Available online at www.sciencedirect.

com

ScienceDirect
Procedia Manufacturing 11 (2017) 2123 – 2130

27th International Conference on Flexible Automation and Intelligent Manufacturing, FAIM2017,

27-30 June 2017, Modena, Italy

Supporting maintenance scheduling: a case study

Patrícia Senraª, Isabel Lopesª*, José A Oliveiraª
ªALGORITMI Research Centre, Department of Production and Systems, University of Minho, Guimarães 4800-058, Portugal

Abstract

This work intends to incorporate in a computerized maintenance management system (CMMS) a scheduling support tool that,
given a set of preventive maintenance tasks and the availability of equipment and technicians, returns a scheduling solution for a
given time period. The tool is based on an algorithm that aims to minimize the total tardiness. The algorithm sorts, initially, the
maintenance tasks by a critically index equipment and by tasks due date. Then the tasks are assigned checking previously if
necessary spare parts are available. An application example is provided to show the proposed algorithm effectiveness.

©©2017
2017TheTheAuthors. Published
Authors. by Elsevier
Published B.V.B.V.
by Elsevier This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the scientific committee of the 27th International Conference on Flexible Automation and
Peer-review under responsibility of the scientific committee of the 27th International Conference on Flexible Automation and
IntelligentManufacturing
Intelligent Manufacturing.

Keywords: Scheduling; maintenance scheduling; preventive maintenace; heuristics, optimization

1. Introduction

Preventive maintenance (PM) is applied at predetermined intervals or according to certain criteria with the
purpose to reduce the probability of failure or degradation [1]. PM scheduling aims to establish “when” and by
“who” each planned task will be processed and its main concern is to find the best sequence of maintenance tasks
for the manufacturing process in each period over a planning horizon [2] so that the company’s needs are satisfied.
Each preventive maintenance task to be scheduled usually involves technicians, equipment and, frequently, spare
parts.
This paper reports part of an ongoing project aimed to improve the computerized maintenance management

* Corresponding author. Tel.:+351 253604762; fax: +351 253604741.

E-mail address: ilopes@dps.uminho.pt

2351-9789 © 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the scientific committee of the 27th International Conference on Flexible Automation and Intelligent Manufacturing
doi:10.1016/j.promfg.2017.07.342
2124 Patrícia Senra et al. / Procedia Manufacturing 11 (2017) 2123 – 2130

system (CMMS) of an automotive company. In the company, the PM scheduling is performed “manually” by a
planner, supported by the CMMS which roles as an information system and solution editor. Therefore, nowadays the
maintenance scheduling is a time-consuming activity due to the high number of planned tasks to be considered and
the constraints to attend: the technician’s availability and their skills; the spare part stock; and, mainly, the downtime
of production lines that defines the equipment availability. Consequently, the current scheduling method evidences
lack of effectiveness and efficiency.
In the automotive company that works 24 hours a day, the planner elaborates a weekly scheduling concerning the
production lines schedule. However, daily, the planner reschedules the tasks to use the unplanned downtime of
production lines, since the automotive company does not allocate time exclusively for preventive maintenance,
except on the weekends. The daily changes in production plan compromise the application of the maintenance
scheduling weekly plan. The tasks allocation process does not follow any method or formal dispatching rule. As an
input on the CMMS, there are around 6000 preventive maintenance tasks that need to be scheduled. Besides regular
tasks, this input includes tasks with predefined time schedule (for instance a compulsory task to comply with the
law) and autonomous maintenance tasks. As output, the algorithm must provide a schedule, which, taking as
example one week, establishes the assignment of approximately 600 maintenance tasks to technicians.In the
manufacturing area, the different production lines have different utilization rates, where some of them have
significantly higher utilization rates than others. Therefore, equipment which is part of the production line with
higher utilization rates should be given priority over other equipment since line stops are much less frequent.
The maintenance scheduling requires a method to allocate the maintenance tasks to the limited resources that
takes into account the objective function and a set of constraints. The objective function must meet the main goal of
the company and the method should be appropriate to solve the problem, achieving a good solution in a short time
[3] since the scheduling method might be used several times a day, and the number of task is very large.
Thus, this paper outlines the development of a scheduling support tool based on an algorithm that aims to
minimize the total tardiness and an objective function defined by the company. The paper is organized in five
sections. Section 2 shows a literature review about scheduling, Section 3 presents the algorithm. Section 4 presents
an application example of the algorithm and, finally, Section 5 shows the main conclusions.

