UTILIZATION OF KEY PERFORMANCE INDICATORS IN PRODUCTION

CONTROL

Vladimir Jovan1, Sebastjan Zorzut1, Alenka Žnidaršič2

1
Jozef Stefan Institute, Jamova 39, SI-1000, Ljubljana, Slovenia
2
Metronik d.o.o., Stegne 9a, SI-1000 Ljubljana, Slovenia

Abstract: A synthesis of plant-wide control structures is recognized as one of the most

important design problem in production management in process industries. This paper
proposes a closed-loop control structure with the utilisation of production Key
Performance Indicators (KPIs) as possible solution to this problem. Production KPIs
represent the translation of operating objectives, such as minimization of production
costs, to a set of measurable variables that can then be used in a feedback control. The
idea of production feedback control using production KPIs as referenced controlled
variables was implemented on a procedural model of a production process for a
polymerisation plant. Some preliminary results show usefulness of proposed
methodology. Copyright © 2006 IFAC.

Keywords: Production control; Productivity, Decision support systems; Closed-loop

control, Model-based control

1 INTRODUCTION subjects require knowledge of management of their

The demands for efficient and flexible production
have established the importance of an autonomous
production management level acting as a kind of
interface between requests from the business
management level and factory-floor production
control (Jovan, 1999). At the production
management level, at least two essential activities are
performed: (i) transformation of a company’s
objectives into results and (ii) optimisation of
production. To fulfil these two basic tasks
successfully, a production manager's decisions must
rely on accurate and on-line information.
In 2001, Forza et al. (2001) discussed the need for
information flow and redistribution of management
responsibility among all management structure
entities in order to achieve highly efficient
production. The authors also claim that management
own object as well as relevant and up-to-date data on
current production. Only in this way and together
with the support of appropriate information
technology, can the production process be managed
successfully. Since 2000, the first research results on
Decision Support Systems (DSS) for the production
management level have appeared. Vicens et al.
(2001) discuss and propose a methodology for the
conceptual design and implementation of a
production DSS, and place this system in the context
of the overall enterprise management structure.
Ahmad et al. (2002) have already defined the
principal measurements used to indicate current
short-term production efficiency. These measures
include the financial indicator, technological
indicator and efficiency indicator. The authors’ main
conclusion is the proposal that these indicators be
taken as the basis for production efficiency
estimation. In the last two years, articles describing
implemented DSS have also appeared. However, the
production management level functions are covered
only partially (e.g. production quality, energy
consumption). In one of these articles the emphasis is
on the implementation of the DSS prototype
(Lohman et al., 2003) using the Balanced Scorecard
Methodology (Kaplan and Norton, 1992; Kaplan and
Norton, 1996).
Generally speaking, the production manager is
usually overloaded with current production process
data during his decision making process with the
main problem being how to extract the relevant
information from this vast amount of data in order to
make fast and correct decisions. One promising way
of solving this problem is the introduction of
production Key Performance Indicators (KPIs) as a
means of reducing the amount of data down to the
most important information on current production
process status. A further step is utilising production
KPIs in closed-loop control scheme for the whole
production plant. To test the usefulness of the
proposed methodology, a model of production
process and production KPIs for a polymerisation
production plant has been developed.
These topics are elaborated in the paper and
organised as follows: Section 2 briefly describes the
basics of KPI methodology; in Section 3 a closed-
loop paradigm of production management is
explained; Section 4 describes the proposed concept
for utilising KPIs in production management in
process industries using the case study of a
polymerisation production process, together with
some preliminary simulation results; Section 5
presents the conclusions.
2 KPI METHODOLOGY
A performance indicator can be defined as “a
variable that quantitatively expresses the
effectiveness or efficiency, or both, of a part of or a
whole process, or system, against a given norm or
target” (Lohman et al., 2003).
A production process involves several business and
technical activities on and around the factory floor.
Its effectiveness can be assessed using information
hidden in a set of current and historical production
data. The problem of extracting the relevant
information from production data for fast and
accurate decision-making can be solved by
introducing a set of production KPIs that show the
operational and mid-term efficiency of the
production. On the strategic management level, the
problem of overall business efficiency in a
production factory is already being solved with this
approach (DeBusk, 2003), while on a production
management level the implementation of KPIs is a
rather new concept. The solution lies in defining an
appropriate set of KPIs that are specific to the
observed production process, and in defining the
strategy for using KPIs to efficiently manage that
process. Recently, a balanced set of general KPIs for
the production management level has already been
introduced (Rakar et al., 2004) and five principal
KPIs for process-oriented productions were defined:
Safety and Environment; Production Efficiency;
Production Quality; Production Plan Tracking; and
Employees Issues.
3 CLOSED-LOOP PRODUCTION
MANAGEMENT PARADIGM
Production is a complex process, consisting of
several operations, interconnected by material,
energy and information flows and restricted by a
range of time, organisational, technological and other
constraints.
The production process must deliver products that
meet customer specifications consistently, while
simultaneously respecting the imperative of being
profitable. The objectives of production management
in process industries are both the effective utilisation
of production resources and maintaining stable
production within given constraints. The weakness of
the production manager’s role is in the subjective
perception of global production aims and also in the
vast amount of data that are not properly classified
according to their importance to the decision making
process.
The desired global production objectives in the
context of production management system can be
more objectively defined as the reference values for
significant measures of plant efficiency, production
plant productivity, mean product quality and others.
These production objectives are often called implicit
objectives as they usually can be expressed only
implicitly as functions of the measurable and
manipulatable variables (Stephanopoulos and Ng,
2000). Since implicit objectives are not directly
measurable, their translation into a set of output
production process variables should be provided.
These output production process variables should
have the following properties (Skogestad, 2004):
(i) They should be more easily measurable,
(ii) It must be possible to handle maintaining their
set-point values by proper adjustments of
manipulatable production process variables, and
(iii) When maintained at the desired optimal set-
points through the feedback control subsystem, they
should inherently contribute to the overall
profitability of a production process.
These variables are denoted in this paper as
“production KPIs” (see Figure 1).
 Economical Technological Energy and
issues constraints Disturbances raw materials

