You are on page 1of 6

Proceedings of the 14th IFAC Symposium on

Information Control Problems in Manufacturing

Bucharest, Romania, May 23-25, 2012

Mathematical Decision Model to Improve TPM Indicators

Lucia Pascale*, Marin Mainea*,
Paul Ciprian Patic*, Luminita Duta*

*Electrical Engineering Faculty, ValahiaUniversity of Targoviste, 24 Unirii Ave., 130065
ROMANIA (e-mail:,,,

Abstract: The maintenance department has a key role so as to ensure the continuous production flow by
up-keeping the equipment at its normal functioning state. Total Productive Maintenance is a concept that
involves everyone in the company. Overall Equipment Effectiveness is the key measure of the tangible
benefits obtainable from the TPM strategy. This paper proposes a decision aid method based on a
mathematical model to improve the OEE indicator and the overall performance of the manufacturing line.
Keywords: Maintenance, line balancing, decision analysis, decision modeling

The goal of the TPM program is to increase production while
achieving zero defects, zero breakdowns and zero accidents,
Total Productive Maintenance (TPM) was developed in Japan in the same time increasing job satisfaction of the worker.
at the beginning of the seventh decade of the last century by
The targets of TPM are to: a) maintain an accident free
the Japanese Institute of Plant Maintenance (JIPM, 2002).
environment, b) enhance the machine efficiency towards
This concept challenges the view that maintenance is no more
100%, c) reduce manufacturing wastes and costs, d) deliver
than a function that operates in the background and only
the goods in time to clients, and e) develop multi-skilled
appears when needed. The objective of TPM is to engender a
sense of joint responsibility between supervision, operators
and maintenance workers, not simply to keep machines The five principles of TPM are (McCarthy and Rich, 2004):
running smoothly, but also to extend and optimise their
1 Adopting improvement activities designed to increase the
global performance (Venkatesh, 2006).
overall equipment effectiveness;
TPM concept brings maintenance into focus as a necessary 2 Improving existing planned and predictive maintenance
part of the enterprise policy. TPM is no longer regarded as a systems;
non-profit activity. Systematically maintenance is scheduled 3 Establishing a level of self-maintenance and cleaning
as a part of the manufacturing day work and is an integral carried out by highly trained operators;
part of the manufacturing process. The goal is to hold 4 Increasing the skills and motivation of operators and
emergency and unscheduled maintenance to a minimum engineers by individual and group development;
level. 5 Applying early management techniques to create reliable
and safe equipment and processes;
This paper is organized as it follows: first a brief presentation
of the TPM concept is done. Principles and characteristics of These principles are supported by seven corresponding
the TPM are presented. In the next section, calculus of some planned maintenance steps meant to guide the standardization
important TPM indicators is given, emphasizing the role of and simplification of maintenance activities (Tajiri and
managing their values during the manufacturing process. A Gotoh, 1999). This stepwise process has the effect of raising
mathematical decision model is developed in the third section the capability of production, maintenance and supervision
for a manufacturing line. By minimising the objective and of releasing specialist resources to focus on the next
function from the proposed model, new improved values of development stage. That is process optimization, a key part of
the TPM indicators are obtained. The results will show that a the active maintainer’s role once breakdowns are brought
well-balanced line provides a maximum value of the Overall under control. The designer of an industrial process has to
Equipment Effectiveness (OEE), one of the most important take into account in his activity the prerequisites demands of
indicators of the TPM strategy. A case study is presented in the maintenance steps (Patic and Pascale, 2010).
the last section.
TPM recognizes the importance of operator involvement
2. TOTAL PRODUCTIVE MAINTENANCE developing a close relationship between him and his machine.
This relation gave birth to the autonomous maintenance
Total Productive Maintenance is a maintenance strategy that which provides the solution for raising equipment condition
involves people in all levels of organisation. It can be to a level where zero breakdowns is possible. TPM tools
considered as the medical science for machine functionality. improve the effectiveness of the transformation process.

