Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
FUZZY LOGIC PRIORITIZING MAINTENANCE WORK IN
OPERATIVE PLANNING
Marinko Aleksić1 and Petar Stanojević2
1
Naval Repair Facility, Montenegro, Tiva
2
Petroleum Industry of Serbia, Serbia, Novi Sad
Dr Marinko Aleksić is Head of the Technology Department of Electronic
Sector in the Naval Repair Facility Tivat (Montenegro Ministry of Defense)
and Visiting Professor at Kotor Maritime Faculty. He was previously First
engineer for Overhaul Technology in NRF Tivat Montenegro, Head of the
Naval overhaul working place in Libya and Head of Maintenance Sector in
the Zenica milk industry in Bosnia and Herzegovina. Dr Marinko Aleksić
was an expert team member in 10 projects which include Type "Kotor"
Frigate reconstructions, Navigational & Torpedo System ORKA 2000,
Command-control & communication system HORIZON, Marine GIS.
He has an 18 years experience working as an engineer and maintenance
manager. He has a BSc in industrial electronics and a PhD in maintenance
engineering. He is also an Expert for RCM / RBI and has particular interest for Maintenance
management, Risk management, FMEA and RCFA. He has authored and co-authored 2 textbooks, and
about 30 articles.
[email protected] Dr. Petar Stanojevic, is a special adviser to the President of the NIS a.d.
(Petroleum Industry of Serbia) management board. He was also the Managing
Director of Corporate Development Department of NIS a.d (Petroleum Industry
of Serbia)., the Deputy Secretary of the state in Ministry of Defense, Professor
at the Military Academy, School of National Defense and Professor at Belgrade
University. Dr Stanojevic was project & program manager of 15 projects
including the Opportunity Confirmation Program with Shell GSI, Restructuring
and Reorganization of NIS a.d., Implementation of risk based approaches in
NIS, and Integ-risk (EU FP7 project) etc.
He has a PhD and MSc in Industrial Engineering and five years experience in
managing the development of a national oil and gas company, ten years of academic and research
experience and five years of experience working as an engineer and maintenance manager. He has
authored and co-authored 3 textbooks, two monographs and about 100 articles. Dr Stanojevic has made
three research study visits abroad and has been an invited speaker at conferences and symposiums.
Abstract
Starting from a statement that a basic problem in the operative planning process is a
unique and automaticated methodology for determination of priorities in maintenance
work, a hierarchical model for determination of maintenance work priorities was made
and fuzzyficated. An original method for modular production of complex fuzzy systems
using Matlab and Symulink was developed. A basic simulation model was made to deal
with a wide range of problems with a possibility to analyse sensitivity of final solution
to different input variables. Application of this model is simple and it has a possibility
to shorten the basic model due to the increase in its use in the operative planning
process.
Key words: maintenance, operative planning, priorities, fuzzy logic, modeling, simulation.
1
Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
1 Problem
As most facilities or plants are still fighting against the maintenance of reactive style,
the probability of breakdowns and other unplanned events which interrupt daily flow of
work is high. Also, in most organizations the workload is excessive and priorities are
constantly monitored and changed. The drawbacks due to an unclear definition of
priorities [1, 2] are as follows:
• Waste of maintenance man-hours on tasks of a relatively low importance
• Crucial tasks lost in maintenance backlog
• Unutilized maintenance capacities
• Dissatisfied shop floor management, personnel and customers
• Lack of faith in the effectiveness of maintenance system and its capacities
• Security issues growing
• Cost enlargement
In fact, without a clear system of priorities, it would be impossible to either implement a
selected strategy or to achieve planned objectives, because there is no way to
differentiate “the important” from “unimportant”.
This is why the methodlogy for priority determination makes a very important segment
of operative planning in maintenance. There are different methods for determining
maintenace work priorities and some of them will be mentioned. The most frequent
situation in an organization is illustrated by a method of “emotional” and “real”
priorities [3]. The text says that a typical “emotional” priority is based on feelings
instead of on objective determination of priority importance.
When determining priority of their own maintenance requirements, people only have in
mind their own scope of activities.