WHAT
Production Operator
WHEN manager PRODUCTION Productivity
using
MANAGEMENT PROCESES
HOW MUCH using DSS + DSS Quality
LEVEL - KPI
Costs
+
-

Fig. 1. Closed-Loop Control Structure of the Production Process

This production control problem can be 4 THE CASE STUDY

mathematically formulated as in (Stephanopoulos
and Ng, 2000): 4.1 Polymer Emulsion Production Process

Let zPROFIT represents the operating profit: The polymer emulsion batch production process is a
typical representative of process-oriented production
z PROFIT = ∑ product values − ∑ raw material costs − where production effectiveness to a great extent
− ∑ production costs . relies on the quality of the production control system.
The production layout consists of several reactors,
dosing vessels, storage tanks and equalizers that are
The feedback-based solution leads to the following used for the production of different products. The
optimisation problem: technological process is defined with a recipe, i.e. the
sequence of operations that have to be performed for
Minimize z PROFIT,sp − z PROFIT = the production of particular product.

min z PROFIT (u, y, d) − z PROFIT (u, y, d) Installed DCS and SCADA systems handle the
u fb , y fb ,sp,C(.) safety, regulatory control and monitoring functions
successfully but often the non-uniform quality of raw
subject to materials, chemical reactions involved, process
down-time due to failures, prolonged operation at
y = h(u, y, d) plant dynamics non-optimum points, long periods of switch-over
g (u, y , d) ≤ 0 process constraints from one mode of operation to another, prolonged
operation with off-spec products, mismatch between
d∈D process disturbances the business production plans and those achieved by
U ≤ u fb ≤ U the plant make the final product quality, yield and
duration of a single batch variable, which renders the
u fb = C(y fb ).
entire production process non-optimal.
where, ufb and yfb denote the manipulations and
measurements involved in the feedback-based The production proceeds in successive batches on
solution and C is the structure of the entire different equipment where on each batch stage
production control system. intermediate products appear and have to be used in
successive stages as soon as possible. In each step
For the case study production process presented in some physical actions (heating, blending) or
this paper the chosen process variables (production chemical reactions are involved that have a
KPIs) are Productivity, Product Quality and significant influence on final product quality. If
Production Costs. None of them are directly production speed is increased, some of the phases
measurable but an estimation of their current values need to be shortened which is usually reflected in
can be made using the combination of directly decreased product quality. If the quality of raw
measurable production process variables. material is low or variable or if the production
process is not stable (due to energy failures,
Maintaining the predefined set points for these two inadequate regulation) then the quality parameters of
variables can be achieved by the proper adjustment the product achieved may not satisfy the prescribed
of some process variables (e.g. quality of raw quality requirements and the product may need to be
materials, production speed, production schedule) recycled in subsequent batches or eliminated.
that are direct inputs in a case study production
process.
A procedural model of the case study production
process has been developed to facilitate
experimenting and the verification of the closed-loop
control structure. The model was designed in the
academically established Matlab, Simulink and
Stateflow simulation environments. The simulated
data are stored in the MS Access database and are
available for different on-line or off-line processing.
4.2 KPIs for Polymerisation Production Process
Given the final objective of stabilising the existing
production process, a promising idea is to introduce a
closed-loop production management concept so that
specific production KPIs serve as the reference
values for the closed loop production control system.