978-3-902661-98-2/12/$20.00 © 2012 IFAC 934 10.3182/20120523-3-RO-2023.00303

INCOM 2012, May 23-25, 2012
Bucharest, Romania

The method most commonly used in Lean Manufacturing to 4. MATHEMATICAL DECISION MODEL
measure the productivity of machinery is the calculus of OEE
Mathematical models with decision variables are often use in
performance indicator.
problems of tasks assignment. The idea is to achieve an
optimal assignment of task on stations so as to minimize the
real cycle time of the line.
Overall Equipment Effectiveness (OEE) is used as an
In the next section the following notations are used:
indicator of how well equipment is used in batch production.
This parameter is obtained in relation to losses that can m number of workstations
impede the equipment efficiency. OEE is a quantitative
indicator. It is the product of the availability, performance n number of tasks
and quality of the equipment (Venkatesh, 2006). j index of a task
i index of a workstation
OEE%  Availability%  Performance%  Quality%
tcy cycle time
Availability measures the percentage of time the machine
runs for production. It could also include a comparison of F the objective function
available size. If a machine is available for production, but it X tasks possible assignment matrix
is stopped for an unscheduled maintenance repair, OEE is Y tasks effective assignment matrix
reduced. The machine is available to run, but it can’t until the
T[n] operational times array
maintenance operations are finished.
Wi workstation i
In fact, availability indicates the performance of the
Maintenance Department. Its value is can be calculated using X is the assignment matrix and defines the possible off-line
equation (1). assignment of the tasks to workstations as it follows:

1 if task j can be assigned to station i

actual run time X (i, j )  
Availability%  100 0 elsewhere
planned run time (1)
1 i  m
Where 1 j n
actual run time=planned run time -idle time One could note that X(i,j) is a decision variable.
Performance measures the production rate compared to rated In equation (3) given above, if the real cycle time is
speed. Performance indicates the operator’s performances minimized, then the speed operating rate increases, so the
and the way the equipment is used: value of the OEE goes towards a maximal value. The
augmentation of the speed operating rate value is possible
Performance%  speed operating rate  net operating rate  100 only if the cycle time of the line is minimized, otherwise this
(2) aim can lead to bottlenecks. Hence, the idea is to balance the
line and to minimize the real cycle time so as to maximize the
Where OEE indicator and the efficiency of the line.
The objective function of present optimization problem is:
standard cycle time
speed operating rate  100 (3)
real cycle time F (t )  min (real cycle time)

number of products  real cycle time In (Duta, 2006a; Duta et al, 2008), the cycle time is the
net operating rate   100 (4) maximal value of the sum of operational times:
actual run time
Quality is given by the percentage of good product produced.
Quality indicates the capability of the process:
tcy  max

j( tasks on Wi )
tij (7)

Introducing the decision assignment variable, the objective

quantity produced right function becomes:
Quality %   100 (5)

   X (i, j)  t
quantity produced
F (t )  min(max ij
) (8)
j( tasks on Wi ) i

INCOM 2012, May 23-25, 2012
Bucharest, Romania

Some terms in brackets before could be 0 depending on the The line is supposed to work in continuous flow, operational
value of the decision variable X(i,j). times are deterministic and production is of flow shop type.
Another assumption is that the line is never starving: there
The minimum value of this function is obtained when the
are always products to manufacture on the line.
manufacturing line is balanced (Duta et al, 2006b). The
effective assignment matrix Y is derived from matrix X and is
obtained after simulation running.
The optimisation problem consists in calculating the minimal
cycle time from the objective function F (equation (8)) taking
into account the four linear constraints below. The input size
of the problem is the length of a binary data representation.
a The non-divisibility constraint : Every task is performed on
a single workstation. A task cannot be divided between

m (9)
 Y (i, j)  1
i 1
j  1..n

b. The precedence constraint

Y (i, j )  X (i, j )  j  1..n,  i  1..m (10)

Fig. 1 Input Excel file
In our study m=5, n=10.
c. Time constraint:cycle time is an upper bound on the The possible assignment matrix is taken from (Mainea et al,
workload assigned to each workstation. 2010) and it fulfills the precedence constraint:

