Simulation of an industrial process of wind towers manufacturing
Paulo M. N. Tomé
paulo.m.n.t@gmail.com
LAETA, IDMEC, Instituto Superior Técnico, Universidade de Lisboa, Lisboa, Portugal
May 2014
Abstract
Lean Manufacturing and Six Sigma methods are often used to maintain and improve the
efficiency of manufacturing processes. Over the last decades efforts were made in order to use
computational simulation for the same purpose.
Simulation is used to obtain a model of an industrial process for the manufacture of wind towers
according to the system of the Portuguese company TEGOPI – Industria Metalomecânica, S.A.
The wind tower is introduced and simulation techniques are presented. Furthermore the case
study is described and presumptions are made for developing the model of the system.
A sensibility and results analysis is conducted for the more relevant parameters and scenarios.
Keywords: Manufacture, wind tower, simulation, model, system.
1. Introduction
Here the manufacturing process of a
tubular metallic tower is considered and
1.1. Wind tower
modelled using simulation.
With early applications for water pumping
and milling, the use of wind energy dates back
to the 2nd century B.C. [1] with the first wind
mills.
In recent years, having developed greatly
trough the 70’s as consequence of the oil
crisis [2], wind energy plays an important role
in today’s society.
In Portugal, according to REN1 [3], at the
end of 2012 almost 23% of the connected
power was obtained using wind.
As of today, most wind towers in use can
be described by the four major components:
Image 1.1 - Typical modern wind tower with tubular
metallic tower [2]
tower, rotor (hub and three blades), nacelle
and foundation [2].
1
Portuguese energy supplier
1
1.2. Computer
Simulation
and
2. Case study
Simulation Methods
Having received data from the company
According to McHaney [4], computer
TEGOPI in order to model the process, the
simulation can be broadly defined as:
obtained model can be used to help the
“Use of a computer to imitate operations of
company evaluating the effect of altering key
a real world process or facility according to
variables in production time.
appropriately developed assumptions taking
the form of logical, statistical, or mathematical
According to the company, a tower is
relationships which are developed into a
produced from smaller elements (Virolas )
model.”
welded to each other in order to become a
2
3
A model is defined as the representation
section of a tower (Tramos ). The sections are
of a system with the purpose of studying it.
then assembled on sight for obtaining the
Only aspects that influence the problem in
complete tower.
study
must
be
considered
and
then
represented in the model. By definition the
model is a simplification of the real system [5].
It is possible to classify computational
simulation according to three categories:
Monte Carlo Simulation (MCS), Continuous
Simulation
(CS)
and
Discrete
Events
Simulation (DES). MCS uses generation of
random
numbers
to
simulate,
Image 2.2 - Tramos
Image 2.1 - Virolas
without
considering time explicitly as a variable. This
A simplification of the process flow used at
method of simulation is defined by Law and
Kelton [6] as being “a scheme using random
TEGOPI is provided by Image 2.1.
numbers that are used to solve deterministic
or stochastic problems where time plays no
role”. CS models use a set of equations in
Preparation
Virola
representation of a system evolution trough
time. DES is characterized by blocks of time
Assemble
Virola
Assemble
Tramo
Welding
Tramo
Inspection
Tramo
when nothing happens until an event changes
the state of the system.
Image 2.1 - Production flow at TEGOPI
For this study, DES methodology is
considered, once it allows to evaluate system
The considered
entities (Virolas) are
behaviour without solving it mathematically.
assumed
The definition
Therefore the operations of “Preparation” and
of important
concepts as
to
have
an
unlimited
stock.
system, system state, entity, attributes, events
and others can be found in references [5] and
2
Metallic arrows
Section of a tower built from multiple “Virolas” welded
together.
3
[7].
2
Finish
“Assemble Virola” do not interfere with the
production and for this reason they are not
As an example, the simplified flow diagram
modelled.
for the lower section S5 is presented in Image
A tower built from 34 Virola elements
3.1.
The simplification
is due to some
welded to form 5 Tramos sections, according
operations being grouped, as is the case of
to Image 2.2 is considered.
“welding” that represents a sequence of 6
different welding operations.
