Computers & Industrial Engineering 149 (2020) 106799

Contents lists available at ScienceDirect

Computers & Industrial Engineering

journal homepage: www.elsevier.com/locate/caie

Dynamic integrated process planning, scheduling and due-date assignment

using ant colony optimization
Halil Ibrahim Demir a , Caner Erden b ,∗
a
Sakarya University, Department of Industrial Engineering, Sakarya, 54187, Turkey
b
Sakarya University of Applied Sciences, Faculty of Applied Sciences, Sakarya, 54187, Turkey

ARTICLE INFO ABSTRACT

Keywords: This paper presents two well-known meta-heuristics which are Genetic Algorithm (GA) and Ant Colony
Integrated process planning and scheduling Optimization Algorithm (ACO) to solve the dynamic integrated process planning, scheduling and due date
Scheduling with due date assignment assignment problem (DIPPSDDA) in which jobs arrive to the shop floor randomly. In this study, it is aimed to
Weighted Dynamic Scheduling
find the best combination of dispatching rule, due date assignment rule and route of all job with the objective
Integrated process planning
of minimizing earliness, tardiness and due-dates of each jobs. 8 different size shop floors for the comparison of
Dynamic scheduling and due date assignment
Ant Colony Optimization
the GA and ACO algorithms performances have been developed. As a result of the experimental study, it was
concluded that ACO algorithm outperformed GA algorithm. In addition, it has been suggested that integrated
approaches can provide more global manufacturing efficiency than individual approaches.

1. Introduction (Genetic Algorithms), PSO (Particle Swarm Optimization), EA (Evo-

lutionary Algorithm), HBO (Honey Bee optimization), SA (Simulated
Traditionally three important manufacturing functions which are Annealing), ACO (Ant Colony Optimization), ES (Evolutionary Strate-
process planning, scheduling and due-date assignment are handled gies), TS (Tabu Search) and some hybrid approaches. But there is a
separately and sequentially. Fierce competition in the global and local single study on DIPPSDDA utilized GA, SA, TS and hybrid of these
market forced firms to produce cheaper and faster also pushed them to algorithms such as; GA/SA and GA/TA (Erden et al., 2019). ACO is
save from costs. Since integrated solutions are better in terms of global a common and well-known meta-heuristic algorithm and also one of
manufacturing performance, integration of these functions gained at- the algorithms that widely used to solve integrated studies. Vinod and
traction recent decades in the literature. If literature is reviewed for the Sridharan (2011) discussed the integration of scheduling and due date
related research, numerous studies can be found on IPPS (Integrated assignment rules in a dynamic production environment using ACO.
Process Planning and Scheduling) which can provide more effective In that study, due-date assignment rules such as PPW (Processing-
process plans and schedules in a manufacturing environment (Chrys- time-plus-wait), TWK (Total Work Content), dynamic TWK and seven
solouris et al., 1985). In the same direction, there is also tremendous dispatching rules such as FIFO (First in First Out), SPT (Shortest Pro-
amount of studies on SWDDA (Scheduling with Due-date Assignment) cessing Time) were integrated. Zhang and Wong (2016a) also focused
which is used to improve the performance of scheduling and due date on solving IPPS problem using ACO system. Leung et al. (2010) applied
assignment functions and make manufacturing more responsive. As a Multi Agent System framed with ACO to solve IPPS with the objective
result, interest in integrated studies continued with the combination of of minimizing makespan in a manufacturing environment. As a result
three functions IPPSDDA (Integrated Process Planning and Scheduling of the study, they concluded that Multi Agent System framed with
and Due-Date Assignment) and DIPPSDDA(Dynamic IPPSDDA). How- ACO can provide a good solution for IPPS. Wang et al. (2014) studied
ever, there are relatively unexplored, especially in a dynamic stochastic a AND/OR graph-based ACO to solve the same problem and try to
environment that need to be further studied in the future. In this study, optimize the same objective with the study of Leung et al. (2010). They
integration of these three manufacturing functions will be discussed. first create a AND/OR graph to represent IPPS and secondly applied
Since integration problems are classified as NP-Hard problems, ACO to solve IPPS based on the graph. Zhang and Wong (2014) applied
several meta-heuristic algorithms have been applied to solve integra- an enhanced ACO to solve IPPS problem in the job shop environment.
tion problems (IPPS, SWDDA, IPPSDDA and DIPPSDDA) such as; GA They tried to minimize makespan, job flow time and computation

∗ Corresponding author.
E-mail addresses: [email protected] (H.I. Demir), [email protected] (C. Erden).
URLs: http://hidemir.sakarya.edu.tr (H.I. Demir), http://cerden.subu.edu.tr (C. Erden).

https://doi.org/10.1016/j.cie.2020.106799

Available online 11 September 2020

0360-8352/© 2020 Elsevier Ltd. All rights reserved.
H.I. Demir and C. Erden Computers & Industrial Engineering 149 (2020) 106799