It is hypothesised that such an approach can
contribute to more stable production and better final
product uniformity and quality. Three production
KPIs were chosen to characterise the case study
production process:
Productivity (also denoted as actual production rate
or production yield). Productivity is defined for the
described production process as the amount of all
products that were produced over a set production
period. All batches finished within the set time
window (production period) are taken into account
and the average amount of products produced in an
hour is calculated.
Mean Product Quality. The Mean Quality KPI is
calculated as the mean value of the quality factors for
the batches completed in the set time window.
Mean Production Costs. Production costs consist of
raw material costs, energy costs, other operating
costs and fixed costs in the set time window. The
mean production costs are calculated as the sum of
all production costs within a time window, divided
by the amount of all products produced in this time
window.
These three KPIs represent the output (controlled)
variables on Figure 1. Maintaining the predefined set
points for these KPIs is achieved by properly
adjusting the manipulated (input) variables, which
are in this case: Raw Material Quality, Production
Speed and Batch Schedule.
Determining the influence of the input variables and
disturbances on the output variables (selected KPIs)
is essential for efficient production control. For
example, when the current Productivity KPI is low,
what can be done to increase it? If the reason lies in
diverse product assortment, a better production
scheduling can be applied. If the production speed is
low, it can be increased. If the Product Quality KPI
value is low and there is a need to improve it than the
quality of raw materials should be increased or
production speed should be decreased or more
frequent cleaning of the reactors may be required. To
decrease the Production Costs KPI, cheaper raw
materials can sometimes be used or production
equipment and employees better utilised through
more effective batch scheduling.
4.3 Simulation Runs
To see how the production speed and the quality of
raw materials affect productivity, product quality and
production costs, a simulation run was performed.
Simulation run was divided in three phases, where
the production speed was changed from low to
normal and finally to high. On Figure 2 the first
phase lasts from 0 to 450 hours of simulation time,
the second phase from 450 to 870 hours and the third
phase from 870 till the end. In each phase the quality
of the raw materials was subsequently increased
from low to very high (0.85 to 1.1). The KPIs for
Product Quality, Productivity and Production Costs
were observed. The KPIs were evaluated every 12
hours for the time window of 120 hours.
Figure 2 represents the response of the Product
Quality KPI. In the first phase of the simulation run
the production speed was low, which represents the
best working conditions in the production process.
Over the simulation run the quality of raw materials
gradually increased and the influence of this change
on the Product Quality KPI can be observed. As
expected, better quality raw materials contribute to
better quality final products.
In the second phase of the simulation run the
production speed is increased, which usually leads to
decreasing production process quality. Raw material
quality changed in the same manner as in the
previous phase: from low to high. The Product
Quality KPI first decreased and when it reached the
bottom it started increasing as in the first phase. The
change in product quality is momentary but the
change in the Product Quality KPI is gradual. The
Raw materials quality and production rate
1.5
1
0.5 Raw materials quality
0
Production rate
0 200 400 600 800 1000 1200
Mean product quality for the time window: 120h
1.5
Mean product quality
1
0.5
0
0 200 400 600 800
Production time(h)
1000 1200
Fig. 2. Response of the Mean Product Quality KPI to
Raw Material Quality and Production Speed
 Raw materials quality and production speed Raw materials quality and production speed
1.5 1.5