Also, everyone in an organization takes advantage of the priority system and always
define their requirements to have priority higher than necessary as, otherwise, it may
happen that their work is never done.
A “real” priority in maintenance work is defined by viewing consequences of undone
work and equipment condition determined by inspection.
“Real” priorities lead towards the idea that most of the work is planned and then
efficiently carried out according to a schedule, which as a consequence has cheaper
maintenance work and more effective dealing.
One of the methodologies for determining priorities, which is further elaborated and
adjusted for practical use, is a relative priority gradation technique or a “REDRIST”
technique. The “REDRIST” index is an output of two criteria. The first criterion is an
index of a relative importance of a facility and the second criterion is an index of a
relative importance of a type of maintenance work. The lower value of “REDRIST”
index [4,5] is, the higher the priority of an operation procedure is.
2
Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
NASA's RCM guide book recommends a priority determination method based on the
consequences of Equipment Failure [6], which is in accordance with a basic principle
for the selection of a maintenance strategy based on RCM methodology.
Priorities are determined by an impact on mission, safety of life and availability of
redundancy.
Like the “REDRIST” technique, [1] also suggests a priority determination technique
based on the following:
• Equipment cruciality
• Effects of a task or work accomplished
• Limitations
Equipment cruciality and effects of a task are first determined and there are nine levels
of importance. A combination of equipment cruciality and effects of work can be cross-
referenced into a matrix to provide a relative weight of each task in comparison with all
other types of tasks. The time frames within which these prioritized tasks are allowed to
appear are marked with colors. Restrictions on the flow of work in the reality add the
third dimension to the process of prioritising and they are labour shortage, availability
of facilities, spare parts etc.
There are simpler methods like a method of dynamic coefficient [7], and more
complicated methods like a method for simulated dynamic operative planning with
changeable arrangement criteria [8].
A large number of companies in the field of process industry, especially in oil and
petrochemical complex use risk matrices as a means for determining the level of
importance of some component within a system (based on the relation of likelihood –
consequences), and then, on the basis of the position that the component has in the
matrix, the priority of maintenance activities is being determined.. For determination of
consequences, business effects are considered (cost of failure), safety, environment,
reputation, etc.
The majority of power plants do not have a formal system for prioritization of
maintenance activities and they are using a logic system instead, based on the fact that
first we have to implement those actions for which the consequence, of failure to
implement, would be a production downtime. Which of the maintenance activities will
be implemented depends also on the budget allocated to maintenance department.
Stated methods do not have an unique approach and emphasize different criteria that
influence maintenance priorities. It is obvious that determination of priorities is better if
as many as possible (all) influential criteria in an organization are taken into account.
However, this brings us to a complex problem which is harder to resolve in practise.
This is why it is resorted to its hierarchical structuring or a multiphase approach which
slows down the methodology. Such complex or multicriteria problems can be resolved
using multicriteria methodologies like, for example, an AHP method [9]. However, in
this case, their fast and practical use is questionable as existing softwares are not
adjusted for such a use. Hence, there is also a problem of a need for instant and practical
use of the priority determination system. Because of this, the problem is simplified and
only one or two influential criteria are taken into consideration.
3
Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
2 Selected method
Starting from an assumption that more crucial criteria need to be taken into consideration
and at the same time to provide a practical and fast solution to a problem, there is a
suggestion to imitate the manner in which a human expert would solve a problem. Experts
do not use mathematical relations but on the basis of their own experience and
knowledge they can establish interdependence between certain criteria and give an
optimal classification according to the priority in a group of more maintenance
requirements. Such a way of priority determining can be presented verbally using
linguistic IF-THEN rules. It appears that an expert ranks every criterion but with
inaccurate and imprecise characteristics. So, if, for instance, Existent time delay is chosen
to be a criterion, it can be described as short, mean or long. It cannot be said where clear
boundaries between and transitions from SHORT to MEAN and from MEAN to LONG
are. Exactly because of this a solution to a problem has been suggested using fuzzy logic
and approximate reasoning. The use of fuzzy logic may turn expert knowledge expressed
inaccurately and verbally [10,11] into a unique and easily comprehensible methodology
for the determination of priorities in operative planning. In fuzzy-logic systems the
knowledge is presented with IF-THEN rules which are a suitable verbal form of human
knowledge. Fuzzy logic efficiently solves the problem of relative gradation of
importance of particular criteria. Fuzzy systems can be easily made. It allows an
introduction of more criteria for analysis and of those which could not have been analized
so far, which increases the quality of decision-making. After the system has been adjusted,
both rough and precise criteria quantification are allowed.