11 0 1 0 0 11 0 0 
n (11)  
 (t ij  Y (i, j ))  t cy  i  1..m  0 1111 0 0 1 0 0 
j 1 X  1 0 11 0 1 0 11 0 
 
d. Real assignment constraint
 0 0 111 0 1111 
 0 0 0 1 0 11 0 11
 
Y (i, j ) {0,1}  i  1..m,  j  1..n (12) Operational times (in minutes) are given bellow:
Task 1 2 3 4 5 6 7 8 9 10
T[n] 38 10 8 26 13 18 25 9 15 30
Minimum standard cycle time is also given and is equal to
The proposed method was implemented at the Romanian
38.4 min.
enterprise ARCTIC Gaesti, a domestic appliances
manufacturer. 1 The proposed solving method is to apply integer
programming on a linear function (8), with linear constraints
Here, the TPM concept was successfully implemented since
(9), (10), (11), (12). To run simulations, the FICO Xpress
2004. This aim was accomplished with the help of a Japanese
Optimization Suite 7 Software was utilized (FICO Xpress,
consultant. The system’s concept is “zero losses” and it
2009). This is a mixed integer programming optimizer which
correlates the augmentation of the production capacity with
has been programmed to handle a broad range of
the costing diminution (Mainea, 2006). The values of the
optimization problems. His quadratic module allows the
necessary parameters to calculate the OEE are registered on
optimization of a quadratic function. The main advantage of
line in Excel files as the one presented in figure 1. Parameters
this software product is that the user works in the Console
are considered input data for our optimization method. The
Mode and can modify the code of the program to suit data of
line has five workstations performing ten different
the problem.
operations, whose operational times are given below.
The XPRESS optimizer uses Branch and Bound technique to
Acknowledgements to Romanian enterprise for information solve mixed integer programming problems.

INCOM 2012, May 23-25, 2012
Bucharest, Romania

The relaxed problem is a linear programming problem and Table 3 Tasks assignment after balancing
can be solved by exploring the tree of solutions using the cut-
off value method. When a better value of the solution is found Wi 1 2 3 4 5
in a solution node, this can act as a cut-off for outstanding
nodes (Dash, 2007). Simulations performed with FICO Task 1 2, 4 3, 6, 8 5, 10 7, 9
XPRESS optimizer give the number of iterations and cuts
made in the solution space.
This effective assignment fulfills constraints from (9) to (12).
After simulations, the minimum value of the real cycle time
After balancing the line, new values for the workstation’s
is obtained: t cy
 43 min. operational times are obtained (Table 4):
The obtained effective assignment matrix is:
Table 4 Operational times after balancing the line
1 0 0 0 0 0 0 0 0 0  Wi 1 2 3 4 5
 
0 1 0 1 0 0 0 0 0 0 38 36 35 43 40
Y  0 0 1 0 0 1 0 1 0 0
 
0 0 0 0 1 0 0 0 0 1 One can notice that the line is better balanced and the cycle
0 0 0 0 0 0 1 0 1 0
  time is minimized.

Simulations were performed on an Intel Core 2Duo T7500 Replacing the values of standard cycle time and real cycle
2.2 GHz Processor with a RAM of 2 GB. time in equations (1) to (4), one can calculate the new values
of availability and performance for each workstation (Table 5
6. RESULTS and Table 6).

In Table 1data before simulation is presented. Table 5 Workstations availabilities after balancing

Table 1 Initial TPM indicators values on workstations in [%] Wi 1 2 3 4 5

are taken from the Excel file (fig. 1)
A[%] 86,84 80,55 77,14 100 92,5


Table 6 Workstations performances after balancing


Wi 1 2 3 4 5

P[%] 88,37 83,72 81,39 95,44 93

19,37 69,22 100,00 13,41
80,72 86,43 92,92 64,82 Further, these results were used to calculate the new values of
the OEE indicator. Final results are presented in Table 7.
78,86 74,92 92,61 54,71
97,49 90,00 97,53 85,57 Table 7 New improved values of the TPM indicators [%]
31,57 65,00 100,00 20,52


The sum of operational times on each station is deterministic


and known and is given in Table 2.