1 1

0.5 Raw materials quality 0.5 Raw materials quality

Production speed Production speed
0 0
0 200 400 600 800 1000 1200 0 50 100 150 200 250 300 350 400
Productivity for the time window: 120h Mean product quality for the time window: 75h
1600
1.2
1400

Mean product quality

1
1200
Productivity (kg/h)

1000 0.8

800 0.6
600
0.4
400
0.2
200

0 0
0 200 400 600 800 1000 1200 0 50 100 150 200 250 300 350 400
Production time(h) Production time (h)

Fig. 3. Response of the Productivity KPI to Raw Fig. 5. Open-Loop Control of the Product Quality
Material Quality and Production Speed KPI.

KPI evaluation algorithm averages the product Figure 4 represents the mean Production Costs KPI.
quality in the set time window, which can be seen in The production costs per product unit increase with
this section of the figure. The second interesting the increasing quality of raw materials. There is a
phenomenon is the influence of production process slight increase in costs with increasing production
quality on product quality. The mean quality of the speed.
products decreased by about 10 %. In the third phase
the pattern repeated. 4.4 Open-Loop Control of a Production KPI

Figure 3 represents the Productivity KPI for the same The following simulation run presents the open-loop
simulation run. The Productivity KPI has slightly control of the Product Quality KPI. The experiment
increasing trend with higher production speed. At the represents the execution of a normal schedule of
beginning of the third phase of the simulation run production jobs using raw materials with normal
high production speed causes the significant decrease quality at normal production speed. After a certain
of productivity. This is the result of coincidence that time period, a disturbance occurs in the form of a
the quality of the production process and also the decrease in the quality of raw materials, which is
quality of the raw materials are low. Consequently, reflected in the considerable decreased value of the
some batches do not attain prescribed quality mean of the Product Quality KPI (see Figure 5).
requirements and they have to be recycled, what
affects Productivity KPI. The appearance of off-spec As an open-loop control action the production
batches can be noticed on the Figure 3 (time period manager then slows down current production speed.
from 900 to 950 hours) where the reactor occupancy The quality of both the production process and final
for some batches is increased due to the need of product gradually increase, and consequently this is
recycling of bad batch entering in the production reflected in the increase in the mean value of the
process. Product Quality KPI. This is not the only possible
action that production manager could take, but in the
Raw materials quality and production rate presented case it was sufficient to eliminate the
1.5
disturbance.
1

0.5 Raw materials quality

0
Production rate 5 CONCLUSIONS
0 200 400 600 800 1000 1200
Mean product cost for the time window: 120h
The ideal plant-wide control system should ensure
200 that the production process is constantly working in
Mean product cost (Sit/kg)

