Beruflich Dokumente
Kultur Dokumente
(5)
where
c
j
i
=
1; if component C
i
requires machining
on resource element RE
j
0; if it does not require machining
on RE
j
8
>
>
>
<
>
>
>
:
Definition of virtual cell boundaries
Virtual cell capability boundaries are defined
using the component clustering approach
based on extended fuzzy C-mean clustering
algorithm with validity measure (Bezdek,
1980). The boundaries of each virtual cell are
represented by RE based centroids
describing the processing requirements of
the corresponding component families. The
objective of the clustering procedure is to
find an optimum set of virtual component
families with maximum similarity between
the processing requirements of the
components in each family and minimum
overlapping between the collective
processing requirements of different
families.
The component family centroids G
i
are
defined by clustering the components in the
component mix based on the similarity of
their processing requirements. The most
distant components which require the largest
number of resource elements are selected as
initial family centroids, which are then
iteratively updated with regard to the newly
clustered components to each family.
Family centroids G
j
are described by
vectors
G
i
= [g
i1
; g
i2
; . . . ; g
in
[ (7)
and reflect the frequency of appearance of
different resource elements as processing
requirements of the components belonging to
each family. The coordinates g
i
in equation
(7) are defined as fuzzy membership
functions, indicating the affinity of the
resource elements to each virtual cell:
Figure 2
Process and virtual cell prototyping methodology an overview
[ 309]
Svetan M. Ratchev
Concurrent process and
facility prototyping for
formation of virtual
manufacturing cells
Integrated Manufacturing
Systems
12/4 [2001] 306315
g
ij
[0; 1[;
X
m
i=1
g
ij
= 1
(8)
The distribution of components around
family centroids is defined on the basis of a
fuzzy membership function
ij
, representing
the degree of lineage of component C
i
to
family centroid G
j
(Bezdek, 1980; Gindy et al.,
1996):
ij
=
1
e
2
(C
j
; G
i
)
1
(f 1)
P
m
i=1
1
e
2
(C
j
; G
i
)
1
(f 1)
X
m
i=1
ij
= 1
(9)
where e
2
(C
j
,G
i
) is a Euclidean norm of
component vectors C
j
and group centroids G
i
and f > 1 is the fuzziness index (Bezdek, 1980).
The coordinates of each group g
ij
are
defined as membership functions of the
resource elements, indicating their affinity to
different groups:
g
ik
=
ik
P
m
j=1
jk
(10)
where
ik
is defined as the weighted sum of
the component coordinates representing RE
k and belonging to group i:
ik
=
P
n
i
j=1
(
ij
)
f
c
jk
P
n1
j=1
(
ij
)
f
(11)
Cell capability boundaries are created by
iteratively updating the family centroids in
Equation (7) and the component membership
functions Equation (9) until a stable partition
is achieved. A compactness and repetition
validity measure R is implemented to
determine the ``optimum'' partition of the
component mix (Ratchev and Gindy, 1996).
The best partition is reached when a
combination of maximum compactness of the
component families and minimum
overlapping between cell boundaries is
achieved (see Figure 3). It is indicated by the
minimum of the validity measure R
r
defined
as the ratio of the average family
compactness to the repetition r of resource
elements in different families:
R
r
=
r
=
P
m
j=1
1
n
j
1
q
j
P
n
j
i=1
ij
e
2
(C
i
; G
j
)
P
m
i=1
1
q
i
(G
i
)
2
(12)
where m is the number of virtual cells, n
j
is
the number of components belonging to
virtual family with centroid G
j
, e
2
(C
i
,G
j
) is
the Euclidean distance between component
C
i
and cell centroid G
j
, and q
j
is the number
of resource elements clustered to G
j
.
The output of the RE based clustering
procedure is an optimum partition of the
component mix into families with their
processing requirements defining the
boundaries of the virtual cells.