2. Literature Review

According to Noemi M. Paz [4], the maintenance scheduling problem can be modeled as a job shop scheduling
problem where:
• a job in the job shop problem is the maintenance work order to process;
• an equipment in the job shop problem is a maintenance worker [4].
In this section the adequate objective functions for maintenance tasks scheduling and scheduling approaches are
reviewed.

2.1 Objective Function

According to Pinedo [3], due dates, deadlines and periodicity are important parameters on preventive
maintenance tasks. These parameters should be respected to guarantee a proper functioning of production lines. The
objective functions most related with maintenance task delays and due dates are the minimization of maximum
lateness (Lmax), minimization of number of tardy tasks (Uj) and minimization of total tardiness (ΣTj) [3,5].
The minimization of maximum lateness is equivalent to minimizing the worst case of the schedule and it is
expressed by Lmax = max (Cj – dj), where Cj is the completion time of task j and dj is the due date of task j (Fig. 1
(a)). The minimization of number of tardy tasks (Uj) is another objective function which represents the number of
tasks that were done after due date. However, minimizing the number of tardy tasks can result in a schedule with
high delay. The total tardiness measure is another performance measure related with the delay expressed by ΣTj = Σ
max {Cj – dj; 0} (Fig. 1(b)). It considers the sum of tardiness of all scheduled tasks. However, if it is intended,
different tasks may carry different priority weights, where the most critical tasks have the higher weights (wj). Apart
from the tardiness penalization, the earliness completion (Ej) of tasks can be also penalized. The representation of
cost functions that consider earliness and tardiness penalties (Ej + Tj) is presented in Fig. 1 (c).
 Patrícia Senra et al. / Procedia Manufacturing 11 (2017) 2123 – 2130 2125

a b c

Fig. 1 -Graphic of due date related objectives: (a) the lateness Lj of job j; (b) the tardiness Tj of job j and (c) cost function in practice [4].

2.2 Scheduling approaches

Ebrahimipour et al. [2] proposes a scheduling model of preventive maintenance tasks in a series-parallel
manufacturing system composed of multiple production lines. The proposed model considers reliability of the
production line, maintenance costs, failure and downtime of system as multiple objectives. In this approach,
different thresholds for available manpower, spare part inventory and periods under maintenance were applied. The
authors consider two types of maintenance activities: adjustment and replacement. A mixed-integer nonlinear multi-
objective model was proposed for this problem. Duffua and Al- Sultan [6] present one of the first mathematical
formulation of the stochastic programming for scheduling maintenance personnel. It incorporates deterministic and
stochastic contributions. Samhouri [7] presents an intelligent method of how to decide whether a particular item
requires opportunistic maintenance or not and if so, how cost-effective this opportunity-based maintenance will be
when compared to a probable future grounding. Opportunistic Maintenance is a systematic method of collecting,
investigating, preplanning and publishing a set of proposed maintenance tasks and acting on them when there is an
unscheduled failure or a repair opportunity. The objective function of the method focuses on the total cost of
maintenance that includes the remaining life cost, down time cost, unplanned down time cost, risk cost, Hazard
function, secondary damage cost due to failure and unit price per finished product or service unit. Manzini et al. [8]
present a resource-constraints mixed integer programming model that minimizes the total cost function which
includes spare parts contributions, the cost of execution of the preventive actions, the cost of additional repair
activity in case of unplanned failure, the cost of personnel of the producer and/or the maintenance provider. This
model can be applied to a short-term plan (e.g. weekly) of preventive actions and resources (spare parts, equipment
and personnel) necessary to conduct maintenance tasks.
Each approach presented in this section aims to address specific cases, once each case requires different
specifications. However, all the presented approaches have the same concern, scheduling the maintenance tasks
required to keep items in a specified condition, guaranteeing a proper performance of equipment. These approaches
were tested with small size instances. Though, our real world instances are very large, which demands a tailored
constructive algorithm to provide effective solutions in real time.
In this paper an initial approach of a scheduling algorithm will be presented and discussed based on a small
instance. In order to validate and to compare the results obtained with the constructive algorithm, the small instance
was also solved by a mixed-integer linear programming (MILP). Through the application of a MILP model, it is
possible to obtain optimal plans considering small instances.