In this way it is possible to achieve significant improvements also in the systems enabling
relatively precise determination of the sequence of maintenance activity implementation,
such as the implementation of the risk matrix. Within these systems maintenance activities
are categorized into the groups depending on their position in the risk matrix, however the
problem of ranking within a group remains unsolved. The solution may be reached by
using the Fuzzy logic methodology which enables the generation of a clear sequence of
activities in the output.
This methodology also enables a significant reduction of the number of errors which may
arise as a consequence of constant recalculations and changing the importance of certain
factors.
3 An outline of the model
A model for setting up maintenance priorities is made for solving a wide range of
problems. Because of this, more criteria had to be taken into consideration which
contributed to the development of a complex model. In order to present the expert-like
way of thinking and to solve problems, the system is created to have a hierarchical and
modular structure. Expert-like way of setting up priorities is turned into an algorithm
and included in the model. A clear relation and interdependence between criteria is
established, which represents a basis for output values in the model. The structure of the
model is adjusted for the use of fuzzy logic and approximate reasoning. The main
following criteria are taken into consideration:
1. Asset item importance,
2. Mission (task) importance,
3. Maintenance action importance,
4. Existent time delay in the maintenance system and
4
Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
5. Logistics and Administrative delay time.
The first three criteria are defined more specifically in a way that Asset item importance
is being defined by taking into account the following: Costs of breakdown,
Consequences of breakdown on major system and Availability of redundancies.
Mission (task) importance criterion is defined by taking into account the following:
Exploitation coefficient, Mission (task) type and Organization subsystem importance
for the wider system. Maintenance action importance criterion is defined by taking into
consideration the following: Acceptable time delay, Time to repair and Comparative
maintenance task importance.
The basic outline of the model of the system consists of five modules hierarchically
organized on three levels, Figure 1, in such a way that output criteria on one level are
input criteria on the next, higher level. Output value in the model is a priority measure
and the name proposed for this value is Maintenance work priority index.
Figure 1. An outline of the basic model
Once the problem is solved on the operative planning level, taking into account a large
number of criteria efficiency of use is called into question. This is why it is emphasised
that a model designed on the suggested criteria is a basic model for a great number of
situations. It provides an opportunity to eliminate redundant criteria for specific systems
achieving the simplicity of its practical use [9]. This requirement could also be met
5
Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
because of fuzzy logic and its possibilities. For instance, if the first three criteria were
directly defined, the model would be reduced from 11 to 5 criteria.
In practice, Maintenance work priority index would be evaluated for every requirement
and their sorting would put requirements in order according to a priority. In Figure 1,
every criterion is put in a separate square and their mutual influence is presented with
arrows. Three criteria on the second level occupy the central place in the model: Asset
item importance, Mission (task) importance and Maintenance action importance.
Each of these criteria is defined with three criteria from lower, third level. General
importance criterion is on the first level as a result of a subliminal influence of the
criteria from the second level. Together with Existent time delay in the maintenance
system and Logistics and Administrative delay time criteria, they determine the output
value for Maintenance work priority index model.
The first model sets up Asset item importance criterion, which is an output value in the
module and depends on criteria 1, 2 and 3 from the third level. The criterion marked
with number 1, Costs of breakdown, shows that a techical device (short TD) is all the
more important as Costs of Breakdown increase. The influence of this criterion is
directly proportionate, thus higher costs contribute to bigger Asset item importance.
The criterion marked with number 2 is named Consequences of breakdown on the major
system. It can be said that as the demand for a technical device for the major system
operation increases, its importance is bigger. Criterion marked with number 3,
Availability of redundancies, shows the possibilities of a replacement of a broken-down
technical device, permanently or until it is repaired, with another device, which would
reduce maintenance requirements emergency.
