Vous êtes sur la page 1sur 5

The following paper was the basis for a presentation at the International Conference on Modeling

and Analysis of Semiconductor Manufacturing (MASM), held at Tempe, Arizona, USA. in 2002.




FACTORY LEVEL METRICS: BASIS FOR PRODUCTIVITY IMPROVEMENT




John P. Dismukes
Manufacturing Value Chain Science Center (MVCS Center)
The University of Toledo
Toledo, OH 43606




KEYWORDS

Overall Equipment Effectiveness (OEE), Unit
Production Process (UPP), Overall Throughput
Effectiveness (OTE), Total Productive
Maintenance (TPM), Theory of Constraints
(TOC)

ABSTRACT

Productivity of a manufacturing line made up
of multiple equipments is measured, in many
cases, as productivity of the bottleneck
equipment given in terms of overall equipment
effectiveness (OEE). For this to hold true, some
limitations apply such as having a series layout
of production equipments operating at fixed
rates, i.e. single product recipe. Manufacturing
lines in many industries are not that simple since
multiple parts are produced through various
architectural combinations of equipments. In
such cases productivity is greatly impacted by
material flow between equipments and by the
interdependence of equipment throughput. This
paper introduces a methodology, based on newly
developed factory level productivity metrics, for
analyzing manufacturing system productivity as
a function of OEE and material flow. The
metrics serve as a foundation for a closed-loop
productivity improvement methodology, which
utilizes several concepts and theories such as the
Theory of Constraints. Application of the
methodology to a real production line is
presented through a case study in flat glass
window assembly.

I. INTRODUCTION

One of the major concepts of TPM
(Ljungberg 1998) is maximizing asset utilization
or equipment effectiveness by gauging its
productivity and focusing improvement tasks to
eliminate its losses. Bottleneck equipments are
usually the first to be involved in the TPM
program.

Another theory that follows the same line of
thought is the Theory of Constraints (Dettner
1997), which suggests that physical constraints
arise due to bottleneck equipments, and that
eliminating losses at a bottleneck will
significantly enhance productivity. TOC also
suggests that non-bottleneck equipments do not
impact productivity to the same extent as
bottlenecks.

Conclusions from both theories, TPM and
TOC, ignore the interdependent relationship
between production steps. Although the idea of
insignificant contribution by non-bottlenecks
holds when compared to gains from focusing on
bottleneck equipments, the impact of production
disturbances is producing greater impacts as a
result of the transition of manufacturers to
continuous production. Single part flow or
continuous production as introduced in lean
manufacturing concepts (Womack 96) is
increasingly gaining positive reputation as the
solution to Just in Time manufacturing.
Furthermore, although disturbances in non-
bottlenecks are known to have minimum if not
negligible impact on throughput, it certainly has
a great impact on cost per unit of product as

determined via activity based costing techniques
which are beyond the scope of this paper.

Manufacturers are increasingly focusing on
equipment productivity management to meet
their profitability goals. But with dynamic
production environments in which bottlenecks
are continuously shifting based on product
demand, the choice of the equipments to be
analyzed change with change of demand. A
systematic methodology is required to manage
the analysis of such production systems.


II. SYSTEM LEVEL PRODUCTIVITY

In order to take all equipments into
consideration when analyzing a production
system, it is required to use a system level metric
(Razzak 2001; Su et al. 2002) analogous to OEE,
and designated as Overall Throughput
Effectiveness (OTE). OEE is the ratio of actual
good parts produced by a UPP to the theoretical
number of parts the UPP is capable of producing
during a certain time period,

Time Total in UPP by oduced Parts of Number l Theoretica
Time Total in UPP by oduced Parts Good of Number
OEE
Pr
Pr
!
(1)

Similarly, OTE is the ratio of good parts
produced by a system to the number of parts the
system theoretically could produce during a
certain time period,

Time Total in System by oduced Parts of Number l Theoretica
Time Total in System by oduced Parts Good of Number
OTE
Pr
Pr
!
(2)

System level metrics are determined as a
function of material flow between UPPs. OTE
of a system is calculated first by calculating OTE
of sub-systems formed by grouping UPPs
according to material flow in predetermined
group structures similar to those in Figure 1.
Sub-system groups are combined further in the
same fashion until the system is reduced to a
single unit, for which an overall value of OTE
results (Razzak 2001; Su et al. 2002).