Allocation of machine tools to virtual cell
boundaries
The next step in the iterative cell formation
process is to transform the virtual cell
boundaries represented by sets of resources
elements into physical machine tools. The
transformation is based on iterative
substitution of the resource elements with
corresponding machine tools to which they
may belong. Each machine tool is evaluated
in terms of the number of resource elements
it has in common with each cell capability
boundary within the available machine
capacity.
Machines are selected using a wide range
of criteria including minimum variety of
machines in the cells; minimum transport
movements in the cell; maximum utilisation;
optimum level of concentration of the
operations (e.g. machining centre against
dedicated machines), etc. The machine
allocation procedure is completed once all
resource elements in the component
partition are substituted by machine tools.
In the simplified illustrative example (see
Appendix) a strategy emphasising minimum
variety of machines and transport
movements is considered as an illustration of
the approach. Accordingly, machines are
selected on the basis of maximum capability,
i.e. preference is given to machines
encompassing the maximum number of
resource elements required by the cell.
Each machine tool is assessed in terms of
the maximum number of common resource
elements to the resource view of each virtual
cell. The machine tools selection procedure is
based on iterative matching of the closest
pair of machine tool/virtual cell vector using
the machine tool membership function.
'
i;j
=
1
e
2
(M
j
; G
i
)
1
(f 1)
P
m
i=1
1
e
2
(M
j
; G
i
)
1
(f 1)
(13)
At each step the machine tool with the
minimum '
ij
is distributed to the
corresponding virtual cell. The machine
allocation procedure is completed, once all
resource elements in the component
[ 310]
Svetan M. Ratchev
Concurrent process and
facility prototyping for
formation of virtual
manufacturing cells
Integrated Manufacturing
Systems
12/4 [2001] 306315
partition are substituted by machine tools.
By analogy with the RE based validity
measure R
r
(Equation (12)) a machine tool
based validity measure R
m
is used to assess
the compactness and separation of the virtual
cells in terms of allocated machine tools.
Processes and facility evaluation
At each iterative step the process and cell
prototypes are evaluated in terms of virtual
cell compactness and expected performance.
At the evaluation stage the planners have the
option of changing the optimisation criteria
for selection of processing alternatives or
modifying the key factors for selection of
machine tools. The approach also allows
different ``what-if'' scenarios to be modelled
and iteratively evaluated in order to find the
best solution. A variety of system constraints
is used in selecting the ``best'' processing
solution among the process planning
alternatives returned by the planning
module. The selection constraints are defined
with regard to the targeted type of production
(e.g. minimum set-ups assume a larger
concentration of operations leading to
selection of a limited number of machines
with higher capabilities; maximising the
number of set-ups allows selection of a wider
variety of machines with limited individual
capabilities). Component level constraints
are also considered, so that requirements like
``machine in the same set-up'' and ``machine
parallel from the same direction'' are taken
into account in deciding the most
appropriate processing and resource
alternatives.
The machine tool allocation is guided by
sets of company specific performance
indicators such as facility utilisation,
resource capacity, reduction of lead times,
minimum inventory, minimum transport
costs, etc. Depending on the selected strategy
for structuring the virtual cellular system
different factors can be combined in defining
the procedures for machine allocation and
validation of the virtual cellular partitions.
Experimental results
The methodology has been experimentally
tested using examples based on production
data from a large industrial company with
virtual cells formed in order to better reflect
the changes which occur in the component
mix. The machining facility is described
through the set of machine tools and their
constituting resource elements (see Table I).