time. As a conclusion, they have found that enhanced ACO is better process plans for uncertain dynamic events and occurrences and to
than other meta-heuristics algorithms. Uslu et al. (2018) solved the obtain better machine loading and higher global benefit. Literature
IPPS problem with a hybrid algorithm. In this study, ACO and GA review studies on IPPS can be found in Li et al. (2000), Phanden et al.
were used together. As a result of the study, it was emphasized that (2020, 2011), Tan and Khoshnevis (2000).
better performances will be achieved if GA and ACO work together. Only the job shop scheduling problem belongs to NP-Hard problem.
In addition, hybrid studies have been proposed for solving the IPPS And integrated studies include scheduling problem. Therefore, meta-
problem. heuristic methods for solving integrated studies have been often used.
As it can be seen from the given studies, ACO is a promising There are some early research used some mathematical model to solve
algorithm for solving integrated studies. This paper also investigates the problem. They are decomposed the problem into process plan
the approach of applying ACO to solve dynamic IPPSDDA in order to selection and loading sub problems and solved later (Hutchison et al.,
minimize earliness, tardiness and due-date time of a DJSS (Dynamic 1991; Nasr & Elsayed, 1990). Later some common meta-heuristics are
Job Shop Scheduling). In this study, process planning is included to the applied. Lihong and Shengping (2012), Morad and Zalzala (1999), Park
integrated system as a contribution to study of Vinod and Sridharan and Choi (2006a, 2006b) used GA (Genetic Algorithms), Amin-Naseri
(2011). The key points of this work can be highlighted as follows. and Afshari (2012) used hybrid GA and Petrović et al. (2016) used
Firstly, since there is categorical data in dynamic IPPSDDA problem. chaotic PSO to integrate process planning and scheduling. Lin et al.
Thus, the probability of belonging to each feature has been used in (1997), Madureira et al. (2003) used GA to solve dynamic job shop
optimization. Probabilities of ants choosing paths is updated in each scheduling problem. Wong et al. (2006) developed agent-based nego-
iteration to improve performance. Secondly, all the weighted due-dates, tiation approach to IPPS in a job shop kind of flexible manufacturing
earliness and tardiness are all properly penalized while only tardiness is environment. Li and McMahon (2007) used simulated annealing-based
penalized in a traditional manufacturing system. Because, according to approach to facilitate the integration and optimization process of IPPS.
JIT (Just in Time) philosophy both earliness and tardiness are undesired Shukla et al. (2008) conceptualized a bidding-based multiagent
and no customer prefer long due dates. Thirdly, earlier research on system for solving integrated process-planning and scheduling prob-
SWDDA generally focused on due-dates which are assigned without lem. Shen et al. (2006) used an agent-based approach to develop
considering the weights of the customers. In this study weighted due- a distributed process planning and scheduling system to allow geo-
date assignment approach is utilized. So, important customers that are graphically dispersed entities to work cooperatively towards the global
given relatively closer due dates are scheduled earlier for closer due- goals. Guo et al. (2009b) used PSO algorithm to optimize IPPS prob-
dates. By applying weighted due-date assignment and later by weighted lem. Liu et al. (2015) proposed ACO algorithm in the solution of
scheduling and by integration of these three manufacturing functions IPPS problem. Xia et al. (2016) solved IPPS problem for dynamic
substantial improvements are tried to be obtained in terms of weighted environment where jobs arrive randomly through the time and ma-
earliness, tardiness and due-date related costs. chine breakdowns are assumed possible. They used a hybrid genetic
The rest of the paper is given as follows. Studies on DJSS, IPPS, algorithm with variable neighborhood search in the solution of DIPPS
SWDDA, IPPSDDA are reviewed in Section 2. Problem definition and (Dynamic Integrated Process Planning and Scheduling). Keddari et al.
data are given in Section 3. Section 4 gives a brief information on ACO (2018) presented a new hybrid method that integrates the two func-
and implementing it to solve problem studied. Following sections give tions in order to minimize the makespan. Their method is made up of a
experimental results and conclusion of the paper. shifting bottleneck heuristic as a starting solution, TS and the kangaroo
algorithm metaheuristics as a global search and they compared their
2. Literature survey results with GA. Yu et al. (2018) studied DIPPS problem by using a
discrete PSO algorithm. Zhao et al. (2006) studied IPPS problem and
Process planning, scheduling and due-date assignment are three im- they used PSO algorithm to balance the machine loads.
portant manufacturing functions that should be handled concurrently In recent years, the IPPS problem has been tried to be solved
to get better global performance. There is plenty of research on IPPS with hybrid algorithms. For instance, GA and tabu search Zhang et al.
and SWDDA such as Baykasoğlu and Özbakır (2009), Chiang and Fu (2013), hybrid artificial bee colony Wang et al. (2013). Also, Yu et al.
(2007), Gordon et al. (2012), Kumar and Rajotia (2006), Min and (2015) tried to solve IPPS problem by using hybrid GA/PSO meta-
Cheng (2006), Yu et al. (2015), Zhao et al. (2018). In this section, a heuristics. Zhang et al. (2014) proposed a hybrid sampling strategy-
brief review of the integration studies is given. based multi objective evolutionary algorithm (HSS-MOEA) to deal with
the IPPS problem. Jin et al. (2015) applied a HBMO (Hybrid Honey
2.1. IPPS (Integrated Process Planning and Scheduling) Bee Mating Optimization), which combines the HBMO algorithm and
variable neighborhood search to solve IPPS problem.
Process planning converts the design data into best methods to The performance of scheduling is affected by the selected dis-
manufacture a part or delivery a service (Khoshnevis & Chen, 1991). patching rules. Many studies on this subject can be found in the
Process planning can be prepared either manually or by using comput- literature (Dominic et al., 2003; Kaban et al., 2012; Sweeney et al.,
ers (Meenakshi Sundaram & Fu, 1988). Process planning is prepared 2019). Baykasoğlu and Özbakır (2010) examined the effect of dispatch-
using computers in CAPP (Computer aided process planning) which ing rules in their study. In this study, different dispatching rules such as
enables to generate alternative process plans easier compared to tradi- SPT, EDD, LPT, ERD, MS were tested at different levels of flexibility. As
tional process planning methods (Yusof & Latif, 2014). Conventionally, a result of the study, dispatching rules have an effect on the scheduling
process plans are prepared before scheduling without considering shop optimization criterion. Sharma and Jain (2014) developed a similar
floor status and conflicting scheduling objectives. Because of conflict- study in a stochastic and dynamic production environment with nine
ing objectives of these functions and inability to communicate the different dispatching rules. These two studies show the benefit of
dynamic changes in the job shop, it is better to integrate these two working with different dispatching rules. There is also a survey article
functions (Nasr & Elsayed, 1990). Before integration studies, process on the effect of dispatching rules (Durasević & Jakobović, 2018).
planners were selecting some desired machines frequently and not
selecting some other machines at all. This was causing unbalanced 2.2. SWDDA (Scheduling with due-date assignment)
machine load at the shop floor level. Thanks to CAPP, it is easier to
prepare process plans also it is better to have alternative process plans Due-dates can be given externally, internally or through negotiation
to balance shop floor (Aldakhilallah & Ramesh, 1999). Unlike tradi- with the customers. At the first case, firm produce according to exoge-
tional way it is better to integrate two functions, prepare alternative nously given due dates and try to optimize production according to

2
H.I. Demir and C. Erden Computers & Industrial Engineering 149 (2020) 106799

these dates. At the second case, firm determine due dates endogenously floor efficiency. Another decision variable that affects the performance
and try to optimize performance and give the most suitable dates for of scheduling is the dispatching rules. In addition, the functions of
the firm. At the third case customer and firm negotiate on due-dates due date assignment together with scheduling were also employed
and they try to agree on due dates that is the most suitable for both to prevent giving an early or late due date to the customers by the
sides (Lawrence, 1994). researchers. With all of this in mind, we can see that current studies
Another important integration problem is SWDDA. The scheduling are focused on working to integrate more functions.
problems included due-date assignment are of permanent interest. Over Scheduling problems assume that there are a limited number of
the last three decades, there are numerous studies on SWDDA and jobs and a limited number of machines in a shop floor. Generally,
lately there are studies on SWDWA (Scheduling with due-window the number of machines that are shown with m the number of jobs
assignment) (Gordon et al., 2002a; Janiak et al., 2015; Li, 2015). At is shown by 𝑛. Typically, 𝑗 indicates the job. Similarly, 𝑖 indicates
the former case suitable common due-date or separate due-dates are the machine (Pinedo, 2012). And the integrated process planning and
assigned to the jobs waiting. At the latter case instead of a single point scheduling is defined as follows: 𝑛 jobs with different size of operations
in time a due-window or windows are tried to be assigned to the jobs and alternative process plans will be assigned to 𝑚 machines with the
waiting. For the surveys on SWDDA it is better to review (Cheng & some objectives (Guo et al., 2009a). In a dynamic IPPS problem there
Gupta, 1989; Gordon et al., 2002b; Lauff & Werner, 2004; Sen & Gupta, are some dynamic events possible throughout the time.
1984) for detailed information on SWDDA. The assumptions made in this study are as follows;
Other related studies can be given as follows. SWDDA studies
with common due date assignment Gordon et al. (2002a), studies 1. There are n jobs and there are more than one operation for each
assign unique due date for every job (Gordon & Strusevich, 2009), job.
single machine environment (Gordon et al., 2002a; Lin et al., 2006; 2. The arrival time of a job is random and follows an exponential
Tuong & Soukhal, 2010), two machine flow shop scheduling (Birman & distribution with 1∕𝜆 average.
Mosheiov, 2004), multi machine environment (Lauff & Werner, 2004) 3. The processing time of each operation is not deterministic and
and job shop environment (Baykasoğlu & Özbakır, 2009). is dispersed normally.
Baykasoğlu et al. (2008) proposed two approaches to determine the 4. Each job has different and its own process route.
due-date times in their study. These approaches are based on statistical 5. If an operation has predecessors before itself, It cannot be done
inferences. It was concluded that the methods developed gave better before all predecessors are performed.
results in terms of mean absolute percent error, mean percent error and 6. Each of the jobs has operations that can be performed on differ-
mean tardiness. ent machines.
After IPPS and SWDDA sub integrations, Demir and his colleagues 7. When a new job arrives in the system, an appropriate due time is
are studying on IPPSDDA and DIPPSDDA (Dynamic IPPSDDA) prob- assigned to the job. If some jobs are currently being processed by
lems as a higher level of integration over a decade (Demir & Taskin, the same machine then jobs to be processed next will be selected
2005). They integrated due-date assignment with IPPS problems and according to dispatching rules.
utilized common meta-heuristics. They used some meta-heuristics such 8. Each operation can only be operated on one machine, and in the
as genetic algorithms, evolutionary strategies, Modified Discrete PSO same way a machine can only perform one operation at the same
and SA (Demir et al., 2017). They commonly used weighted due- time. It is assumed that the machines will not breakdown.
date assignment methods. By applying weighted due-date assignment
approach valuable customers were given earlier due dates and less im-
3.1. Experimental design
portant customers were given relatively longer due dates and important
customers were scheduled earlier using weighted dispatching rules. By
As an example of the problem in a dynamic job shop, in this paper,
doing so, substantial savings are tried to be obtained through weighed
8 shop floors are tested with different job sizes and number of machines
earliness, tardiness and due-date related penalty costs.
which are given in Table 1. For instance, at the smallest shop floor there
Up to now, there is only a single study is carried out on DIPPSDDA
are 25 jobs, 5 machines and each job have 5 alternative routes and at
problem. Erden et al. (2019) applied GA, SA, TS and their hybrids with
each route there are 10 operations. Operation times of each operations
GA to the single example of this problem. In this study, ACO is applied
varies in between practically 1 and 30 according to the formula (12+𝑧 ∗
to solve DIPPSDDA problem as the structure of the problem is suitable
6) in which 𝑧 is the z-value of standard normal distribution mainly in
to utilize ACO meta-heuristic.
between −3 and +3. All the data produced specifically for the problem
3. Problem definition are attached to the article as supplementary files. Data information is
given in Section 6.
In this study unlike previous research in the literature more man-
ufacturing functions are integrated. Traditionally process planning, 3.2. Rules and formulas
scheduling and due-date assignment functions were being handled sep-
arately and sequentially. During the recent decades there are numerous When a new job arrives to the system, at first, due date assignment
research on IPPS and SWDDA where process planning and scheduling rule must be determined to solve the problem as shown in block
functions are integrated at the previous case and scheduling and due- diagram 1. In this study, 9 different due date assignment rules discussed
date assignment functions are integrated at the latter case. So, this with different customers’ weights. In addition, new rules have been
higher integration problem is quite a novel problem to study further. added according to the multiple parameters of a rule. Thus, the number
The problem of this study is to optimize the job shop problem in a of rules increased to 36. While the weighted rules take into account
dynamic environment by considering process planning, scheduling and the customer weight, in other rules the weights of the customers are
due date assignment date as an integrated problem. New job arrivals not taken into account. For instance, when determining the due date
have been taken as a dynamic event. The job shop is thought to be in the SLK (Slack) rule, the process time of 𝑖th job(𝑝𝑖 ) is added to the
instantly prepared at t=0, but in a way that a job can arrive at any arrival time of 𝑖th job(𝑎𝑖 ) as given in Eq. (1). The 𝑞 value is calculated
time to the system. Thus, the conditions of disruption of productions by multiplying the average processing time of all jobs (𝑃𝑎𝑣 ) value with
in real shop floors were predicted and process plans and schedules a certain coefficient. In the WSLK (Weighted SLK) rule, as in the SLK
were created. It is shown that working with alternative process plans rule, operations are performed and eventually the 𝑞 value is multiplied
in scheduling will introduce significant improvements to the shop by a value inversely proportional to the 𝑖th job weight (𝑧𝑖 ). Here, too, 𝑞