Calculating system productivity in terms of
OTE provides robust metrics that are very useful
especially when multiple part types are produced
by the same system. A dynamic production
system is one that is producing multiple part
types each with a different recipe of production
such as seen in the semiconductor industry. In
these environments bottlenecks are continuously
shifting from one equipment to another. Non-
bottleneck equipments may not have a direct
impact on throughput especially when space is
allocated for building a limited level of
inventory.

Figure 1. Generic Sub-system Combinations





III. CLOSED LOOP IMPROVEMENT
METHODOLOGY

The closed loop improvement methodology
presented in this paper (Razzak 2001) utilizes
some of the concepts exhibited in the teachings
of Total Productive maintenance (TPM) and the
Theory of Constraints (TOC). It provides a
powerful tool for managing productivity
improvement activities. The improvement
methodology is a systematic process comprising
12 steps schematically represented in Figure 2.
The following is a summary of each step:

1. Flowchart system architecture:
" Allocate beginning and ending UPPs to
specify system boundaries.
" Layout all UPPs and identify their
interconnectivity.
" Recognize sub-systems based on
material flow.

2. Define productivity parameters:
" Determine production characteristics of
each UPP (ideal processing rates, buffer
levels, production capacity, and
transportation rates to and from UPPs).

" Assign time intervals for measurement
frequency.

3. Apply productivity metrics:
" Prepare calculation algorithms such as
in a spreadsheet.
" Enter data from step 2.

4. Collect data:
" Obtain data from production reports
(manually or electronically).
" Measure or estimate actual data missing
from reports.

5. Determine productivity:
" Determine productivity of each UPP.
" Determine productivity of sub-systems.
" Determine system productivity

6. Identify bottleneck and associated losses:
" Identify system bottleneck.
" Evaluate bottleneck productivity by
analyzing A
eff
, P
eff
, and Q
eff
.
" Determine losses at bottleneck.

7a. Identify upstream losses:
" Analyze losses occurring at all UPPs
upstream from the bottleneck.
7b. Identify downstream losses:
" Analyze losses occurring at all UPPs
downstream from the bottleneck.

8. Identify constraint:
" Based on information from steps 6, 7a
and 7b determine the constraint
responsible for the status of system
productivity.

9. Decide how to relieve constraint:
" Use simulation to assess various
improvement scenarios.

10. Manage improvement decision:
" Determine priorities of improvement
tasks based on simulation results. This
step involves the organizational culture
involvement in exploiting the
improvement tasks by subordinating
everything else in the system to
accomplish productivity improvement.

11. Eliminate constraints:
" Execute improvement tasks.

12. Find new constraints:
" Repeat the process for continuous
improvement.

Repeating the cycle takes one of two different
paths depending on the improvement
implemented. If no changes are required in the
system design then the cycle may be repeated
from step 6. On the other hand, if improvement
introduces a change to the system design the
cycle may be repeated from step 1 or 2
depending on a change in the system layout or
productivity parameters.


Figure 2. Closed Loop Improvement Methodology



IV. CASE STUDY


A case study was performed on a
manufacturing system producing a single part
type product, specifically a window subassembly
manufactured by Pilkington, North America, and
sold to a major automotive manufacturer. As
shown in Figure 3, the system involved 15 UPPs
connected in series and parallel; hence the name
SP15. Based on real production data during an
8-hour shift, OEE for each UPP is calculated as
the product of Availability, Quality, and
Performance efficiencies. Results are shown in
Figure 4. Using sub-system algorithms, OTE of
each sub-system is calculated. For example, the
sub-system made up of UPPs M1, B1, and C1
had an OTE of 0.51. Further grouping of UPPs
and sub-systems revealed a system OTE of 0.69.

The manufacturer production schedule
resulted in an OTE of 0.86. Comparing actual
results versus the target limits shown in Figure 5
suggest that the system underperformed by 20
percent.