3. The Scheduling Algorithm

Preventive maintenance scheduling requires three resources: manpower (technician), material (spare parts) and
equipment [4]. The scheduling support tool for this problem should be concerned with due dates of maintenance
tasks and availability of all resources, and should be able to elaborate a short-term plan (weekly or monthly). It
should also consider the equipment availability because in the company there are equipment with long operating
periods, and consequently, with limited availability for maintenance tasks.
A scheduling support tool that meets the company’s requirements is being developed. This support tool is based on a
preventive maintenance scheduling algorithm (PMS) that can be divided in three parts with specific objectives:
2126 Patrícia Senra et al. / Procedia Manufacturing 11 (2017) 2123 – 2130

• Identification of the scheduling period and equipment availability for preventive maintenance in this period;
(step 1of the algorithm)
• Tasks ordering (steps 2 and 3);
• Allocation of tasks to each technician based on the defined order and equipment availability (steps 5 – 6).
The tasks ordering is made in two steps. In the first step, all the maintenance tasks are ordered by a criticality
index for equipment, designated by Business Impact Rating (BIR), in descending order. Then, in the second step, for
each maintenance tasks with the same BIR, a second sorting is done by due dates. The tasks will be assigned
following this last order. Both ordering have the aim to minimize delays on the execution of maintenance tasks,
especially in the most critical equipment. Maintenance tasks with the same BIR and due date will be sorted by their
processing times in descending order. These orderings, besides being used to prioritize the assignment of tasks, will
also allow the visualization of all the maintenance tasks that should be executed in a short–term horizon and to be a
support in the negotiation between manufacturing and maintenance departments to provide time for preventive
maintenance performing. The visualization is possible through “Maintenance Tasks Table”, a table in the company’s
CMMS that contains all tasks ordered to be scheduled.
The main objective of BIR is the impact evaluation of equipment breakdown on productivity. BIR is obtained
through two criteria: CA – Equipment utilization and CB – Availability of alternative equipment. The “equipment
utilization” criterion shows the production line’s criticality. A line that is being used more than other lines is more
critical because it has less time available to perform maintenance tasks and less capacity slack to be able to recover
from downtimes and meet deadlines when there are delays. The criterion “availability of alternative equipment”
intends to show if there are equipment in the production line that can do the same operations as the equipment that is
being assessed. The existence of alternative equipment can decrease the impact of a breakdown since production can
continue, using the other equipment. Each criterion can be classified in three levels, where the higher level is the
most critical. Each combination of two criteria was analyzed and for each one, a classification with four levels was
assigned as presented in Table 1.

Table 1 - Business impact rating

Equipment utilization Availability of alternative equipment BIR
3 2 4
3 1 3
2 2 3
2 1 2
1 2 2
1 1 1

The planner defines and introduces in the scheduling tool a scheduling period. Consequently, it will be possible
to identify the equipment with available intervals during this period. Consecutively, the equipment should be
ordered by BIR and the maintenance tasks associated to the available equipment should be imported from the
“Maintenance Tasks Table”.
The allocation of maintenance tasks as previously mentioned, should consider the availability of spare parts,
equipment and technicians. So, the algorithm starts the allocation process with the equipment with higher BIR. This
way, the allocation is done in parallel among equipment with the same the BIR taking into account due date of
maintenance tasks, and in serial among the equipment with different BIR value. The ordering by processing times
prioritizes the tasks with higher durations.
While the equipment is available and there are also available technicians, the PMS will check if the task requires
spare parts, and if they are available. In case of unavailability of the spare parts, spare part acquisition is requested
and the maintenance task will not be scheduled. In contrast, if the spare part is available the PMS will search an
available and competent technician. The same process happens when the maintenance task does not require spare
part. The allocation of maintenance task to technician starts selecting the technician with, at least, the lowest level
competence required. If there is an available and competent technician, the maintenance task is scheduled. When a
maintenance task is scheduled, its data information is updated. At the end, after checking if all the maintenance tasks
imported can be scheduled during the availability time window of equipment, the PMS will return a schedule
solution. The PMS structure is the following:
Step 1. Identify the time windows availability of equipment and technicians;
Step 2. Sort the maintenance tasks by BIR in descending order;
 Patrícia Senra et al. / Procedia Manufacturing 11 (2017) 2123 – 2130 2127