Since each process requires different
machine and worker resources, in order to be
able to evaluate the impact of changing the
availability of any of them in the model, the
modelled system has:

10 production lines

20 vehicles

26 different resources

12 worker profiles

Around 500 operations
Image 2.2 - Sections of the considered tower [9]
3. Computational model
The production of a subsection consists of
the
operations
of
transportation
of
With the purpose of modelling the studied
the
system
elements to the production line, welding,
in
an
objective
way,
a set
of
presumptions was defined, such as:
inspection, assembling of accessories and
other operations. For each production line,
i.
depending on the section in production,
The production orders are to be
specified for the sections (Tramos)
around 65 to 90 different operations can be
ii.
executed in order to manufacture that section.
The created entities are transported to
the production line by vehicles and workers
Image 3.1 - Simplified flow diagram of the manufacturing process used for section S5
3
simultaneously
iii.
production line locations into consideration.
The vehicles mentioned in ii travel in
After realizing these transportations, the used
specific networks and have a limit for the
vehicles stay idle at the unload point. For
maximum transporting weight
accepting a transportation request a condition
All vehicles
“driven” by two
is evaluated. This condition specifies that the
workers, where the profile of the worker is
chosen vehicle is from the ones that are
depending on the entity type to transport.
available and with a load capacity that is able
iv.
v.
are
At the production line the entities are
to carry the element, but with the lowest
processed according to the array of operations
possible carrying capacity.
specified by the company
vi.
All operations are performed by at least
Whenever a shift is interrupted or
one worker sometimes using resources to
finished, the workers assigned to the running
perform the task.
operations are released
vii.
Having had access to the average times of
At the beginning of a shift where
the different operations only but knowing that
interrupted of pending operations exist, those
this times are not constant, a time distribution
are
is
resumed
according
to
the
priority
associated to the respective entities.
adequate
According
to
to
model
reference
this
[7],
behaviour.
a
triangular
distribution can be used when the actual time
In addition, two specific production lines
distribution is unknown but it is reasonable to
are assigned to the production of each specific
assume that minimum (a) and maximum (b)
section due to weight constraints of the
possible values exist and that the most likely
vehicles.
value (c) is inside this interval. Triangular
Production orders are given
for the
distributions are defined by the following
sections by specifying a start date and a
function:
destination production line. As soon as one
(
element reaches the assembly stand of one
(
production line, the next element of the
( )
production sequence is created and sent to
(
(
{
this same production line. Initial element
)
)(
)
)
)(
)
priorities can be specified for each entity and
destination production lines.
In the model the average value provided
If an order for the production of a section
by the company is considered to be the most
is given and the destination production line is
likely value (c). The minimum and maximum
busy, the model will create the entity, but this
values (a and b) are obtained applying a
entity will not be allowed to leave the stock
percentage deviation around c. The model
area until the destination line finishes the work
considers 5% deviation around c, but it is
in progress. Then becoming available to
possible to change this rule easily by using the
processes the new ordered section.
spread sheets attached to the model.
The model takes transportation times
according to the distances from stock to
4
Multiple profiles 4 of workers are defined
(12
in
total)
according
to
distance (Model*2) for the production of
TEGOPI’s
section S5 in production line 8.
specifications. These profiles are not flexible to
Time (hours)
perform a task from another profile. The same
applies for resources. This implies that an
interrupted operation can only be resumed
when
the needed
worker profile(s)
and
resource(s) are available to attend this task.
Workers and resources “capacities”, costs and
950
850
750
650
550
450
350
250
150
832
752
519 508
301
Total time
reliability logic (the last only for resources) are
245
Stopage time
Model*1
Net time
Model *2
able to be feed easily and changed into the
Graph 4.1 - Production times: section S5 in 8
model through the attached spread sheets.
For the later purpose of model validation
As expected, the net time needed for the
reliability logic is not considered since it is not
production of this section decreases when
mentioned in the supplied data.
average distances are considered. In more
detail, a drop in occupation times of both
workers (PF1 and PF8) and vehicles can be
4. Results Analysis
observed in Graph 4.2
According to Kelton [8], only for the
simplest
models
one
is
able
to
prove
300
categorically the aspects of correct modelling.