3
H.I. Demir and C. Erden Computers & Industrial Engineering 149 (2020) 106799

Table 1
Characteristics of shop floors.
Shop floors # of Jobs # of Machines # of Routes Processing times # of Operations
SF1 25 5 5 (12 + 𝑧 ∗ 6) 10
SF2 50 10 5 (12 + 𝑧 ∗ 6) 10
SF3 75 15 5 (12 + 𝑧 ∗ 6) 10
SF4 100 20 5 (12 + 𝑧 ∗ 6) 10
SF5 125 25 3 (12 + 𝑧 ∗ 6) 10
SF6 150 30 3 (12 + 𝑧 ∗ 6) 10
SF7 175 35 3 (12 + 𝑧 ∗ 6) 10
SF8 200 40 3 (12 + 𝑧 ∗ 6) 10

In PPW (Processing-time-plus-wait), processing time is multiplied by

a 𝑘 coefficient and a constant that is determined according to 𝑃𝑎𝑣 value
is added to the arrival time. In WPPW (Weighted PPW) rule again two
coefficients which are inversely determined according to the weights
of the jobs are used as multiplier with both k and q values. In PWW
and WPPW rules three different k coefficients and three different q
constants are used (Cheng & Gordon, 1994). So with the multipliers and
constants nine rules are used for each PPW and WPPW rule formulas
are given as follows:
𝑑𝑖1 = 𝑎𝑖 + 𝑘𝑥 𝑝𝑖 + 𝑞𝑥
(5)
𝑑𝑖2 = 𝑎𝑖 + 𝑧1𝑖 𝑘𝑥 𝑝𝑖 + 𝑧2𝑖 𝑞𝑥
The rules that are addressed after the due date rule are the dispatch-
ing rules. These rules are about which job to choose after the end of
the job in front of a machine. In this study, 11 different dispatching
rules were used. Like the due date rules, the dispatching rules include
weighted and unweighted rules. ATP (Apparent Tardiness Cost) rule,
which is a composite dispatching rule, uses the following formula, and
the job with a low index value is determined as the next job (Lee &
Pinedo, 1997).
𝑝𝑖 𝑚𝑎𝑥(𝑠𝑙𝑎𝑐𝑘,0)
𝐼𝑖 = 𝑒 𝐾𝑝 (6)
𝑤𝑖
Fig. 1. Block Diagram of the Problem Flow. MS (Minimum Slack) also prefers a job with the highest index value
and slacks and priority values are determined as follows:
𝑠𝑙𝑎𝑐𝑘 = 𝑑𝑖 − 𝑝𝑖 − 𝑡
values have three different values. Thus, 3 different rules for each SLK (7)
𝐼𝑖 = (𝑠𝑙𝑎𝑐𝑘)𝑤𝑖
and WSLK are evaluated (Baker & Bertrand, 1981; Chang, 1996).
Other common dispatching rules in literature which are applied
𝑑𝑖1 = 𝑎𝑖 + 𝑝𝑖 + 𝑞𝑥 are SPT (Shortest Processing Time), LPT (Longest Processing Time),
(1) SOT (Shortest Operation Time), LOT (Longest Operation Time), EDD
𝑑𝑖2 = 𝑎𝑖 + 𝑝𝑖 + 𝑞𝑥 𝑧𝑖 (Earliest Due Date), ERD (Earliest Release Date), SIRO (Service In
In the TWK (Total Work Content), the process time of the job is Random Order), FIFO (First In First Out) and LIFO (Last In First Out).
multiplied by a 𝑘 coefficient and added to the time of arrival. 3 different Besides, unweighted dispatching rules weighted dispatching rules
k coefficients make up 3 different rules. Similarly, the WTWK(Weighted are also applied such as WATC (Weighted Apparent Tardiness Cost),
TWK) takes into account inversely weighted coefficients (Cheng, 1987). WMS, WSPT, WLPT, WSOT, WLOT, WEDD, and WERD rules are also
Weighted WTWK formulas are given as follows: used to minimize global weighted penalty function.
𝑑𝑖1 = 𝑎𝑖 + 𝑘𝑥 𝑝𝑖 After the due date for a job is determined, shop floor operations
(2) are carried out together with other jobs according to the dispatching
𝑑𝑖2 = 𝑎𝑖 + 𝑘𝑥 𝑝𝑖 𝑧𝑖 rules. Then the time of the job to leave the shop is compared to the
NOPPT (Number of Operations Plus Processing Time) also consid- due date time. If the job is completed earlier than the due date time,
ers processing time plus number of operations (𝑜𝑖 ). In other words, early completion is involved. If the work is delivered after the due date
processing times are added to the value of the number of operations time, the penalty for late completion is carried out. Here a job shop with
are multiplied with coefficient 𝑘. WNOPPT (Weighted NOPPT), on the one shift is studied and there are 480 min per day within 8 h. All the
other hand, number of operations are multiplied with both k and a 𝑧𝑖 weighted due date, earliness and tardiness related costs are penalized.
coefficient inversely determined to the jobs weights. WNOPPT formulas Penalty terms used in this study can be given as follows;
are given as follows: 𝐷𝑖
𝑃 𝐷𝑖 = 8𝑤𝑖
𝑑𝑖1 = 𝑎𝑖 + 𝑝𝑖 + 5𝑘𝑥 𝑜𝑖 480
(3) 𝐸𝑖
𝑑𝑖2 = 𝑎𝑖 + 𝑝𝑖 + 5𝑘𝑥 𝑜𝑖 𝑧𝑖 𝑃 𝐸𝑖 = 𝑤𝑖 (5 + 4 )
480 (8)
RDM (Random-allowance Due-date) adds a value that matches the 𝑇
𝑃 𝑇𝑖 = 𝑤𝑖 (8 + 6 𝑖 )
normal distribution, with an average of 3𝑃𝑎𝑣 , and a standard deviation 480
𝑃𝑎𝑣 , to the arrival time of the job. RDM rule formulas are given as 𝑃𝑖 = 𝑃 𝐷𝑖 + 𝑃 𝐸𝑖 + 𝑃 𝑇𝑖
follows:
Where 𝑤𝑗 is the weight of 𝑗th job. 𝐷 is the due date, 𝐸 is the
𝑑𝑖 = 𝑎𝑖 + 𝑁(3𝑃𝑎𝑣 , 𝑃𝑎𝑣 ) (4) earliness and 𝑇 is the tardiness of job 𝑗th job.𝑃 𝐷𝑗 is penalty for due