As indicated in Figure 4, computation starts at
the equipment level then sub-system level and
expands to the system level (left to right
approach). After realizing that the system did
not meet production demand, the analysis phase
is done in a drill-down fashion from the
system level to identify the root-causes at the
equipment level which created the most
significant deviation from target (right to left
approach).

According to the closed loop improvement
methodology (section III), the analysis of SP15
at this stage has covered steps 1 through 5. Sub-
system level OTE results indicate that losses had
occurred in the sub-system groups of M2-B2- C2
and F-G-H-I-J-K. The latter group includes UPP
I, which is identified based on OEE analysis as
the theoretical bottleneck with losses in
performance efficiency due to reduced flow of
input material. Since the former group (M2-B2-
C2) is upstream from the bottleneck, this
suggested that the constraints lay within this
group. Inspecting efficiency results of each UPP
in the group concluded that UPP M2 experienced
significant losses in performance and
availability. Further scrutiny of production data
identified these losses to be due to unscheduled
breakdowns and idle times.

The last steps in the improvement
methodology involve eliminating the constraints
and finally repeating the cycle. Proposals for
eliminating the constraints from this system were
straightforward since the production line is a
simple one relative to other complex lines in
other industries. The improvement tasks
involved adding maintenance jobs for the
equipments during non-scheduled times, and
reviewing operator assignments at the equipment
to minimize idle times.

It should be noted here that in cases when
multiple losses are witnessed during a certain
period of time then eliminating losses follows a
priority list. This becomes even more crucial as
the types of losses are continuously changing in
magnitude and location within the system. In
such cases, losses with the greatest impact on
system productivity are included on top of the
list. To generate the priority list of improvement
tasks simulation models are created. These
simulation models use the actual conditions and
test them against multiple improvement
scenarios to identify their impact on overall
productivity. Refer to Razzak 2001 for a
detailed description of the case study.

Figure 3. SP15 Layout


Figure 4. SP15 Actual OEE and OTE Results





Figure 5. SP15 Target OEE and OTE




V. CONCLUSIONS

Using system level productivity metrics
provides a tool for identifying the root cause of
production losses. This is particularly useful
when multiple part types are produced in various
architectural combinations of material flow.

The case study analyzed in this paper utilized
the systematic closed loop improvement
methodology to identify the losses that caused
the 20 percent negative deviation from
production target limits. Although several other
losses occurred during the production period, the
most significant losses were found to be due to
availability and idle time at a single upstream
UPP.

Since the manual implementation of the
closed loop methodology becomes more difficult
as the system becomes more complex, and the
fast paced production environments require
expediting the decision making process of
improvement tasks, then a computer-based
application is recommended to facilitate this task
and offer ongoing productivity assessment and
improvement. A web-based application utilizing
the methodology presented is currently being
implemented by Effective Metrix, Inc. to
evaluate the utility of the methodology and the
associated algorithms for several clients in the
plastics and composites industries.


VI. REFERENCES

Dettner H. W.; Goldratts Theory of Constraints, ASQC
Quality Press, Milwaukee, 1997.

Dismukes J. P.; Q. Su; S. H. Huang; M. A. Razzak, G. Wang;
and R. Kothamasu, Hierarchical Methodology For
Productivity And Improvement Of Production Systems,
International Patent Convention Application, Filed
December 19, 2001.

Ljungberg O.; Measurement of Overall Equipment
Effectiveness as a Basis for TPM Activities, International
Journal of Operations and Production Management, Vol. 18
No. 5, 1998, pp.495-507.

Razzak M. A.; Metrics For Factory Level Productivity, MS
Thesis, University of Toledo, May 2001.

Su Q.; J. P. Dismukes; S. H. Huang; and M. A. Razzak,
Factory Level Metrics for Manufacturing Productivity,
submitted to IEEE Transactions on Robotics and
Automation, May 2001, revised February 2002.

Womack J. P. and D. T. Jones Lean Thinking, Simon &
Schuster, 1996.


BIOGRAPHY


John P. Dismukes is Professor, Department of Chemical and
Environmental Engineering, and Director, Manufacturing
Value Chain Science Center (MVCS Center) at The
University of Toledo.

Vous aimerez peut-être aussi