) investigated the 2-machine no-wait flow shop with a
maintenance operation in one of the machines and Cmax minimization. Ruiz et al. (2007)
developed a criterion to schedule preventive maintenance operations and jobs to minimize
Cmax in the permutation flow shop. The intervals time between preventive maintenance
operations are generated by models based on statistical reliability. Allaoui et al. (2008)
studied the jointly scheduling of jobs and a maintenance operation in a 2-machine flow
shop with Cmax minimization. The maintenance activity is scheduled inside to a given
interval between [0; T]. Choi et al. (2010) proved that the ordered m-machine permutation
flow shop with maintenance operation greater than one is strongly NP-hard. The authors
considered that jobs are resumable case and the maintenance operations are scheduled in
advance. Safari et al. (2010) proposed the adaptation of metaheuristics for the m-machine
permutation flow shop with machines subject to condition-based maintenance to
minimize the expected Cmax. The authors considered shocks as the main reason for
degradation of each machine, where the number of shocks is represented by a Poisson
distribution. Khelifati & Benbouzid-Sitayeb (2011) proposed a multi-agent system for the
joint maintenance and job scheduling for the m-machine flow shop to minimize the sum
of Cmax and temporal deviation of MP operations. Sitayeb et al. (2011) developed some
strategies to joint scheduling jobs and maintenance operations for the m-machine flow
shop with Cmax minimization. The authors established that maintenance operations are
carried out in T known periods within intervals that allows flexibility of maintenance
actions. Chiu & Shih (2012) proposed models to minimize Cmax in a 2 machine flow shop
with maintenance operations and rush orders. Hnaien & Yalaoui (2013) propose a mixed
integer linear programming (MILP) and metaheuristics to minimize Cmax and the sum of
earliness and tardiness due to maintenance in the 2-machine flow shop with maintenance
operations based on frequencies. Vahedi-Nouri et al. (2013) and Vahedi-Nouri et al.
(2014) investigated the integration between learning effects of jobs and preventive
maintenance operations in the non-permutation flow shop with sum of tardiness and
maintenance costs minimization and total flow time minimization, respectively.
Preventive maintenance frequencies are defined according to predefined interval times
(Vahedi-Nouri et al. (2013)) or according to a given number of processed jobs (Vahedi-
Nouri et al. (2014)). Wang & Liu (2014) evaluated the scheduling of jobs and preventive
maintenance operations in a two-machine flow shop with Cmax minimization. The authors
established the interval between preventive maintenance operations through Weibull
distribution. Yu & Seif (2016) extended the concept of diverse and flexible condition-
based maintenance by Bock et al. (2012) to the flow shop scheduling problem to minimize
the sum of tardiness and maintenance costs. Xu et al. (2018) proposed mixed integer
programming formulations for 2-machine flow shop with machine availability constraints
Cmax minimization. The authors considered two separated problems: when the
unavailability occurs only in the first machine and the second is always available and
vice-versa. Seif et al. (2018) investigated the integration between maintenance levels and
flow shop with sum of tardiness and MP costs minimization. Additionally, the authors
considered the age of machine and the impact of the machines health on the processing
time of jobs. Miyata et al. (2019a) extended the preventive maintenance approach by Ruiz
et al. (2007) to m-machine no-wait flow shop with total flow time minimization,
evaluating the performance of constructive heuristics to solve the problem. Miyata et al.
(2019b) applied the flexible and diverse maintenance activities by Yu & Seif (2016) to
the m-machine no-wait flow shop with sequence-dependent setup times with Cmax
minimization. Seif et al. (2019) applied the concept of maintenance levels to the stochastic
scheduling in the m-machine permutation flow shop with sum of tardiness and MP costs
minimization. The authors included uncertainties in both processing times and duration
of maintenance activities. Perez-Gonzalez et al. (2020) considered the 2-machine flow
shop with fixed periodic maintenance and Cmax minimization. A survey of scheduling
problem with availability machines can be seen in Ma et al. (2010).