Table 2 Initial operational times for each station (min)
Wi 1 2 3 4 5
86,84 88,37 100,00 76,74
38 44 38 33 39
80,55 83,72 92,92 62,66

In the previous paragraph, the value of the real cycle time 77,14 81,39 92,61 58,14
was obtained after minimizing the value of the function F
100 95,44 97,53 93,08
from the equation (8). The effective assignment of the tasks
after simulations is given by matrix Y and also in Table 3. 92,5 93 100,00 86,03

INCOM 2012, May 23-25, 2012
Bucharest, Romania

One could notice that, after applying the proposed method, Figure 4 is a print screen of TPM indicators displayed in real
OEE and machine performances are maximized and are time by Doruk system2. Different OEE values are shown on
almost equally distributed on the line (Fig. 2 and 3). That the screen. They are displayed in the same time on
means that workstations are equally loaded, thus increasing computer’s display and on the monitoring panels from the
the efficiency of the line. workshop walls.

Fig. 2 Initial values of TPM indicators

Fig. 4 Screen of the monitoring system

TPM is a new manufacturing approach implemented in
Romania and the results in increasing the productivity are
over expectations.
An optimization method for the Overall Equipment
Effectiveness maximization, the most important indicator of
the Total Productive Maintenance, is presented in this paper.
The method is based on the line balancing accomplishment
after minimizing the real cycle time using a linear
programming model. The originality of the approach is the
presence of decision variables in the objective function as
well as in model constraints. After performing simulations on
the mathematical model, TPM indicators values are
Fig. 3 TPM indicators after balancing the line
significantly improved.
Results of the above optimization methodology are visualized
The results were implemented at ARCTIC Gaesti enterprise
on the monitoring system and presented on screens like that
which is in present in the third stage of the TPM
from figure 4.
implementation concept.
A computerized system monitors the work on line and
Further research will take into account the multicriterial
displays the values of OEE for each workstation in real time.
character of TPM indicators optimization.
Integrating our method in the informational system gives the
possibility to control tasks assignment on workstations and to
increase OEE values to almost 100% which is the maximal
performance of the equipment.

Doruk Automation Company, Turkey is an industrial
equipment supplier

INCOM 2012, May 23-25, 2012
Bucharest, Romania


Dash A. (2007), Xpress Optimizer Reference Manual, Dash

Optimization Ltd., Canada
Duta, L. (2006a) Introduction dans l’étude des systèmes de
désassemblage – Programmation et contrôle, Expert
Publishing House Bucharest (in French language)
Duta L, Filip F G., Henrioud J.M. (2006b), Applying Equal
Piles Approach to Disassembly Line Balancing Problem,
ELSEVIER, Vol 16, pp. 152-157
Duta L, Filip F. G., Caciula I. (2008): Real Time Balancing
of Complex Disassembly Lines, Proceedings of the 17th
IFAC World Congress, ELSEVIER, Volume# 17 | Part#
1, pp 913-91
JIPM (2002), TPM Encyclopaedia, Japan Institute of Plant
Mainea M., (2006) Quality Management. Design,
implementation and development of the quality system in
industrial enterprises, PhD Thesis, Polytechnic
University of Bucharest
Mainea M., Duta L, Patic P. C., Caciula I., (2010) A Method
to Optimize the Overall Equipment Effectiveness,
Manufacturing Control, Production and Logistics
Conference MCPL 2010, Coimbra, Portugal
McCarthy D, Rich N, (2004), Lean TPM – a Blueprint for
Change, Elsevier Butterworth-Heinemann
Patic P.C., Pascale L., (2010) CAD in Industrial Application,
Bibliotheca Publishing House, Romania
Rich, N. (2002) Turning Japanese, PhD Thesis, Cardiff
Tajiri M., Gotoh F. (1999), Autonomous Maintenance in
seven steps, Productivity Inc., USA
Venkatesh J (2006)A Introduction to Total Productive
Maintenance (TPM),Plant Maintenance Resource
FICO Xpress Optimization Suite 7, (2009)