250
Time (hours)
Therefore, the model should be tested in a
way that is possible to verify that it behaves in
the desired manner. First the model sensibility
200
150
50
analysed. Afterwards, data provided from the
0
is
used
to
evaluate
111 105
100
to changing key variables is tested and
company
242 236
21.0 9.6
PF1
model
PF8
Model*1
performance. Multiple versions of the model
Vehicles
Model *2
are used to verify this point.
Graph 4.2 - Resources occupation: section S5 in 8
4.1. Sensibility Analysis
In fact, when the complete model is
Hence that transportation of elements from
considered and production orders that lead to
the stock location to the production line takes
the manufacture of 4 towers are given,
a considerable amount of time, Graph 4.1
Model*2 needs about 3 days less of labour for
shows the production time using calculated
finishing the production.
distance (Model*1) or an smaller average
The
order
in
which
operations
are
executed is related to the priority given to the
entities (Virola) in processing. Once that the
4
In this context a welding specialist has a different
profile then a quality controller and so on.
number of resources is limited, situations can
5
occur where not
enough
resources
are
Due date
available to attend to all requested operations.
Model*3
Model*4
Therefore, the priority assigned to entities
1st tower
28-Feb-13
15-Feb-13
plays an import role in the system and by
2nd tower
4-Apr-13
25-Feb-13
3rd tower
10-Apr-13
21-Mar-13
4th tower
30-Apr-13
26-Mar-13
changing this parameter model results should
be affected. According to TEGOPI the priority
Table 4.2 - Limited VS Unlimited resources
is defined to be highest for entities of section
S5 followed by the entities of sections S1, S4,
As expected, the model reacted to the
S3 and therefore, S2 entities were given the
increase of resources by reducing significantly
lowest priorities of the set. This configuration is
the delivery times. Note that for obtaining the
used in both Model*1 and Model*2.
results above the number of workers used is
A new set of priorities was defined for
much higher than the ones scheduled by
Model*3 by keeping the relative priority of the
TEGOPI. Graph 4.3 and Graph 4.4 show the
different sections but increasing priority of the
comparison of used resources.
sections that will form the two first towers. This
Number of Workers
way assuring that priority is given to an older
production order and by doing so a better
approximation to the real system should be
achieved. A summary of due dates obtained
with the models considered up to this point
can be observed in Table 4.1.
60
50
40
30
20
10
0
1st Shift
Due date
th
4 tower
Tegopi
Model*1
Model*2
Model*3
8-May-13
3-May-13
30-Apr-13
2nd Shift
3rd Shift
Model*4
Graph 4.3 – Workforce: TEGOPI VS Model*4
Table 4.1 - Due dates summary
14
12
attention was placed into the allocation of
10
workers to operations. Since specific profiles
8
of workers are used to perform operations on
6
the job-shop using resources of some kind,
4
varying the available number of workers and
2
their profiles should have a major effect on the
0
Arc
ArcoSubmerso
Carro chanfrar
EasyLaser
Lix2
Lix4
Macarico
Rebarbadeira
RF1
Semi
UltraSom
SO213
SO257
SO283
SO294
SO296
In the construction of the model, special
delivery dates. This effect can be observed
through analysis of the results obtained with
Model*4. Model*4 is built from Model*3 but
Tegopi
setting an unlimited number of both workers
Modelo*4
and resources.
Graph 4.4 – Resources TEGOPI VS Model*4
6
4.2. Validation
obtained for machine occupation. Workers
utilization varies up to 4.8%.
300
comparing delivery times of the different
250
scenarios
with
data
provided
from
Time (hours)
Validation of the model is achieved by
the
company.
200
150
100
50
4.2.1.Individual production line
PF1
PF2
PF3
PF4
PF5
PF6
PF7
PF8
PF9
PF10
PF11
PF12
0
With knowledge of the average processing
times of each operation to be performed in
Validation Scenario
Model *2
each section, it is possible to obtain an
estimated average time for the total delivery
Graph 4.6 – Workers occupation: Section S5
time of each section. For this scenario
Time (hours)
unlimited resources are used so that delays
resulting from lack of resources do not occur.
Graph
4.5
to Graph 4.7
show the
occupation of the resources to manufacture
10
Validation
Scenario
9
8
Model *2
7
Total Time
section 5S for the cases: Validation scenario
Graph 4.7 – Vehicles occupation: section S5
and Model*2. Model*2 is chosen because it
considers average distances from stock to
Regarding transportation times there is a
production, thus matches the data better for
15.2% difference between the validation data
validation.
and
Model*2.
around
case
21
Model*1
hours
of
was
total
transportation time would be needed.