Thus, this criterion has an inversely proportionate influence on the output value, Asset
item importance, in the first module.
The second module determines Mission (task) importance criterion. It is an output value
in the module which depends on criteria 4, 5 and 6 on the third level. The criterion
marked with number 4, Operation coefficient, takes into account the extent to which the
device is used or should be used. The criterion marked with number 5, Mission (task)
type, takes into consideration the mission or task the technical device is used for. At a
given time some tasks or missions have different priorities which determine a higher or
lower value of the Mission (task) type. The criterion marked with number 6,
Organisation subsystem importance for the system, takes into consideration how much
an organisation subsystem is important for the operation of the system or to what extent
the failure of a technical device will reduce production or overall effects of the major
system. As for the influence of the specified three criteria from the third level on the
value of Mission (task) importance criterion, it can be said that they have a directly
proportionate influence.
It may be recommended to make a difference between Mission (task) importance and
Asset item importance because, at some point in time, the two Assets of the same type
may be present in missions of different levels of importance for the facility/company. It
has been shown that this kind of difference is important also for military systems, for
which the methodology was actally originally developed.
6
Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
As in the first two modules, Maintenance action importance (MA) also depends on the
criteria 7, 8 and 9 from the third level and together they make the third module. The
criterion marked with number 7 is specified with the name Acceptable time delay. There is
usually a period of time during which the fulfilment of requirements can be delayed
without incuring additional expenses or causing reduction of the quality of service and this
period of time is defined as Acceptable time delay. It can be said that the higher the value
of Acceptable time delay criterion, the lower is the Maintenance action importance. The
criterion marked with number 8, Time to repair, is time needed for the performance of a
required maintenance action. Shorter maintenance actions should have higher priority.
Thus, Time to repair has an inversly proportional influence on the value of Maintenance
action importance criterion. The criterion marked with number 9, Comparative
maintenance task importance, takes into account the consequences of a possible delay of
maintenance work completion in cases like a danger of an accident, resulting damages and
interruption in production. Here it is easy to make a gradation among maintenance actions
which are related to different safety issues, environmental issues, reputation issues etc. In
the future it might be possible to make a certain improvement of this model by using a risk
matrix or by modifying the Table 1. It should be considered that preventative tasks
including statutory required maintenance actions (inspections and similar) will never have
higher priorites than corrective actions and that their acceptable time delay is always
significant, and thus they will always have a lower priority. In order to prevent a possibility
of these and other maintenance actions of lower priority not being implemented ever, the
factor of Existent delay time is introduced which tells us how much a certain action will
stay in the backlog and the longer this time is, the priority of its implementation is higher.
Comparative maintenance task importance influence on the value of Maintenance action
importance criterion is directly proportional.
It can be said that the criteria 10 and 11 present system restrictions. The criterion
marked with number 10 is named Logistics and Administrative delay time.
Determination of the value of this criterion is based on logistics and administrative
maintenance waiting time and it represents a deviation of a real system from an ideal
maintenance system, which would provide all indispensable resources in any place at
any time. It can be said that this criterion has an inversely proportional influence on
Maintenance work priority index value.
The last criterion marked with number 11 is named Existent time delay in maintenance
system. Experience has shown that in a maintenance workshop a technical device for
some reason often has to be put aside for a period of time before it is maintained. If
there are always more prioritised reqirements, it may happen that the technical device is
never maintained. If Existent time delay is long, the technical device should be awarded
higher priority and in that way take its turn for maintenance. Hence, the value of
Existent time delay in the maintenance system has a directly proportional influence on
the value of Maintenance work priority index.
This is one of the problems that have not had an acceptable solution so far.