4
H.I. Demir and C. Erden Computers & Industrial Engineering 149 (2020) 106799

date, 𝑃 𝐸𝑗 is the penalty for earliness and 𝑃 𝐷𝑗 is the penalty for 4.2. Ant colony system
tardiness of 𝑗th job. 𝑃𝑗 is the total penalty for job 𝑗th job.
Finally, for the objective function, both early completion, late com- Dorigo and Gambardella (1997), Gambardella and Dorigo (1996)
pletion and the duration of the given due-dates have been punished. created the original ant system with three basic improvements. Bal-
Thus, it was tried to ensure that the shop floors should be produced ancing the situation transition rule with the search of new roads and
with the philosophy of JIT. That is why we determine 𝑇 𝑃 which is the the development of existing roads is facilitated. After completing the
total penalty for all the jobs as follows; ant’s tours, pheromone is updated on the road where only the best
round is formed. In addition to this update, local pheromone updates
∑
𝑛
𝑓𝑚𝑖𝑛 = 𝑇 𝑃 = 𝑃𝑖 (9) are available.
𝑗=1 The local pheromone update is implemented every step of the way
by all ants when creating the tour. Each ant follows this rule only on the
4. Ant colony optimization algorithm last path it took. By reducing the number of pheromones on the roads
traveled in iteration, ants in subsequent iteration are encouraged to
Because only the scheduling problem is in the NP-Hard class. Inte- choose other paths and produce different solutions. Thus, the probabil-
grated problems have even a greater solution space. Hence, heuristic ity of many ants producing the same solution is reduced in an iteration.
algorithms should be used to solve the integrated problem. ACO is
a meta-heuristic method developed for solving difficult optimization
𝜏𝑖𝑗 ← (1 − 𝜑)𝜏𝑖𝑗 + 𝜑𝜏0 , ∀(𝑖, 𝑗) ∈ 𝐿 (12)
problems. It is inspired by the collective discharging pheromone and
tracking behavior of ants (Dorigo & Di Caro, 1999). Ants are communi- Here, the evaporation parameter, ranging from 𝜑, [0, 1] shows the
ties that communicate with the chemical traces they leave around them, number of pheromones at the beginning of 𝜏0 . With the local pheromone
producing solutions with the information they obtain from each other. update, pheromones are evaporated from the paths that ants pass
Although ants seem simple when viewed individually, they have a very through, making them less preferred. Thus, ants explore other paths
complex social life as a community. When ants search for food, they and the algorithm does not stagnate. Universal pheromone update is
first discover randomly around their nests. As soon as an ant finds food, performed just like in Eq. (13).
it evaluates the quantity and quality of the resource and takes some of
𝜏𝑖𝑗 ← (1 − 𝑝)𝜏𝑖𝑗 + 𝑝𝛥𝜏𝑖𝑗𝑚𝑏𝑒𝑠𝑡 , ∀(𝑖, 𝑗) ∈ 𝐿𝑚𝑏𝑒𝑠𝑡 (13)
the food it finds in its nest. On the way back, the ant leaves a chemical
trace that affects other ants on the ground. The amount of chemicals Here, the ant that found the best solution up to that time can
left may vary depending on the quantity and quality of food and guides leave traces of pheromones. Since the amount of pheromone dropped is
other ants to find their food source (Blum & Dorigo, 2004). Thanks to determined by taking the weighted average of the previous pheromone
this order, very complex tasks can be done extremely perfectly, which value and the new pheromone value with the help of the 𝑝 parameter,
is in no way possible for a single ant to do. As a result of observing the the attractiveness of this path does not increase excessively and the
behavior of ants, ‘‘ant algorithms’’ models emerged, and these models algorithm is attached to the best solution up to that time.
inspired new algorithms that could be used to solve optimization and
distributed control problems. This trail-and-trail behavior of ants has 4.3. Proposed ACO
inspired ant colony optimization algorithms (Dorigo & Stützle, 2004)
and it has proven itself to solve many NP-Hard problems. With the The structure of the problem discussed consists of discrete and
additions to the ant system, the algorithm has been strengthened and categorical data and the elitist ant colony algorithm has been preferred
it has become one of the successful ACO algorithms. Besides, it has to solve the problem. In other words, only the best ant can increase
been used to solve many scheduling problems (Dorigo & Stützle, 2003; the amount of pheromones in the path. The pheromone amount of the
Merkle et al., 2002; Stützle, 1998; Zhang & Wong, 2016b) with different preferred pathways left by the other than the best ant is reduced in
strategies like MAX–MIN feature, elitist strategy (Zhang & Wong, 2017). each iteration. The adapted version of the problem according to the
ant colony can be described as in Fig. 2.
4.1. Elitist ant colony optimization The path chosen by every ant in the colony represents a solution.
The ant completes its path through three different types of stations.
Elitist ACO was developed by Dorigo et al. (1991) for the first time. The due date assignment rule is determined at the first station. The
This method is called elitist because it is like the best solution used dispatching rule is determined at the second station. In the third type
in genetic algorithms. More pheromones are left in the best solution stations, routes to be assigned to jobs are determined. Therefore, the
(𝑇 𝑚𝑏𝑒𝑠𝑡 ) from the moment the algorithm starts working, increasing the number of stations that the ant must visit depends on the number of
probability of this path being preferred. jobs in the shop floor. The number of stations is two more than the
number of jobs. When the ant completes the path, the quality of the
∑
𝑚
solution is calculated using objective function. Of all ants, ant with the
𝛥𝜏𝑖𝑗𝑘 + 𝑒𝛥𝜏𝑖𝑗𝑚𝑏𝑒𝑠𝑡 , → ∀(𝑖, 𝑗) ∈ 𝐿 (10)
𝑘=1
best quality is determined as the best ant. In the next iteration, the
amount of pheromones on the path of the best ant is increased. So, the
In this equation, the amount of pheromones 𝛥𝜏𝑖𝑗𝑘 that ant 𝑘 leaves pheromone amount on this path is increased. The paths that all ants
on the route is calculated as in the Eq. (10). The 𝛥𝜏𝑖𝑗𝑚𝑏𝑒𝑠𝑡 calculated as go through are reduced by the formula, which reduces the probability
in Eq. (11). It shows the number of 𝑒 elitist ants in the equation. If of paths. Here, the minimum probability probability is updated to
the optimal value of the number of elitist ants is used according to 0.001 whenever probabilities are below than 0 in order not to have 0
the problem, the algorithm can find better solutions with much less probabilities for any station. Eventually all possibilities are updated to
iteration. 1 sum. An instance form of a colony is illustrated in Fig. 3 (see Fig. 3).
{
1
, if (𝑖, 𝑗) is better than 𝑇 𝑚𝑏𝑒𝑠𝑡 There are 10 ants in the colony and each ant must visit a route
𝑐 𝑚𝑏𝑒𝑠𝑡 (11) including station 1, 2 and 3 respectively. The initial colony is created by
0, 𝑒𝑙𝑠𝑒
assigning an equal probability to due date assignment rule, dispatching
Similar to Eq. (10) in this equation, the route of ant 𝑘, which has rule, and routes for jobs. Then the ant with the highest quality solution
found the best solution to date, is left on the paths along the 𝑇 𝑚𝑏𝑒𝑠𝑡 in an in the initial colony is called the best ant. The amount of pheromones
inverse proportion to the 𝐶 𝑚𝑏𝑒𝑠𝑡 , the length of this route. This increases is increased according to the quality of the best ant solution. The
the probability of being chosen for the best path found. amount of pheromones is evaporated in the preferred paths of other

5
H.I. Demir and C. Erden Computers & Industrial Engineering 149 (2020) 106799

Fig. 2. An ant path representation.