Graph 4.8 compares the net production
Time (hours)
time used.
ARC.
Arco-Submerso
Carro chanfrar
Easy-laser
Maçarico
Rebarbadeira
Semi
SO213
SO294
Ultra-sons
VS - Linha8
Time (hous)
considered,
100
90
80
70
60
50
40
30
20
10
0
In
Validation Scenario
850
750
650
550
450
350
250
150
Total time
Model *2
Stopage
time
Validation Scenario
Net time
Modelo*2
Graph 4.5 – Machine occupation: section S5
Graph 4.8 – Production time: section S5
Deviations between 0.2% and 2.1% when
compared with the validation data were
7
4.2.2.Validation scenario
4.2.3.Unlimited resources
Having already verified model sensibility to
A version of the model with unlimited
key parameters, the considered models can
resources and average paths (Model*4) is
now be compared to the expected due dates
used. This allows the model to get closer to
from the validation data.
the data for validation.
Product
ion
Order
1st
tower
2nd
tower
3rd
tower
4th
tower
Base de dados
Due Date
Due Date
Model*
1
Model*
2
Model*
3
Production
Order
Due Date
11-Jan-13
13-Feb-13
15-Feb-13
nd
18-Jan-13
22-Feb-13
25-Feb-13
rd
23-Jan-13
15-Mar-13
21-Mar-13
th
28-Jan-13
19-Mar-13
26-Mar-13
st
11-Jan
13-Feb
11-Mar
26-Feb
28-Feb
1 tower
18-Jan
22-Feb
22-Mar
21-Mar
4-Apr
2 tower
23-Jan
15-Mar
29-Apr
1-May
10-Apr
3 tower
28-Jan
19-Mar
8-May
3-May
30-Apr
4 tower
Table 4.3 – Due dates: Multiple scenarios
Table 4.4 – Due dates: Model*4
Straight away the models seem to contain
For the first and second towers results are
errors. However, it is very important to
matching the expected since the production
consider the following:

lines start free and times are not affected by
According to the supplied data there
delay. For the next towers delay occurs since
are shifts where worker profiles are not
the production lines are occupied when the
available (e.g.: PF6 is not available on the
production order reaches the system. To
third shift). Meanwhile, on the data for
eliminate this effect, a scenario for the
validation this situation appears not to occur
production of towers three and four only was
since there is no indication of delayed
simulated. Results in Table 4.5 show that the
operations as result of missing workers.

Due date
Model*4
due dates in this situation are inside the
In simulation it often occurs, that after
acceptable limits for the model.
the end of a shift the next shift does not have
enough resources to resume all interrupted
Base de dados
operations. Priority operations are resumed
Production
Order
Due Date
Tower 3
23-Jan-13
26-Feb-13
27-Feb-13
Tower 4
28-Jan-13
4-Mar-13
5-Mar-13
first. This can lead to delay due to lack of
resources in low priority operations.

Model does not consider workers to be
flexible.
Table 4.5 – Due dates tower 3 and 4: Model*4
Consider an operation that uses the
worker profile above (PF6) and that is
interrupted in the end of the 2
nd
Due date
Model*4
For achieving these results, Model*4 used
shift. This
145 workers. TEGOPI scheduled 84 workers.
operation can only be resumed after one shift
A situation in which the number of workers
of interruption, affecting the production time
used is higher than the number scheduled is
due to this delay.
verified. The number of workers PF5 of the 1
st
shift is chosen because it represents the
8
4.2.4. Approximation to production
bigger deviation. Four workers are scheduled
by the company and ten were used by
In this version of the model another
BusyTime (h)and nunber of occorences
Model*4.
approach is used. Knowing that for completing
160
the scheduled production of the four towers
140
the company took about three months. The
120
model will be used to achieve this delivery
100
time using a number of workers similar to the
80
one scheduled by TEGOPI.