4 Model fuzzification
Defining certain criteria and establishing their interdependence is not the end of the
modeling process because there is a need to translate expert knowledge into an
algorithm and introduce it into the model using IF-THEN rules. As the described outline
7
Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
of the model for the interdependence system is adjusted to fuzzyfication, it is modeled
as a complex fuzzy system consisting of five fuzzy modules hierarchically organised on
different levels. Every module is one fuzzy system consisting of three input linguistic
variables and one output linguistic variable, as shown in Figure 2. The modular
principle is used for easier system analysis and, by this, easier verbal description of its
operation i.e. for drawing up IF-THEN rules. The linguistic variable Comparative
maintenance task importance, has triangle forms of memebership functions of input
linguistic values with the following parameters: SMALL (0,0,1), RELATIVE (0,0.5,1),
BIG (0.5,1,1). Other linguistic variables also have triangle input linguistic values with
the following parameters: Acceptable time delay - SHORT (0,0,1), MEAN (0,0.5,1),
LONG (0,1,1). Time to repair - SHORT (0,0,1), MEAN (0,0.5,1), LONG (0,1,1).
Comparative maintenance task importance - SMALL (0,0,0.5), RELATIVE (0,0.5,1),
BIG ( 0.5,1,1).
Figure 2. A fuzzy-module
Modelling of every input linguistic variable is done using three linguistic values. This
sets up an input base of rules for one module counting 27 rules all told. Researches have
shown that this is a number of rules which may include experts’ knowledge and which
are not at the same time loaded with data experts cannot control. In case no modular
principle has been used, an input base of 311=177147 rules would have to be used. It is
clear that no expert could express his knowledge in this way. Thus, the introduction of
the knowledge of the system was commenced by modelling hierahical model with the
total of five bases of rules with 27 rules each, which is 135 rules all told. After
8
Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
optimisation of the number of rules, for which a special method was devised, the system
now has 69 IF-THEN rules all together [9]. The defuzzification method chosen was a
Centre of Gravity method (COG) since it is common and practical for the development
of the fuzzy system and it ensures required continuity and graduality of an outcome
[12].
It can be said that the model imitates an expert or a decision maker and a user has a
clear insight into the expert knowledge through the IF-THEN form of rules. Using these
rules, the expert qualitatively describes the system which is later adjusted for finding
more precise solutions. Also, every rule clearly indicates how to reach the goal, that is,
increase maintenance efficiency. An example of the following rule states that
requirements are fulfiled:
IF AVAILABILITY OF REDUNDANCES IS GREAT AND COSTS OF
BREAKDOWN LOW AND CONSEQUENCES OF BREAKDOWN ON THE MAJOR
SYSTEM MINOR THEN ASSET ITEM IMPORTANCE IS SMALL.
5 A type of simulation model
A full-scale interdependence model, outlined as a complex fuzzy system, cannot be
designed in the Fuzzy Logic Toolbox for Matlab [13] since this software package is not
designed for modular development of complex systems. For this reason, the problem is
solved by simulation in the Matlab standard module for simulation – Simulink, as it
links blocks in a simple manner and also supports the fuzzy systems modeling. Matlab
is an adequate package since it can generate an original code in C language. This
provides a basis for integration into already made programs whose purpose is to
automate the maintenance planning process. Figure 3 shows a block of schemes of a
full-scale simulation model.
9
Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
Figure 3. A block of schemes of a simulation model
A graphic presentation of the model using blocks was used as the most convenient and
fastest way to edit a model and achieve desired results. Sensitivity testing is a very
complex task, thus, a very complex simulation model was produced because of which
groups of blocks are pooled. The scheme presents it as one block, and it contributes to
model’s clearness. The picture shows that the simulation model consists of two major
blocks: the fuzzy system and sensitivity analysis subsystem. It is obvious that input
values are entered into blocks with the names of every value. Rectangles display output
value reading. The simulation model has all the facilities needed for designing and
verification of a model and a user can set aside only a fuzzy system block illustrated on
Figure 4.
10
Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
Figure 4. The fuzzy system
For a user (practicioner) it represents a system which has eleven input variables and one
outcome, which represents a preference index for maintenance prioritizing. The user
does not need to know how the system operates, but there is a possibility to observe in
detail how either a simulation model or a fuzzy system module operates. During
analysis, the observation of results is provided as every fuzzy module has a display for
result reading.