Fig. 3. Colony representation.

ants, which is reduced. Thus, improvement is attempted by increasing initialization;

the possibilities of the paths preferred by the ant with the highest 𝑎=𝑘 𝑃 𝑎=𝑘 , 𝑃 𝑎=𝑘, 𝑗=𝑖 are initialized
All parameters such as 𝑃𝑑𝑑𝑟𝑢𝑙𝑒 𝑑𝑠𝑝𝑟𝑢𝑙𝑒 𝑟𝑜𝑢𝑡𝑒𝑠
quality solution in each iteration. These iterations are continued until
randomly.;
the specified number of iterations is reached. Pseudo codes of the
𝑆𝑡𝑒𝑝 ← 0 ;
proposed ACO algorithm is given in the Algorithm 1.
𝑎=𝑏𝑒𝑠𝑡 = big number ;
Notations
while step < 𝑁𝐼 do
• 𝑋𝑑𝑑𝑟𝑢𝑙𝑒 : Due Date Assignment Rule Vector Generate 𝐶𝑠𝑡𝑒𝑝 colony;
• 𝑃𝑑𝑑𝑟𝑢𝑙𝑒
𝑎=𝑘 : Due Date Assignment Probabilities Vector of 𝑘th ant 𝐾 ← 0;
for k ∈ 𝑁𝐶 do
• 𝑆𝑑𝑑𝑟𝑢𝑙𝑒
𝑎=𝑘 : Selected Due Date Assignment Rule of 𝑘th ant
Generate a path for 𝑎𝑘 using probabilities;
• 𝜏𝑑𝑑𝑟𝑢𝑙𝑒
𝑎=𝑘 : Pheromone amount of 𝑘th ant for the selected Due Date
Evaluate the quality score of 𝑎𝑘 ;
Assignment Rule
if 𝑎=𝑘 < 𝑎=𝑏𝑒𝑠𝑡 then
• 𝑋𝑑𝑠𝑝𝑟𝑢𝑙𝑒 : Dispatching Rule Vector 𝑎𝑏𝑒𝑠𝑡 = 𝑎𝑘 ;
• 𝑃𝑑𝑠𝑝𝑟𝑢𝑙𝑒
𝑎=𝑘 : Dispatching Rule Probabilities Vector of 𝑘th ant 𝑎=𝑏𝑒𝑠𝑡 = 𝑎=𝑘 ;
• 𝑆𝑑𝑠𝑝𝑟𝑢𝑙𝑒
𝑎=𝑘 : Selected Dispatching Rule of 𝑘th ant 𝑘+ = 1;
• 𝜏𝑑𝑠𝑝𝑟𝑢𝑙𝑒
𝑎=𝑘 : Pheromone amount of 𝑘th ant for the selected Dispatching else
Rule 𝑘+ = 1;
𝑗=𝑖
• 𝑋𝑟𝑜𝑢𝑡𝑒 : Route Station Vector of 𝑖th job end
𝑎=𝑘, 𝑗=𝑖
• 𝑃𝑟𝑜𝑢𝑡𝑒 : Route Probabilities Vector of 𝑘th ant and 𝑖th job end
𝑎=𝑘, 𝑗=𝑖
• 𝑆𝑟𝑜𝑢𝑡𝑒 : Selected Route Station of 𝑘th ant and 𝑖th job Update probabilities ;
𝑎=𝑘, 𝑗=𝑖 Increase the pheromone quantities (probabilities) of the stations
• 𝜏𝑟𝑜𝑢𝑡𝑒 : Pheromone amount of 𝑘th ant for the selected route of
where the best ant passes.;
𝑖th job
Reduce pheromones (probabilities) at stations where all ants pass;
• 𝑅𝑎=𝑘 : Route of 𝑘th ant (solution)
𝑠𝑡𝑒𝑝+ = 1;
• 𝐶𝑘 : Colony of 𝑘th iteration
end
• 𝑎𝑘 : 𝑘th ant
Algorithm 1: Pseudo code of proposed ACO
• 𝑎𝑏𝑒𝑠𝑡 : Best ant
• 𝑁𝐶: Number of ants in a colony
• 𝑁𝐼: Number of iterations
• 𝑁𝐷𝐷: Number of due date assignment rule given on ACO and a structure in which ACO performed better than GA
• 𝑁𝐷𝑆𝑃 : Number of dispatching rules was shared. Although GA yielded better results in some shop floors,
• 𝑁𝑅: Number of routes of each job it was observed that ACO achieved more best results than GA. The
• 𝑎=𝑘 : Fitness value of 𝑘th ant most important shop floors for the study are the 7th and 8th shop
floors. Because there are more jobs and machines in these shop floors.
In addition, the experiments in the 8th shop floor were tried with 5
different number generators and the results were averaged. According
to the results obtained, the average performance of ACO was superior
5. Experimental results to GA. This demonstrates the strong structure of ACO in this particular
problem. The Figs. 4–11 graphs show improvements in each iteration.
We have developed 8 different size shop floors for the comparison As the number of work and machines in the shop floor increased, the
of the GA and ACO algorithm performances. Since both algorithms number of iterations was reduced. Because of the large number of
were population based, the performances of the two algorithms were iterations, there are difficulties in working in large shop floor.
compared with the same number of iterations. GA is conserved by The Figs. 12–19 graphs below show the best, average and worst
its form in (Erden et al., 2019). In this study, more emphasis was values obtained for each of the two algorithms. The best result for GA

6
H.I. Demir and C. Erden Computers & Industrial Engineering 149 (2020) 106799

Fig. 4. Shop Floor 1. Fig. 7. Shop Floor 4.

Fig. 5. Shop Floor 2. Fig. 8. Shop Floor 5.

Fig. 6. Shop Floor 3. Fig. 9. Shop Floor 6.

is the performance value of the best chromosome in the last population Table 2 shows the assignment results of the all jobs. According to
and the best ant for ACO is the best value. The average value was the table, information about the time of arrival of the work, time of
taken as the mean of the solutions of all chromosomes in the last departure, assignment of due dates, early completion, late completion
population in GA and as the average of the performances in the last and penalties are given for each job. Accordingly, the robustness of the
colony in ACO. The worst value was taken as the worst value in the assigned delivery dates can be observed.
last population in both algorithms. Thanks to these graphs, average and
worst performance solutions were evaluated along with the best values. 6. Discussion of results
Fig. 20 shows the Gantt scheme for assignments for the works in the
first shop floor. Since the Gantt charts in larger shop floors are difficult In this study one of the popular metaheuristics ACO algorithm is
to read, the Gantt chart is only provided for the small shop floor. In the tested and compared with another popular metaheuristics GA and with
Gantt chart, information about the arrival of the works is given with the random solutions. In order the results to be more reliable numerous
help of arrows. shop floor configurations are tested and biggest shop floor is tested

7
H.I. Demir and C. Erden Computers & Industrial Engineering 149 (2020) 106799

Fig. 13. Comparisons for Shop Floor 2.

Fig. 10. Shop Floor 7.

Fig. 14. Comparisons for Shop Floor 3.

Fig. 11. Shop Floor 8.

Fig. 15. Comparisons for Shop Floor 4.

Fig. 12. Comparisons for Shop Floor 1.

for five different random seed numbers five times and average of these
results were taken for interpretation.
Table 3 shows the experimental results of the study. If first genera-
tion of ants is considered as the random solutions of the problem and
if the average of first generation is taken as ordinary solution quality
and compared with the results of ACO and GA, it can be seen how
metaheuristics provide dramatic improvement. For instance for the 7th
shop floor, average of first generation is 1151 and penalty function is
reduced to 897 if GA is used and reduced to 834 if ACO metaheuristics
is applied. Similarly if 8th shop floor is considered, this shop floor is
specifically tested for five random seeds five times and average of these
values are taken for interpretation. Average of first generations is 1354 Fig. 16. Comparisons for Shop Floor 5.