The results of the iterative virtual cell
formation for a sample set of 50 components
are shown in Table II. It can be observed that
the best compactness and separation between
the cells are achieved in the case of two cells
(R
m
= 0.61). Although there are shared
capabilities in terms of required resource
elements (e.g. REs 1, 5, 6, 9), the boundaries of
the virtual cells are clearly defined in terms
of machine tools without any overlapping
Figure 3
Compactness and separation measure for validation of virtual cell
boundaries
Table I
Machine tool capabilities represented by REs
Machine
Id. Type
Resource
elements
1 Turret drill RE4 RE5
2 Slotting machine RE4 RE9
3 Turning lathe RE1
4 Shaper RE7
5 Turning lathe RE1 RE3
6 Horizontal milling machine RE6 RE5 RE3
7 Drill RE5
8 Turning lathe RE1, RE3
9 Turning lathe RE1
10 Bridge mill RE6 RE5 RE4
11 Turning lathe RE1
12 Slotting machine RE9
13 Shaper RE7
14 Radial drill RE4 RE5
15 Grinder RE8
16 Drill RE4
17 Grinder RE8
18 Turning lathe RE2
19 Drill RE4 RE5
20 Milling machine RE6
21 Grinder RE8
22 Grinder RE8
23 Turning lathe RE1 RE2
24 Borer RE4 RE5 RE3
25 Turning lathe RE2
Notes: RE1 turning (1); RE2 turning (2); RE3
boring; RE4 drilling; RE5 drilling/reaming; RE6
milling; RE7 shaping; RE8 grinding; RE9 slotting
(vertical shaping)
[ 311]
Svetan M. Ratchev
Concurrent process and
facility prototyping for
formation of virtual
manufacturing cells
Integrated Manufacturing
Systems
12/4 [2001] 306315
between the cells. The case of two cells also
provides the best balance between the
number of parts distributed to each cell
compared with the cases with three and four
cells.
The flexibility which the use of resource
elements provides for configuration and re-
configuration of virtual cells is further
demonstrated in Figure 4. After the first step
of the algorithm is completed (refer to Table
II) the boundaries of two virtual cells are
defined using eight resource elements (the
membership of the resource elements to each
of the cells is shown in brackets, e.g. RE2
belongs only to virtual cell 1 with = 1, while
RE1 belongs to cell 1 with = 0.47 and to cell 2
with = 0.53). Two different cases of machine
allocation could be considered. In the first
case (refer to the partition with two groups,
as described in Table II) the machine tools
are selected on the basis of cells with
maximum compactness (i.e. maximum
Table II
Virtual cell formation results for a sample set of 50 components
Number
of cells Cell
Number
of parts Resource elements
a
Machine tools
a
Validity
measure R
m
2 1 26 1 (047), 2(1), 3(1), 5(062), 6(053),
9(056)
10(1), 3(1), 14(1), 2(1),
4(1)
0.61
2 24 1 (053), 4(1), 5(038), 6(047), 8(1),
9(044)
22(1), 23(1), 6(1), 12(1)
3 1 9 2(0.35) 5(0.55), 3(0.69), 7(1) 23(0.78), 17(1), 4(1) 0.65
2 19 6(0.60), 4(0.38), 8(1), 1(0.32),
9(0.36)
10(1), 3(1), 14(1), 2(1)
3 22 2(0.65), 4(0.62), 5(0.45), 1(0.68),
9(0.64), 6(0.40), 3(0.31)
22(1), 13(1), 6(1), 12(1),
23(0.22)
4 1 13 2(0.52), 5(0.62), 3(0.83), 9(0.49),
1(0.15)
23(0.87), 22(1), 2(1) 0.68
2 9 4(0.51), 1(0.48), 9(0.51), 8(0.34) 3(1), 7(1), 12(1), 14(1)
3 15 2(0.35), 4(0.49), 5(0.16), 1(0.37),
6(1), 3(0.17)
5(1), 10(1), 17(1),
23(0.13)
4 13 8(0.66), 5(0.22), 7(1), 2(0.13) 16(1), 1(1), 4(1), 24(1)
Note:
a
The membership coefficients of the resource elements and the machine tools are shown in parentheses
Figure 4
Flexible virtual cell prototyping using REs
[ 312]
Svetan M. Ratchev
Concurrent process and
facility prototyping for
formation of virtual
manufacturing cells
Integrated Manufacturing
Systems
12/4 [2001] 306315
similarity between the components in each
family in terms of processing requirements)
and minimum overlapping (minimum shared
machine tools). If resource sharing between
two or more cells is the preferred option due
to capacity considerations, it can be easily
accommodated by relaxing the constraints on
cell overlapping at the machine allocation
stage. This is illustrated by the second case
(partition 2), where the same cellular
boundaries are transformed into a smaller
number of machine tools (seven compared
with nine in the first case) with a limited
sharing of resources being allowed (a turning
lathe and a slotting machine are shared by
the two cells for processes described by
resource elements S2 and T1).