New version submitted:


In most of conventional scheduling problem models, machines are considered always
available for processing. However, in real world, machines can be unavailable by a period
of time because of breakdowns or maintenance activities reasons (Ruiz et al. (2007), Bock
et al. (2012), Yu and Seif (2016)). Although most of the literature considers separately
maintenance operations and job scheduling, both activities are closely related with each
other. While maintenance actions consume potential production time, postponing them
may increase the occurrence of machine breakdowns Ruiz et al. (2007). Furthermore, the
joint scheduling of jobs and maintenance can optimize the utilization of resources (Ruiz
et al. (2007), Xu et al. (2010), Miyata et al. (2019b), Miyata et al. (2019a)), reduction of
costs (Yu and Seif (2016), Xu et al. (2010), Xiao et al. (2016), Xu et al. (2018)), increase
of safety and reliability of system and Distributed blocking flow shop with setup times
and maintenance operations ensure the availability of machines (Ruiz et al. (2007)). In
literature, maintenance activities is described under several perspectives. Maintenance
operations can be considered fixed and flexible (Aggoune (2004), Kubzin and Strusevich
(2005), Allaoui et al. (2008), Choi et al. (2010), Chiu and Shih (2012)), described as
quantitative models (Ruiz et al. (2007), Miyata et al. (2019a)), described as actions to a
certain pre-defined interval of time (Sitayeb et al. (2011), Vahedi-Nouri et al. (2013),
Vahedi-Nouri et al. (2014)) and frequencies (Hnaien and Yalaoui (2013)). Also, one can
see the integration between condition-based maintenance and scheduling (Safari et al.
(2010), Bock et al. (2012), Yu and Seif (2016), Seif et al. (2018), Seif et al. (2019), Miyata
et al. (2019b)) and integration between scheduling of jobs and periodic maintenance
(Perez-Gonzalez et al. (2020)). A survey of scheduling problem with availability
machines can be seen in Ma et al. (2010).

2. Figure 1 seems too ambiguous. Its resolution should be improved.


Answer: We try to improve the resolution of Figure 1. The new version of the Figure 1
can be seen in the new version of the manuscript. We also dispose as follows the Figure
1 from the first version submitted and the current version.

First version submitted:

Figure 1. 2-machine blocking flow shop with sequence-dependent setup times and
maintenance operations

New version submitted:


Figure 1. 2-machine blocking flow shop with sequence-dependent setup times and
maintenance operations

3. Most notations in section 2.1 is repeated as the notations of the problem definition,
which seems unnecessary.
Answer: We remove repeated notations in section 2.1 and added all the notations in
section 2.

4. The procedure of LSPM,RE is hard to comprehend. A figure to illustrate the


example of LSPM,RE can be help to understand.
Answer: We generate Figure 2 in order to illustrate part of the functioning of LS PM,RE.
Figure 2 can be seen in the new version of the manuscript. We also dispose Figure 2 as
follows:
Figure 2. Reassignment of maintenance activities of level 1 of the first machine

5. The complete spellings of "EHPF2-B" should be given when it first occurs.


Answer: We modify the part of the text that mention for the first time EHPF2-B. The new
version of the text can be seen highlighted in red in the new version of the manuscript.
We also dispose as follows the first version submitted that mention it and the current
version submitted.

First version submitted:


For example, during the functioning of EHPF2-B, only Ruiz et al. procedure is used to
incorporate the PM operations to partial sequence of jobs.

New version submitted:


For example, during the functioning of extended HPF2 with both adjacent jobs
(EHPF2_B), only Ruiz et al. procedure is used to incorporate the PM operations to partial
sequence of jobs.

6. The last line in Page 14 and the line after equation (24), "Where" should be
"where".
Answer: We modify “Where” to “where”.

7. It is suggested that the algorithm complexity analysis should be given.


Answer: We added the algorithm complexity analysis (complexity time) of the algorithm
in Section 3.5. We also dispose as follows the part of the text that explain the time
complexity of the algorithm.
For the IG\_NM, one can assume that the number of jobs in each factory is nf and the
iterations in the “while” loop of IG is 𝛿1 . The time complexity of testing a certain job in
all positions of one factory (as in the first part of reconstruction phase of IG and insertion
neighborhood structure) is O(nf × m × (l + 1)). The complexity of the second part of
reconstruction is O(d × (F × nf × m × (l + 1))). One can assume that 𝜔1 is the “while”
loop of insertion and permutation neighborhood structure. The time complexity of each
neighborhood structure is O($\omega_{1}$ nf ² × F × m × (l + 1)) One can assume that
𝛿2 is the number of iterations in the “while” loop of VNS (local search) phase. The time
complexity of VNS is $O$(2 × 𝛿2 × (𝜔1 × nf × F × m × (l + 1))). Thus, the time complexity
of IG is O(nf × m × (l + 1) × ((d × (1 + F) + 2 × 𝛿2 × 𝜔1 × (nf × F)).

8. Throughout the whole algorithm, each component of the IG is referred from the
existing algorithms and methods. The contributions and novelties about the
algorithm perspective should be clarified and reflected.
Answer: We try to bring more clarity of how we added some new elements to our
proposed IG. We rewrite the parts of section 3 that describes the reconstruction phase and
VNS (local search phase). The main differences were added in these two parts of the
structure of the algorithm. The parts added to the text can be seen in Section 3.