60
Model*4 is used as the starting point and
40
then, the number of workers is limited based
20
on the previous usage time. Workers with less
0
than 10 hours of use are discarded for the next
run. Using the previous example of PF5_t1, in
Model*5 the number of workers of this profile
BusyTime
Occurrences
8h
16h
is limited to 6. This way Model*5 uses 96
workers but still more machines than the
24h
number scheduled by the company. Now
limiting the number of machines according to
Graph 4.9 – PF5 1st shift usage: Model*4
specification (Model*6) and reducing five more
Hence profiles PF5_t1[7] to PF5_t1[10]
workers (Model*6.1) the model has provided
are used for less than a eight hours shift,
the following results.
results appear to show that the model lacks
some logic for task selection preventing that,
Due date
for example, a task with more than eight hours
Model*5
Model*6
Model*6.1
duration starts close to the end of a shift.
In a situation when more workers are
available,
a
larger
number
of
machine
Machine resources used
resources is used according to Graph 4.10.
14
12
10
8
6
4
2
0
Torre1
15-Feb-13
19-Feb-13
21-Feb-13
Torre2
25-Feb-13
4-Mar-13
5-Mar-13
Torre3
21-Mar-13
22-Mar-13
28-Mar-13
Torre4
26-Mar-13
5-Apr-13
9-Apr-13
Table 4.6 – Due dates: Model*6 and Model*6.1
With delivery time of approximately three
months, Model*6.1 uses a total of 91 workers
Meaning 8.3% above the number of workers
specified by the company. Although more
workers are used in simulation, this value is
Tegopi
according to the company’s expectation for the
Modelo*4
model’s performance without considering a
flexible worker force.
Graph 4.10 – Allocated machines: Model*4
9
5. Conclusions
http://www.cresesb.cepel.br/content.php?ci
d=tutorial_eolica.
The obtained model allows analysis of the
[
dynamic behaviour of the studied system.
A. Hemami, Wind Turbine Technology,
Although the model represents all of the
2] 5 Maxwell Drive, Clifton Park, NY 12065-
production process in a fairly detailed way,
2919, USA: Cengage Learning, 2012.
special emphasis was given to workers
[
allocation to tasks. Nevertheless it is easy to
3] "Energia Eólica em Portugal 2012," 2013.
change other parameters and test more
[
scenarios.
Breaking
the model,
individual
Publishing ApS, 2009.
detail
by
verifying
worker
[
and
Simulation, New Jersey: PRENTICHEA L-L
ensure correct modelling of the system.
lN TERNATIONAL SERIEINS lN
It is possible to establish that both
DUSTRIAL AND SYSTEMENSG
considered traveling distance and priority logic
INEERING, Fourth Edition.
influence the response of the model. But
[
above all the model is sensitive to changes in
Edition, McGraw-Hill Book Company, 2000.
the specified profiles. This is due to the
[
demand of at least one worker to execute
States of America: Elsevier Inc, 2007.
The due dates supplied for validation are
met when using unlimited resources (Model*4)
individually
T. Altiok and B. Melamed, Simulation
7] Modeling and Analysis with Arena, United
each task from the productive process.
evaluating
A. M. Law and W. D. Kelton,
6] Simulation Modeling and Analisys, Third
the work force capacity and distribution among
when
J. Banks, J. S. C. II, B. L. Nelson and
5] D. M. Nicol, Discrete-Event System
resources utilization and completion time to
and
R. McHaney, Understanding Computer
4] Simulation, Roger McHaney & Ventus
down
production lines were studied and analysed in
greater
R. Redes Energéticas Nacionais,
[
the
W. D. Kelton, J. S. Smith and D. T.
8] Sturrock, Simio and Simulation: Modeling,
performance of each production line under the
Analysis, Applications - Third Edition, Simio
same assumption.
LLC, 2013.
By reorganizing workers profiles it was
[
possible to achieve the necessary time to
9] http://www.tegopi.pt/?mod=pdinamicas_pa
complete the production orders by using a
g&id=2&_item=1.
8.3% higher number of workers than the
[
specified by the company. This result is
F. Chance, J. Robinson and J. Fowler,
10 "Supporting manufacturing with simulation:
acceptable to the company hence the model is
]
under the specification of non-flexible workers.
model design, development, and
deployment," San Diego, CA., 1996.
6. References
[
TEGOPI. [Online]. Available:
S. d. S. Brito, "CRESESB," 2014.
1] [Online]. Available:
10