an optimal manner. As a result of the plant-wide

150
focus, a plant-wide control problem possesses certain
100 characteristics that are not encountered in the design
of control systems for single units, such as the
50
following (Stephanopoulos and Ng, 2000):
0
(i) The variables to be controlled by a plant-wide
0 200 400 600
Production time(h)
800 1000 1200
control system are not as clearly or as easily defined
as for single units;
Fig. 4. Response of the Mean Production Costs KPI to (ii) Local control decisions, made within the context
Raw Material Quality and Production Speed of single units, may have long-range effects
throughout the plant;
(iii) The size of the plant-wide control problem is
significantly larger than that for the individual units,
making its solution considerably more difficult.
Despite evident progress in this research field
(Stephanopoulos and Ng, 2000; Skogestad, 2000;
Skogestad, 2004) the design of an optimal
production control system is still an unreachable
objective. This paper proposes an approach to
measuring and presenting the attainment of
production objectives in the form of introducing
production KPIs. Using the KPI approach the
implicit production objectives were translated into
measurable values that can be extracted from
existing production data. In this way the production
control concept and the role of a production manager
are slightly changed; instead of monitoring and
controlling several tens and hundreds of process
variables at a low production level, a production
manager monitors and controls only a few major
production KPIs with the aim of achieving the most
important implicit production objectives, e.g. high
product quality, high productivity and minimal
production costs.
The procedural model of the case study production
process has been developed and used in a number of
simulation runs. The preliminary simulation results
presented indicate that this work could evolve
towards the implementation of a production KPI-
based control system in a real industrial plant. The
intention in future is to improve the existing
production process model, validate it rigorously and
incorporate it into a Decision Support System for
production control in the polymerisation plant that
was used as the case study production process in this
paper.
6 ACKNOWLEDGMENT
This work was supported in part by the 6 RTD th
Framework Programme under Contract MRTN-CT-
2004-512233, project title: PRISM – Towards
Knowledge-Based Processing Systems and by
Slovenian research agency under Contract 3211-06-
000023 .
REFERENCES
Ahmad M. M., Dhafr N. (2002): Establishing and
improving manufacturing performance
measures, Robotics and Computer Integrated
Manufacturing 18, Pergamon Press, pp. 171-
176.
DeBusk, G. K., Brown, R. M., Killough, L. N.
(2003): Components and relative weights in
utilization of dashboard measurement systems
like the Balanced Scorecard, The British
Accounting Review, Volume 35, Issue 3, pp.
215-231.
Forza C., Salvador F. (2001): Information flow for
high-performance manufacturing, Inter. Journal
of Production Economics 70, Elsevier Press, pp.
21-36.
Jovan, V., (1999): The Relationship between
Planning and Production Activities in Process
Industries, Proceedings of the 7th
Mediterranean Conference on Control and
Automation, June 28-30, 1999, Haifa, Israel,
1999; pp 1982-1989.
Kaplan, R. S., Norton, D. P. (1992): The Balanced
Scorecard - Measures that drive performance,
Harvard Business Review, January–February,
pp. 71–79.
Kaplan, R. S., Norton, D. P. (1996): Translating
Strategy into Action: The Balanced Scorecard,
Harvard Business School Publishing
Corporation, Boston
Lohman, C., Fortuin, L., Wouters, M. (2004):
Designing a performance measurement system:
A case study, European Journal of Operational
Research 156, Elsevier press, pp. 267-286.
Rakar, A., Zorzut, S., Jovan, V. (2004): Assesment
of production performance by means of KPI,
Proceedings of the Control 2004, 6-9
September, 2004, University of Bath, UK.
Skogestad S. (2000): Self-optimizing control: the
missing link between steady-state optimization
and control, Computers and Chemical
Engineering 24, pp. 569-575.
Skogestad S. (2004): Control structure design for
complete chemical plants, Computers and
Chemical Engineering 28, pp. 219-234.
Stephanopoulos G., Ng C. (2000): Perspectives on
the synthesis of plant-wide control structures,
Journal of Process Control 10 (2000), pp 97-
111.
Vicens, E., Alemany M. E., Andres C., Guarch J. J.
(2001): A design and application methodology
for hierarchical production planning decision
support system in an enterprise integration
context, Int. Journal of Production Economics
74, Elsevier Press, pp. 5-20.