8
H.I. Demir and C. Erden Computers & Industrial Engineering 149 (2020) 106799

Table 2
Schedule results of shop floor 1.
Job 𝑤𝑖 𝑎𝑖 𝑑𝑖 𝐷𝑖 𝐸𝑖 𝑇𝑖 PE PT PD PT
0 1.0 71.0 198.0 214.0 16.0 0.0 5.1 0.0 2.4 7.5
1 0.6 76.0 192.0 252.8 60.7 0.0 3.6 0.0 1.9 5.6
2 0.3 83.0 283.0 286.8 3.8 0.0 1.7 0.0 1.1 2.8
3 0.3 88.0 259.0 268.2 9.2 0.0 1.7 0.0 1.0 2.7
4 0.6 98.0 260.0 290.5 30.5 0.0 3.5 0.0 2.1 5.6
5 1.0 124.0 284.0 315.4 31.4 0.0 5.3 0.0 3.2 8.5
6 0.3 133.0 404.0 321.0 0.0 83.0 0.0 2.3 1.0 3.4
7 0.6 188.0 337.0 372.7 35.6 0.0 3.5 0.0 2.0 5.5
8 0.6 196.0 335.0 345.8 10.7 0.0 3.4 0.0 1.7 5.0
9 0.6 218.0 365.0 388.0 23.0 0.0 3.4 0.0 1.9 5.3
10 0.3 288.0 477.0 476.0 0.0 1.0 0.0 2.0 1.0 3.0
11 0.6 353.0 508.0 512.9 4.9 0.0 3.3 0.0 1.8 5.1
12 0.7 367.0 625.0 560.7 0.0 64.3 0.0 4.5 2.1 6.6
13 1.0 371.0 491.0 529.8 38.7 0.0 5.3 0.0 2.7 8.0
14 0.6 377.0 663.0 565.0 0.0 98.0 0.0 4.8 2.1 6.8
Fig. 17. Comparisons for Shop Floor 6.
15 1.0 393.0 597.0 611.4 14.4 0.0 5.1 0.0 3.6 8.8
16 0.6 408.0 735.0 573.5 0.0 161.5 0.0 5.3 1.8 7.1
17 0.6 416.0 743.0 599.5 0.0 143.5 0.0 5.1 2.0 7.2
18 1.0 429.0 561.0 585.5 24.5 0.0 5.2 0.0 2.6 7.8
19 0.6 458.0 763.0 612.3 0.0 150.7 0.0 5.2 1.7 6.9
20 1.0 467.0 640.0 640.4 0.4 0.0 5.0 0.0 2.9 7.9
21 0.3 468.0 840.0 688.7 0.0 151.3 0.0 2.6 1.2 3.8
22 1.0 492.0 677.0 680.0 3.0 0.0 5.0 0.0 3.1 8.2
23 0.6 496.0 800.0 715.5 0.0 84.5 0.0 4.7 2.4 7.1
24 1.0 537.0 710.0 744.2 34.1 0.0 5.3 0.0 3.5 8.7

Where 𝑤𝑖 weight of job 𝑖, 𝑎𝑖 arrival time of job 𝑖, 𝑑𝑖 completion time of job 𝑖, 𝐷𝑖

given due-date time of job 𝑖, 𝐸𝑖 earliness of job 𝑖, 𝑇𝑖 tardiness of job 𝑖, PE penalty for
earliness, PT penalty for tardiness, PD penalty for due-date, PT total penalty.

and this penalty value is reduced to 1098 in average in case of GA and

984 in average in case of ACO algorithms.
At the five shop floors which are SF1 (Shop Floor 1), SF3, SF4, SF7
Fig. 18. Comparisons for Shop Floor 7. and SF8, ACO gave much better results. Especially SF8 is tested five
times and averages were taken and ACO is found much better. GA is
found slightly better at the other shop floors, SF2, SF5 and SF6. Even in
these shop floors ACO gave good results and GA was only found slightly
better than ACO. If averages of final iterations are considered for eight
shop floors, ACO outperformed other solutions and six times ACO gave
better averages of the shop floors compared to the GA.

7. Conclusion

As a conclusion, in this study, DIPPSDDA problem which is a

novel problem in literature is discussed. More specifically, DIPPSDDA
problem has dynamic scheduling integrated with process planning and
due-date assignment function. As a dynamic event, it is examined that
the behavior of the jobs change dynamically. Highlights of the paper
can be given as follows. (i) Process planning dynamic scheduling and
Fig. 19. Comparisons for Shop Floor 8. the due-date assignment functions are integrated using GA and ACO
algorithms. (ii) Objective function is to minimize earliness, tardiness
and due dates of all jobs. (iii) 9 due date assignment and 11 dispatching
rules have been integrated. The experimental results show that the

Table 3
Experimental results.
SF1 SF2 SF3 SF4
Best Avg Worst Best Avg Worst Best Avg Worst Best Avg Worst
First generation 184.1 202.8 227.0 348.0 370.7 417.5 385.3 432.2 464.8 525.8 610.6 666.7
GA 175.8 176.4 188.4 304.6 308.8 349.1 377.5 378.5 386.4 486.4 503.8 533.0
ACO 164.6 171.0 182.1 305.9 310.9 322.4 354.3 366.4 394.8 481.9 485.6 492.3
SF5 SF6 SF7 SF8
Best Avg Worst Best Avg Worst Best Avg Worst Best Avg Worst
First generation 829.4 869.6 1055.5 701.9 801.2 918.8 910.8 1151.2 1363.0 1101.5 1354.5 1817.3
GA 636.0 693.3 838.1 690.7 692.7 701.9 897.1 898.5 910.8 1098.3 1099.8 1108.6
ACO 643.7 654.5 661.4 702.7 707.3 918.7 834.4 838.8 844.5 984.5 994.9 1297.7

9
H.I. Demir and C. Erden Computers & Industrial Engineering 149 (2020) 106799

Fig. 20. Gantt chart for shop floor 1.