Furthermore, we thank Reviewer #1 for the recommendation and precious comments,


since it was relevant to improve the quality of the paper.

Best regards,

The authors.
REVIEWER #2

This paper proposes an iterated greedy algorithm for the DBFSP with setup times and
maintenance operations. The research problem is worth of investigation because can help
companies to improve their efficiency since considering the maintenance operations
together with the productive operations allows minimizing idle times. However, I have
some concerns that need to be addressed:
1. There are some recent papers for solving the distributed flow shop with sequence-
dependent setup times and the parallel blocking flow shop with sequence-dependent
setup times, that should be revised because, at least one of them, proposes a similar
algorithm than the one proposed in this manuscript. I suggest to explain the
differences, if any, and compare your algorithm against these one.
References:
Huang, J-P, Pan, Q-K.; Gao, L. (2020) An effective iterated greedy method for the
distributed permutation flowshop scheduling problem with sequence-dependent
setup times. Swarm and Evolutionary Computation, 59, 100742
Huang, J-P, Pan, Q-K.;Miao, Z-H; Gao, L. (2021) Effective constructive heuristics
and discrete bee colony optimization for distributed flowshop with setup times.
Engineering Applications of Artificial Intelligence, 97, 104016
Ribas, I.; Companys, R.; Tort-Martorell, X. (2021). An iterated greedy algorithm
for the parallel blocking flow shop scheduling problem and sequence-dependent
setup times. Expert Systems with Applications, 184, 115535
Ribas, I.; Companys, R. (2021). A computational evaluation of constructive
heuristics for the parallel blocking flow shop problem with sequence-dependent
setup times. International Journal of Industrial Engineering Computations, 12 (3):
321-328

8. Page 23. In the line before section 4.1 it is said that the algorithm is compared with
the one proposed by Ruiz (2019) and Shao et al (2020) because are the most recent
metaheuristics for the DBFSP. But, as I mentioned before, recently an algorithm for
the DBFSP with sequence dependent setup times was proposed in Ribas et al (2021).
Answer: We thank the reviewer for the suggestions. We implemented all the
metaheuristics and we added them to the computational results. For this reason, most of
the computational results Section were modified. The modifications can be seen in the
new version of the manuscript submitted. All the references were included to the text.

2. In the notation presented in section 2, I propose to add to complete the list


3. In section 2.1 is not necessary to repeat the notation defined before. Thus, I suggest
including only the new notation.
Answer: We added the notation MAt,j in Section 2. We removed the repeated notation
and included the remaining notations in the list of notations of Section 2. The modification
of the list of notations can be seen in the new version of the manuscript submitted.

4. Page 12, line 3. If I am not wrong the first job in the sequence has to be J2. This
applies to all examples where the sequence appears.
Answer: We modify all the sequences where J2 was not included.

5. I recommend adding a pseudocode of LSPM,RE (page 12)


Answer: We added a pseudocode of LSPM,RE in the section that mentioned it. A new
pseudocode was generated and it is called Alg. 1 - Heuristic procedure to reassign PM
activities.

6. Page 14. In section 3.1, seven lines below the title of the section appears the
reference "Ruiz et al (2019)" if I'm not wrong this reference has to be Ruiz et al
(2010).
Answer: We thank the reviewer for this comment. We modify the reference to Ruiz et al.
(2010).

7. Page 15. Eq 24. Please revise this equation because there are extra symbols in it (-
-, xx)
Answer: We revised the equation and other constraints of MILP which were in the same
context. The modifications can be seen in the Sections that each equation belongs to.

9. Which is the stopping criterion used? It is not said. Is it the same when calibrating
the algorithm than when running the experiments?
Answer: We added to the text the part that mention the stopping criterion used. We use
the number of iterations as stopping criterion for the metaheuristics and limit time of
3,600 seconds to MILP. In order to measure the efficiency of the algorithm we also use
computational time. The part of the text that mention the stopping criterion used to
evaluate the algorithms can be seen as follows.

New version submitted:

The stopping criterion was set as the number of iteration Nit, in which Nit = {1, 2, 3, 4,

5}. Computational times of the methods were also recorded with the aim to evaluate the
efficiency of the algorithm to solve the problems.

10. Pag 31 the theoretical explanation is not necessary.


Answer: We remove the part of text of Section 4 that explain ANOVA and non-
parametric Kruskal Wallis Statistical test. The modifications can be seen in the new
version of the paper.

Furthermore, we thank Reviewer #2 for the recommendation and precious comments,


since it was relevant to improve the quality of the paper.

Best regards,

The authors.

You might also like