In order for this simulation model to operate, certain problems related to the following
needed to be solved:
1. Merging of fuzzy modules in a unique system
2. Specification of the interchangeable variables base of rules
3. Reduction of interval in every module output [14]
4. Restriction of input values in the simulation model from upper and lower side,
depending on the trust interval of every input variable.
5. Sensitivity analysis by variation of one of the entries and its graphic
presentation.
11
Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
To show a graphic presentation as clearly as possible and to perform a simulation as
simply as possible initiation and determination of graphic presentation is automatic
which is left out if sensitivity analysis is not required.
The subsystem for fuzzy model sensitivity analysis provides a variation of certain input
values from their minimal to their maximal value in the universe of discourse.
Sensitivity analysis is done after an ordinal number of an input variable is entered in a
sub-block called Varying input. The result of the variation is given in a graphic form on
Figure 5 which represents a change of priority index depending on a changed variable.
A constant value of other variables is entered in headings which, in such a case,
represent analysis parameters. Sensitivity analysis should be planned and performed
together with the use of a graphic presentation of a group of possible solutions for every
module.
Figure 5. Variation of the variable
After verification and validation of the model an independent program is made since
Matlab generates C-code. The program may be linked with Excel where input and
output results are charted, sorted by the order of priority for any number of maintenance
requirements.
12
Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
Table 1. Comparative maintenance task importance criterion
6. An example of use
For practical use, in [9,4] references are given on how to determine the input criteria
value. Using these references the user should be able to rate the value of every input
criteria. An example is given in Table 1. with references for general numerical values of
Comparative maintenance task importance criterion. It is easy just to click on
maintenance task term and to have all the other data input automatically.
This model has been used for a longer period of time in Electronic Division in Naval
Repair Facility Tivat for operative planning requirements. The Model and all IF-THEN
rules were tested and validated on the example of military ships. For a complete
procedure of 50 maintenance requirements or so, a trained user (planer) may finish in 45
13
Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
minutes which justifies the use of this model. The Model improved the planning
possibilities; however, its further application was practically stopped because the former
Yugoslav Navy was disbanded. The program is combined with the existing software for
network planning and project management especially because this software has solved
the problem of distribution and balancing of resources.
The disadvantage of this approach is that it has to be adjusted to concrete problems. In
that case a simpler model is better, because, for example for military ships, it is
impossible to adequately determine the cost of downtime. Also, there is a problem of
connecting with different CMMS. However, this does not diminish model's qualities.
7. Conclusion
One of the basic problems in operative planning process is a unique and automatic
methodology for maintenance work prioritizing. Determination of maintenance system
priorities is modeled by the use of fuzzy logic with a purpose to make improvements in
the planning process i.e. predominantly in business system efficiency and maintenance
systems. The model for determination of maintenance work priorities is built by taking
into consideration the following: more criteria, requirements of comparative gradation
of certain influential criteria importance; presentation of inaccurate knowledge
requirements i.e. requirements of translation of verbally expressed expert knowledge
into a clear and visible algorithm. Application of fuzzy logic may turn expert
knowledge expressed inaccurately in a verbal form into a unique and easily
comprehensible priority determination methodology when making operative plans.
Fuzzy logic effectively solves the problem of comparative gradation of importance of
certain criteria and more criteria have been introduced for consideration by now even
those that could not have been considered so far, which improves the quality of
decision-making. The complex system modeling process applying fuzzy logic suggests
an original way for modular development of fuzzy system using Matlab and Simulink.
A simulation model was suggested with a possibility of sensitivity analysis of a final
solution to certain input variable variation. The model was made to represent a basic
model for a wide range of problems but maximally simple to use. After the system has
been adjusted, it is allowed both rough and very precise quantification of criteria and
output results are also very precise so it is virtually impossible for two problems to
obtain the same priority (as it is a case in risk matrix for example). A relatively simple
shortening of the basic model was provided for specific and practical use which has
increased its use in operative planning process if one of important requirements is the
immediacy of problem solving. When maintenance priorities are set up, distribution and
balancing between the obtained and required capacities is done using network planning
or project management software.