Step 3. For each BIR value:

3.1 Sort the maintenance tasks by due date.
3.2 When the maintenance tasks have the same BIR and due date, order by processing time in descending
order.
3.3 Return a “Maintenance Tasks Table”.
Step 4. (Scheduling period)
4.1Define the scheduling period.
4.2 Filter the available equipment for the pre-selected scheduling period:
4.2.1 Order the equipment by BIR in descending order;
4.2.2 Import the maintenance tasks previously ordered associated to each available equipment in
the scheduling period.
Step 5. (Allocation for each equipment following the order defined in 4.21.)
5.1 While the equipment is available and there is competent technician:
5.1.1For each imported maintenance tasks ordered by BIR and due date:
5.1.1.1 If the maintenance task requires spare-parts:
5.1.1.1.1 The spare part is available: Go to step 5.1.1.3.
5.1.1.1.2 The spare part is not available: Request the spare part acquisition and
the maintenance task returns to “Maintenance Tasks Table” to be scheduled
later.
5.1.1.2 Else: Go to step 5.1.1.3.
5.1.1.3 Filter the technician with a level of competence equal or higher than it is required.
5.1.1.3.1 For each technician:
5.1.1.3.1.1 If the technician is available: Schedule the maintenance task
and go back to 5.1.
5.1.1.3.1.2 Else: Maintenance task returns to “Maintenance Tasks
Table” to be schedule later and go back to 5.1.
Step 6. Return the solution.

4. Application case

For the application case, it is considered that the technician team is composed by three technicians competent to
perform all maintenance tasks to be scheduled and their availability is presented in Table 2. Furthermore, for
simplification, it is considered that the maintenance tasks do not require spare parts. The defined scheduling period
starts at 10:00 a.m. of February 9th, 2016 and ends at 17:00 p.m. of February 10th, 2016 where only few equipment
have availability windows (Table 2). The respective maintenance tasks imported from “Maintenance Tasks Table”
are presented in Table 3.

Table 2 - Availability of Technicians and Equipment

Availability
Technician’s Equipment’s
Designation Work Schedule Designation BIR Day Hour
09/02/2016 18:00-24:00
T1 06:00 – 14:30 GP01 4
10/02/016 24:00 - 7:00
T2 14:30 – 23:00 ASY10_0010 3 10/02/2016 06:00 - 17:00
09/02/2016 18:00 - 24:00
T3 23:00 – 06:00 FCT50_0010 2
10/02/2016 24:00 - 04:00
BUR50_0301 1 10/02/2016 11:00-24:00
09/02/2016 10:00-20:00
VT17 4
10/02/2016 16:00-24:00
The tasks ordering allows to understand the maintenance tasks that should be scheduled first. In cases of tasks
with the same BIR and due date, they were ordered by processing time in descending order. This way, the available
time of equipment will be filled by the maintenance tasks with higher durations first.
2128 Patrícia Senra et al. / Procedia Manufacturing 11 (2017) 2123 – 2130