proposed ACO can solve the DIPPSDDA problems and can provide a Baker, K. R., & Bertrand, J. (1981). An investigation of due-date assignment rules with
probabilistic computation of the algorithm. constrained tightness. Journal of Operations Management, 1(3), 109–120.
Baykasoğlu, A., Göçken, M., & Unutmaz, Z. D. (2008). New approaches to due date
The future studies can be listed below:
assignment in job shops. European Journal of Operational Research, 187(1), 31–45.
• Similar problem can be solved with current meta-heuristic al- Baykasoğlu, A., & Özbakır, L. (2010). Analyzing the effect of dispatching rules on
the scheduling performance through grammar based flexible scheduling system.
gorithms such as Artificial Bee Colony, Cuckoo Search Algo-
International Journal of Production Economics, 124(2), 369–381.
rithm, Salp Swarm Algorithm, Firefly Algorithm, Wasp Swarm Baykasoğlu, A., & Özbakır, L. (2009). A grammatical optimization approach for
Algorithm, Hunting Search etc. and their comparisons can be integrated process planning and scheduling. Journal of Intelligent Manufacturing,
given. 20(2), 211–221.
• Other dynamic events such as the breakdowns of the machines Birman, M., & Mosheiov, G. (2004). A note on a due-date assignment on a two-machine
flow-shop. Computers & Operations Research, 31(3), 473–480.
and the cancellation of the jobs in the scheduling might be
Blum, C., & Dorigo, M. (2004). The hyper-cube framework for ant colony optimization.
included in the problem and new studies might be carried out. IEEE Transactions on Systems, Man and Cybernetics, Part B, 34(2), 1161–1172.
• A multi-purpose programming model could be developed taking Chang, F.-C. R. (1996). A study of due-date assignment rules with constrained tightness
into account the makespan value. in a dynamic job shop. Computers & industrial engineering, 31(1–2), 205–208.
Cheng, T. (1987). Optimal TWK-power due-date determination and sequencing.
International Journal of Systems Science, 18(1), 1–7.
CRediT authorship contribution statement
Cheng, T., & Gordon, V. (1994). Optimal assignment of due-dates for preemptive
single-machine scheduling. Mathematical and Computer Modelling, 20(2), 33–40.
Halil Ibrahim Demir: Writing - review & editing, Supervision, Con- Cheng, T. C., & Gupta, M. C. (1989). Survey of scheduling research involving due date
ceptualization, Methodology, Data curation, Writing - original draft. determination decisions. European Journal of Operational Research, 38(2), 156–166.
Caner Erden: Methodology, Data curation, Visualization, Investigation, Chiang, T., & Fu, L. (2007). Using dispatching rules for job shop scheduling
with due-date based objectives. International Journal of Productions Research, 45,
Software, Validation, Writing - original draft.
3245–3262.
Chryssolouris, G., Chan, S., & Suh, N. (1985). An integrated approach to process
Appendix A. Supplementary files planning and scheduling. CIRP Annals, 34(1), 413–417.
Demir, H. I., Erden, C., Uygun, O., & Kokcam, A. H. (2017). Solving process planning
The data used for this study is given as supplementary files and in WATC scheduling and WPPW due date assignment concurrently using modified
discrete PSO and simulated annealing algorithms. (p. 472).
in this github link. Problem data is given as follows:
Demir, H. I., & Taskin, H. (2005). Integrated process planning, scheduling and due-date
assignment (Ph.D. thesis), Sakarya: Sakarya University.
• The data belonging to the shop floors produced and saved as a
Dominic, P. D. D., Kaliyamoorthy, S., & Kumar, M. S. (2003). Efficient dispatching rules
‘.txt’ file. for dynamic job shop scheduling. International Journal of Advanced Manufacturing
• Arrival time ‘arrivals shop floor number.txt’. Technology, -1(1), 1.
• the machine sequences ‘machine numbers shop floor number.txt’ Dorigo, M., & Di Caro, G. (1999). Ant colony optimization: a new meta-heuristic.
• processing times ‘operation durations shop floor number.txt’. In Proceedings of the 1999 congress on evolutionary computation-CEC99 (Cat. No.
99TH8406)): Vol. 2 (pp. 1470–1477). IEEE.
• weights ‘weights-shop floor number.txt’.
Dorigo, M., & Gambardella, L. M. (1997). Ant colony system: a cooperative learning
Supplementary material related to this article can be found online approach to the traveling salesman problem. IEEE Transactions on Evolutionary
Computation, 1(1), 53–66.
at https://doi.org/10.1016/j.cie.2020.106799.
Dorigo, M., Maniezzo, V., & Colorni, A. (1991). Positive feedback as a search strategy.
Milan, Italy: Dipartimento di Elettronica, Politecnico di Milano.
References Dorigo, M., & Stützle, T. (2003). The ant colony optimization metaheuristic: Algorithms,
applications, and advances. In Handbook of metaheuristics (pp. 250–285). Springer.
Aldakhilallah, K. A., & Ramesh, R. (1999). Computer-integrated process planning and Dorigo, M., & Stützle, T. (2004). Ant colony optimization. MIT Press.
scheduling (CIPPS): Intelligent support for product design, process planning and Durasević, M., & Jakobović, D. (2018). A survey of dispatching rules for the dynamic
control. International Journal of Productions Research, 37(3), 481–500. unrelated machines environment. Expert Systems with Applications, 113, 555–569.
Amin-Naseri, M., & Afshari, A. J. (2012). A hybrid genetic algorithm for integrated Erden, C., Demir, H. I., & Kökçam, A. H. (2019). Solving integrated process planning,
process planning and scheduling problem with precedence constraints. International dynamic scheduling, and due date assignment using metaheuristic algorithms.
Journal of Advanced Manufacturing Technology, 59(1–4), 273–287. Mathematical Problems in Engineering, 2019.