Currently the model is operationally used in the Electronic Division in Naval Repair
Facility Tivat – Montenegro. Implementation of this solution with some other user
requires approximately three consultants who would work on adjustment and
finalization for 3 months and then three months on testing and validation of the model
in question (the presented, basic model, was developed for about 3 months). The
consultants would cover the areas of Fuzzy logic, Software development and
Maintenance. The staff of the client would have to be engaged in completing a specific
questionnaire and in a required number of workshops, and also to nominate one or two
persons as process owners who will apply the system in the future. From all the above
information, it can be concluded that duration and costs of implementation are low, and
14
Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
benefits would be higher in proportion to the costs of a specific system, because the
savings would be greater. In cooperation with the client, this model would be
customized to his needs, constraints, strategy and goals which would be used as a basis
for performing required quantifications and determination of IF-THEN rules. In this
way it would be possible to influence particularly the resolution of safety or
environmental issues depending on the client’s priority. There would be no other issues
except for the cautiousness and reserve of the client to use new and sophisticated tools.
In addition to obvious advantages in terms of more precise determination of priorites,
increased safety, avialability and reliability of assets, it is certain that in this way it is
possible to realize savings through a better planning of resources (including
subcontractors), and it might even be possible to extend backlog (from the currently for
petrochemical industry acceptable 3 months to 4 and more). Better planning certainly
may contribute also to resolving of various organisational issues, such as issuing
permits to work or obtaining permits to access certain assets.
8. References
[1] Mather D.: "Work Order Priorities"
http://www.plant-maintenance.com/ maintenance_articles.shtml
[2] Mather D.: "Work Order Execution and Data Capture"
http://www.maintenanceworld.com/articleArchive.htm
[3] Idhammar, C.: "Prioritize Your Maintenance Work Objectively", Maintenance
Planning Articles, http://www.maintenanceworld.com (February 2001).
[4] Stanojevic P.: "Influence of the Technical Factors on the Maintenance Systems
Organizational Structure", Doctoral Dissertation, Faculty of Mechanical
Engineering Library ( Kraljice Marije Street 16, 11120 Belgrade 35, Serbia),
Belgrade, 1997.
[5] Grothus, H.: "Maintenance Costs Planning", OMO, (Vatroslava Lisinskog Street
12a, 11000 Belgrade, Serbia) Belgrade, 1977.
[6] Reliability Centered Maintenance Guide for Facilities and Collateral Equipment,
National Aeronautics and Space Administration - NASA, 2001.
http://www.hq.nasa.gov/office/codej/codejx/Assets/Docs/RCMGuideMar2000.pdf
[7] Zelenović, D.: "Production System Management (Anatomy of Management
Procedure)", Naučna Knjiga, (Gračanička Street 16, 11000 Belgrade, Serbia)
Beograd, 1981.
[8] Radenković, B.: "Organisation System Modeling and Simulation Using a
Simulation System GPSS/FON", The Faculty of Organisational Science Library,
(Jove Ilića Street 154, 11000 Belgrade, Serbia)Beograd, 1991.
[9] Aleksić, M.: "Fuzzy Logic Model for Operative Planning in Maintenance
Systems", M.Sc.Thesis, Military Academy Library, (Pavla Jurišića Šturma Street
33, 11000 Belgrade, Serbia) Beograd, 2002.
15
Fuzzy logic prioritising maintenance work. OMMI, 2008, Volume 5, Issue 3, Dec.
www.ommi.co.uk
[10] Subašić, P.: "Fuzzy Logic and Neuron Networks", Tehnička knjiga, (Vojvode
Stepe Street 89, 11000 Belgrade, Serbia) Beograd, 1997.
[11] Zadeh, L.: "The Concept of a Linguistic Variable and Its Aplication to
Approximate Reasoning", Information Science, 1975.
[12] FuzzyTECH 5.31 Professional, http://www.fuzzytech.com
[13] Fuzzy Logic Toolbox for Use with Matlab. http://www.mathworks.com
[14] Jantzen, J.:"Tech. report No98-864", Technical University of Denmark,
Department of Automation, 1998. http://fuzzy.iau.dtu.dk/
16