Table 3 - List of maintenance tasks filtered and ordered

Equipment Maintenance Tasks
index Line Designation BIR Designation Due Date Duration
t1 2N16 GP01 4 Replacing the Stator and Rotor of the Dispensing Pump 09/02/2016 108
t2 2M05-2 VT17 4 Cleaning the flow tank 09/02/2016 34
t3 2N16 GP01 4 Curtains replacement of in EPM's 10/02/2016 240
t4 2M05-2 VT17 4 Check flow quantity 10/02/2016 14
t5 2M05-2 VT17 4 Machine Check 10/02/2016 11
t6 2N05 ASY10_0010 3 Replacing Vacuum Pumps (suction cups) 10/02/2016 4
t7 2N05 FCT50_0010 2 Replace interface needles 09/02/2016 96
t8 2N05 FCT50_0010 2 Internal cleanliness 09/02/2016 34
t9 2I05 BUR50_0301 1 Maintenance level 1 aparaf. Guides and suction cups (ASI) 09/02/2016 30
t10 2I05 BUR50_0301 1 O-ring replacement 09/02/2016 6
t11 2I05 BUR50_0301 1 Preventive Maintenance PC - Cooler and Mass 10/02/2016 53
t12 2I05 BUR50_0301 1 Cleaning the oven glasses 10/02/2016 32

Analyzing the Table 3, the allocation process will initiate with t1 task with a duration of 108 min. The equipment
associated is GP01 which has the following available periods: 18:00 to 24:00 at February 9th and 24:00 to 07:00 at
February 10th during the schedule period, as shown in Fig. 2. The technicians T2 and T3 are available during the first
available period of GP01, and the technicians T3 and T1 are available during the second period. All technicians are
available at different instants. In order to respect the due date and minimize the delays, the maintenance tasks will
be allocated at the first available instant that all resources are available. Thus, t1 will be allocated to technician T2 at
18:00 (Fig. 2). The same process will be executed for the second maintenance task (Fig. 3) and so on until one of the
resources finishes. In Fig. 4 is presented a Gantt chart for all the equipment that have available periods during the
schedule period, with all the tasks allocated.

09/02/2016 10/02/2016
Hours 11 12 13 14 15 16 17 18 19 20 21 22 23 24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
GP01 t1
VT17
ASY10
FCT50
BUR50

Fig. 2 - Gantt chart with first scheduled task.

09/02/2016 10/02/2016
Hours 11 12 13 14 15 16 17 18 19 20 21 22 23 24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
GP01 t1
VT17 t2
ASY10
FCT50
BUR50

Fig. 3 - Gantt chart with second scheduled task.

 Patrícia Senra et al. / Procedia Manufacturing 11 (2017) 2123 – 2130 2129

09/02/2016 10/02/2016
Hours 11 12 13 14 15 16 17 18 19 20 21 22 23 24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
GP01 t1 t3
t4
VT17 t2 t5
ASY10 t6
t7
FCT50 t8
t9 t11
BUR50 t10 t12

Fig. 4 - Gantt chart for the equipment.

In this example, there are few maintenance tasks to be allocated, hence the available period of equipment are not
totally filled and there were not conflict situation during the allocation process. The PMS aimed to perform the
maintenance task allocation on the beginning of the equipment’s availability period, which allows a margin to
possible delays without compromising the task accomplishment in time. It is important to respect the due dates of
each maintenance task to ensure the production quality and the proper performance of each equipment. The diagram
in the Fig. 5 is the result of the allocation process that considers the availability of all resources: equipment, spare
parts and technicians.

09/02/2016 10/02/2016
Hours 11 12 13 14 15 16 17 18 19 20 21 22 23 24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
t2 t6 t9 t10 t11 t12
VT17 ASY01 BUR50
T1
t1 t7 t8 t4 t5
GP01 FCT50 VT17
T2
t3
GP01
T3

Fig. 5 - Gantt chart for technicians.

In Fig. 5, it can be noticed that t9 and t10 were scheduled with delay due to equipment unavailability (BUR50). In
this example, the total tardiness obtained was ΣTj = 2 (days), there is 2 tardy jobs in total (ΣUj =2; t9 and t10) and the
maximum lateness (Lmax) is one day.
In this company the technicians are also allocated for production work and corrective maintenance. Therefore,
although the distribution of preventive maintenance tasks and the total completion time (ΣCj) are not equal between
the three technicians (see Table 4), the remaining hours of the shifts could be occupied with production
support and corrective maintenance interventions.