10
H.I. Demir and C. Erden Computers & Industrial Engineering 149 (2020) 106799

Gambardella, L., & Dorigo, M. (1996). Solving symmetric and asymmetric TSPs by ant Min, L., & Cheng, W. (2006). Genetic algorithms for the optimal common due
colonies. In Proceedings of IEEE international conference on evolutionary computation date assignment and the optimal scheduling policy in parallel machine earli-
(pp. 622–627). IEEE. ness/tardiness scheduling problems. Robotics and Computer-Integrated Manufacturing,
Gordon, V., Proth, J. M., & Chu, C. (2002). Due date assignment and scheduling: SLK, 22(4), 279–287.
TWK and other due date assignment models. Production Planning and Control, 13(2), Morad, N., & Zalzala, A. M. S. (1999). Genetic algorithms in integrated process planning
117–132. and scheduling. Journal of Intelligent Manufacturing, 10(2), 169–179.
Gordon, V., Proth, J. M., & Chu, C. (2002). A survey of the state-of-the-art of common Nasr, N., & Elsayed, E. A. (1990). Job shop scheduling with alternative machines. The
due date assignment and scheduling research. European Journal of Operational International Journal of Production Research, 28(9), 1595–1609.
Research, 139(1), 1–25, PMID: 2017. Park, B. J., & Choi, H. R. (2006). A genetic algorithm for integration of process planning
Gordon, V. S., & Strusevich, V. A. (2009). Single machine scheduling and due date and scheduling in a job shop. In Lecture notes in computer science (pp. 647–657).
assignment with positionally dependent processing times. European Journal of Springer Berlin Heidelberg,
Operational Research, 198(1), 57–62. Park, B. J., & Choi, H. R. (2006). A genetic algorithm for integration of process planning
Gordon, V., Strusevich, V., & Dolgui, A. (2012). Scheduling with due date assignment and scheduling in a job shop. In Lecture notes in computer science (pp. 647–657).
under special conditions on job processing. Journal of Scheduling, 15(4), 447–456. Springer Berlin Heidelberg.
Guo, Y., Li, W. D., Mileham, A. R., & Owen, G. W. (2009). Applications of particle Petrović, M., Vuković, N., Mitić, M., & Miljković, Z. (2016). Integration of process
swarm optimisation in integrated process planning and scheduling. Robotics and planning and scheduling using chaotic particle swarm optimization algorithm.
Computer-Integrated Manufacturing, 25(2), 280–288. Expert Systems with Applications, 64, 569–588.
Guo, Y. W., Li, W. D., Mileham, A. R., & Owen, G. W. (2009). Optimisation of integrated Phanden, R. K., Jain, A., & Davim, J. P. (Eds.), (2020). Integration of process planning
process planning and scheduling using a particle swarm optimisation approach. and scheduling: approaches and algorithms. In Science, technology, and management,
International Journal of Productions Research, 47(14), 3775–3796. Boca Raton: CRC Press/Taylor & Francis Group.
Hutchison, J., Leong, K., Snyder, D., & Ward, P. (1991). Scheduling approaches Phanden, R. K., Jain, A., & Verma, R. (2011). Integration of process planning and
for random job shop flexible manufacturing systems. The International Journal of scheduling: a state-of-the-art review. International Journal of Computer Integrated
Production Research, 29(5), 1053–1067. Manufacturing, 24(6), 517–534.
Janiak, A., Janiak, W. A., Krysiak, T., & Kwiatkowski, T. (2015). A survey on scheduling Pinedo, M. (2012). Scheduling. 29, Springer.
problems with due windows. European Journal of Operational Research, 242(2), Sen, T., & Gupta, S. K. (1984). A state-of-art survey of static scheduling research
347–357. involving due dates. Omega, 12(1), 63–76.
Jin, L., Zhang, C., & Shao, X. (2015). An effective hybrid honey bee mating optimization Sharma, P., & Jain, A. (2014). Analysis of dispatching rules in a stochastic dynamic
algorithm for integrated process planning and scheduling problems. International job shop manufacturing system with sequence-dependent setup times. Frontiers of
Journal of Advanced Manufacturing Technology, 80(5–8), 1253–1264. Mechanical Engineering, 9(4), 380–389.
Kaban, A. K., Othman, Z., & Rohmah, D. S. (2012). Comparison of dispatching rules in Shen, W., Wang, L., & Hao, Q. (2006). Agent-based distributed manufacturing process
job-shop scheduling problem using simulation: a case study. International Journal planning and scheduling: a state-of-the-art survey. IEEE Transactions on Systems,
of Simulation Modelling, 11(3), 129–140. Man and Cybernetics Part C: Applications and Reviews, 36(4), 563–577.
Shukla, S. K., Tiwari, M. K., & Son, Y. J. (2008). Bidding-based multi-agent system
Keddari, N., Mebarki, N., Shahzad, A., & Sari, Z. (2018). Solving an integration
for integrated process planning and scheduling: A data-mining and hybrid tabu-
process planning and scheduling in a flexible job shop using a hybrid approach. In
SA algorithm-oriented approach. International Journal of Advanced Manufacturing
Computational intelligence and its applications (pp. 387–398). Springer International
Technology, 38(1–2), 163–175.
Publishing.
Stützle, T. (1998). An ant approach to the flow shop problem. In Proceedings of the 6th
Khoshnevis, B., & Chen, Q. M. (1991). Integration of process planning and scheduling
European congress on intelligent techniques and soft computing: Vol. 3 (pp. 1560–1564).
functions. Journal of Intelligent Manufacturing, 2(3), 165–175.
Sweeney, K. D., Sweeney, D. C., & Campbell, J. F. (2019). The performance of priority
Kumar, M., & Rajotia, S. (2006). Integration of process planning and scheduling in a
dispatching rules in a complex job shop: A study on the Upper Mississippi River.
job shop environment. International Journal of Advanced Manufacturing Technology,
International Journal of Production Economics, 216, 154–172.
28(1–2), 109–116.
Tan, W., & Khoshnevis, B. (2000). Integration of process planning and scheduling-a
Lauff, V., & Werner, F. (2004). Scheduling with common due date, earliness and
review. Journal of Intelligent Manufacturing, 11(1), 51–63.
tardiness penalties for multimachine problems: A survey. Mathematical and Computer
Tuong, N. H., & Soukhal, A. (2010). Due dates assignment and JIT scheduling with
Modelling, 40(5–6), 637–655.
equal-size jobs. European Journal of Operational Research, 205(2), 280–289.
Lawrence, S. R. (1994). Negotiating due-dates between customers and producers.
Uslu, M. F., Uslu, S., & Bulut, F. (2018). An adaptive hybrid approach: Combining
International Journal of Production Economics, 37(1), 127–138, PMID: 603.
genetic algorithm and ant colony optimization for integrated process planning and
Lee, Y. H., & Pinedo, M. (1997). Scheduling jobs on parallel machines with
scheduling. Applied Computing and Informatics.
sequence-dependent setup times. European Journal of Operational Research, 100(3),
Vinod, V., & Sridharan, R. (2011). Simulation modeling and analysis of due-date as-
464–474.
signment methods and scheduling decision rules in a dynamic job shop production
Leung, C., Wong, T., Mak, K., & Fung, R. (2010). Integrated process planning and
system. International Journal of Production Economics, 129(1), 127–146.
scheduling by an agent-based ant colony optimization. Computers & Industrial
Wang, J., Fan, X., Zhang, C., & Wan, S. (2014). A graph-based ant colony optimization
Engineering, 59(1), 166–180.
approach for integrated process planning and scheduling. Chinese Journal of
Li, C. L. (2015). Improved algorithms for single-machine common due window Chemical Engineering, 22(7), 748–753.
assignment and scheduling with batch deliveries. Theoretical Computer Science, Wang, L., Zhou, G., Xu, Y., & Liu, M. (2013). A hybrid artificial bee colony algorithm for
570(C), 30–39. the fuzzy flexible job-shop scheduling problem. International Journal of Productions
Li, X., Gao, L., Zhang, C., & Shao, X. (2000). A review on integrated process planning Research, 51(12), 3593–3608.
and scheduling. Journal of Intelligent Manufacturing, 5(2), 51–63. Wong, T. N., Leung, C. W., Mak, K. L., & Fung, R. Y. K. (2006). An agent-based
Li, W. D., & McMahon, C. A. (2007). A simulated annealing-based optimization negotiation approach to integrate process planning and scheduling. International
approach for integrated process planning and scheduling. International Journal of Journal of Productions Research, 44(7), 1331–1351.
Computer Integrated Manufacturing, 20(1), 80–95. Xia, H., Li, X., & Gao, L. (2016). A hybrid genetic algorithm with variable neighborhood
Lihong, Q., & Shengping, L. (2012). An improved genetic algorithm for integrated search for dynamic integrated process planning and scheduling. Computers &
process planning and scheduling. International Journal of Advanced Manufacturing Industrial Engineering, 102, 99–112.
Technology, 58(5–8), 727–740. Yu, M., Yang, B., & Chen, Y. (2018). Dynamic integration of process planning and
Lin, S.-W., Chou, S.-Y., & Chen, S.-C. (2006). Meta-heuristic approaches for minimizing scheduling using a discrete particle swarm optimization algorithm. Advances in
total earliness and tardiness penalties of single-machine scheduling with a common Production Engineering & Management, 13(3), 279–296.
due date. Journal of Heuristics, 13(2), 151–165. Yu, M., Zhang, Y., Chen, K., & Zhang, D. (2015). Integration of process planning and
Lin, S.-C., Goodman, E. D., & Punch III, W. F. (1997). A genetic algorithm approach scheduling using a hybrid GA/PSO algorithm. International Journal of Advanced
to dynamic job shop scheduling problem. In ICGA (pp. 481–488). Manufacturing Technology, 78(1–4), 583–592.
Liu, X., Ni, Z., & Qiu, X. (2015). Application of ant colony optimization algorithm Yusof, Y., & Latif, K. (2014). Survey on computer-aided process planning. International
in integrated process planning and scheduling. International Journal of Advanced Journal of Advanced Manufacturing Technology, 75(1–4), 77–89.
Manufacturing Technology, 84(1–4), 393–404. Zhang, L., Gao, L., & Li, X. (2013). A hybrid genetic algorithm and tabu search for
Madureira, A., Ramos, C., & Silva, S. d. C. (2003). Using genetic algorithms for dynamic a multi-objective dynamic job shop scheduling problem. International Journal of
scheduling. In I14th annual production and operations management society conference. Productions Research, 51(12), 3516–3531.
Meenakshi Sundaram, R., & Fu, S.-s. (1988). Process planning and scheduling — Zhang, W., Gen, M., & Jo, J. (2014). Hybrid sampling strategy-based multiobjective
A method of integration for productivity improvement. Computers & Industrial evolutionary algorithm for process planning and scheduling problem. Journal of
Engineering, 15(1–4), 296–301. Intelligent Manufacturing, 25(5), 881–897.
Merkle, D., Middendorf, M., & Schmeck, H. (2002). Ant colony optimization Zhang, S., & Wong, T. N. (2014). Integrated process planning and scheduling: an
for resource-constrained project scheduling. IEEE Transactions on Evolutionary enhanced ant colony optimization heuristic with parameter tuning. Journal of
Computation, 6(4), 333–346. Intelligent Manufacturing, 29(3), 585–601.

11
H.I. Demir and C. Erden Computers & Industrial Engineering 149 (2020) 106799

Zhang, L., & Wong, T. (2016). Solving integrated process planning and scheduling Zhao, C., Hsu, C.-J., Lin, W.-C., Liu, S.-C., & Yu, P.-W. (2018). Due date assignment
problem with constructive meta-heuristics. Information Sciences, 340–341, 1–16. and scheduling with time and positional dependent effects. Journal of Information
Zhang, S., & Wong, T. N. (2016). Studying the impact of sequence-dependent set- and Optimization Sciences, 39(8), 1613–1626.
up times in integrated process planning and scheduling with E-ACO heuristic. Zhao, F., Zhu, A., Ren, Z., & Yang, Y. (2006). Integration of process planning
International Journal of Productions Research, 54(16), 4815–4838. and production scheduling based on a hybrid PSO and SA algorithm. In 2006
Zhang, S., & Wong, T. N. (2017). Flexible job-shop scheduling/rescheduling in dynamic international conference on mechatronics and automation (pp. 2290–2295). IEEE.
environment: a hybrid MAS/ACO approach. International Journal of Productions
Research, 55(11), 3173–3196.

12