Table 4 - Maintenance tasks scheduled to each technician

Technician Number of maintenance tasks allocated ΣCj

T1 6 159
T2 5 263
T3 1 240

This instance was also solved by MILP and the same solution was obtained. It is a successful outcome since the
PMS was able to provide an optimal solution. Through the analysis of this example, it is possible to identify some
specific situations that need special attention and study:

• The maintenance tasks with small durations should first be scheduled on the gaps between the shifts and lunch
time, consequently the longer available intervals of equipment would be filled by the maintenance tasks with
longer durations.
2130 Patrícia Senra et al. / Procedia Manufacturing 11 (2017) 2123 – 2130

• It would be interesting to apply the opportunistic maintenance concept, i.e. when a corrective maintenance
intervention is performed, preventive maintenance tasks with small processing time could be performed, allowing
the increase of the efficiency of the maintenance task execution.
• The ordering of maintenance tasks adopted only considers the equipment’s criticality, it does not consider the
impact of maintenance task non-achievement on the equipment performance. This can lead to situations that
available time windows of equipment are totally occupied with maintenance tasks schedule with low impact on
equipment performance that could be schedule later, instead of being occupied with maintenance tasks with
higher impact on performance equipment. Therefore, this criterion will be considered in the ordering method.
• The objective function only considers the tasks delays.

5. Conclusion

An algorithm is being developed to construct a quick scheduling solution, taking into account availability of
equipment and technicians, processing times of maintenance tasks, due date, competence level required and
downtime of production line. The algorithm orders the maintenance tasks by a criticality index equipment and due
date and the allocation to technicians occurs after verifying the availability of required spare parts. Thus, the
algorithm that supports the scheduling support tool is composed by identification of availability times, an ordering
method and an allocation method, and aims to minimize the total tardiness.
Through an application example, it was possible to show the proposed algorithm’s effectiveness. The ordering
method adopted in this initial approach can lead to situations that decrease the efficiency of scheduling, such as the
maintenance tasks related to equipment with lower BIR value that have higher impact on the quality and
performance of the equipment are not scheduled, because the algorithm only prioritizes the equipment with higher
BIR. The same goes to spare parts, where this method can lead to situations where the spare parts are allocated to
equipment with higher BIR value, when it can be more urgent to allocate them toequipmentwith lower BIR value (if
the spare required is the same).
In this first computing experience, it was possible to conclude that the PMS is a promising scheduling approach
method. Future work will be focused on tasks ordering considering the impact of not performing the maintenance
task as well as the possible tolerance in due date for a given task. After the complete definition of the algorithm, the
associated computational application will be developed to be integrated in the company’s CMMS.

Acknowledgements

This work has been supported by the Portugal Incentive System for Research and Technological Development.
Project in co-promotion nº 002814/2015 (iFACTORY 2015-2018).

References

[1] EN 13306. European Standard: Maintenance Terminology 2010.

[2] Ebrahimipour V, Najjarbashi A, Sheikhalishahi M. Multi-objective modeling for preventive maintenance scheduling in a multiple production
line. J Intell Manuf 2013;26:111–22. doi:10.1007/s10845-013-0766-6.
[3] Pinedo M. Planning and Scheduling in Manufacturing and Services. 2009. doi:10.1007/978-1-4419-0910-7_7.
[4] Paz NM, Leigh W. Maintenance scheduling: issues, results and research needs. Int J Oper Prod Manag 1994;14:47–69.
[5] Leung JYT. Handbook of scheduling: algorithms, models, and performance analysis. CRC Press; 2004.
[6] Duffuaa SO, Al-Sultan KS. A stochastic programming model for scheduling maintenance personnel. Appl Math Model 1999;23:385–97.
doi:10.1016/S0307-904X(98)10009-4.
[7] Samhouri MS. An intelligent opportunistic maintenance (OM) system: A genetic algorithm approach. TIC-STH’09 2009 IEEE Toronto Int
Conf - Sci Technol Humanit 2009:60–5. doi:10.1109/TIC-STH.2009.5444428.
[8] Manzini R, Accorsi R, Cennerazzo T, Ferrari E, Maranesi F. The scheduling of maintenance. A resource-constraints mixed integer linear
programming model. Comput Ind Eng 2015;87:561–8. doi:10.1016/j.cie.2015.06.006.