The methodology has been further tested
using four sample sets of components
sequenced for production over different
periods of time. The component sets were
specifically selected to demonstrate the
effects of the fluctuations in the component
variety on the structure of the virtual
cellular facility. In order to validate the use
of resource elements the virtual cell
formation was repeated using the traditional
machine tool based description of the
component processing requirements and the
shop capabilities. The results of the
experimental tests are summarised in
Figure 5. It could be observed that, due to the
improved assessment of the component
similarity at sub-machine level and more
precise definition of cell capability
boundaries, the virtual cells created using
REs were consistently more compact
(improvement of 38 per cent on average in
terms of compactness and repetition
measure). The advantages of the approach
are also evident from the lower number of
repeated (shared) machine tools for the
majority of the partitions (average reduction
of 41 per cent). Although sharing of resources
is one of the features of the virtual
manufacturing concept, there are always
advantages of defining well separated virtual
cells with limited or no overlapping, which is
easier to achieve when using REs.
The superior quality of formation of
virtual cells using resource elements is
further underlined by the consistently more
compact solutions in terms of number of cells
when using resource elements (refer to
Figure 5). A comparison of the overall
number of machine tools selected for the
``optimum'' partitions also shows the
superior performance of the RE based
formation methodology. The distinctively
lower number of machine tools selected
using REs is due again to the fact that by
describing the capabilities of the
manufacturing facility at sub-machine level
the overall form-generating capabilities of
the individual machine tools can be captured,
assessed and better utilised.
Conclusions
The proposed methodology supports the
general trend for shortening of the product
life cycle by allowing manufacturing systems
to be easily and more often renewed and re-
configured in order to maximise the
utilisation of the available resources in
manufacturing new products. The reported
conceptual framework contributes to the
development of unified approach for product
requirement assessment and flexible re-
configuration of machining facilities. It can
be extended to allow existing facilities in
workshops with small to medium production
volumes of production to be ``re-used'' by re-
configuring them in accordance with the
changes of the market environment and
product demand.
The concurrent process and virtual cell
prototype methodology can also facilitate
planners in obtaining more realistic and
timely manufacturing information and
feedback at different planning stages. A
fundamental requirement in virtual cell
formation is the availability of choices in
terms of routeing alternatives for each
component. Using resource elements
different routes can be generated by
translating the virtual cell boundaries into
different sets of machine tools. This allows
multiple choices in configuration and re-
configuration of the virtual cells in order to
select for each component mix the most
Figure 5
Virtual cell formation results for four sample component sets using RE
and machine tool based decision making
[ 313]
Svetan M. Ratchev
Concurrent process and
facility prototyping for
formation of virtual
manufacturing cells
Integrated Manufacturing
Systems
12/4 [2001] 306315
efficient combination of workstations in
terms of machine tool capability and capacity
utilisation.
The reported product, process and facility
capability models provide a new generic
representation of the production knowledge
required for decision making in a virtual
cellular environment. The form-generating
schemata and resource elements describe
uniquely the distribution of processing
capabilities among different machine tools
by capturing the ``exclusive'' and ``shared''
facility capability boundaries. The model
facilitates the decision-making process by
providing greater flexibility in allocating
production resources and higher
responsiveness to changes in the component
mix and production priorities in dynamic
facility reconfiguration.
The concurrent process and virtual cell
prototyping methodology facilitates the
decision-making process at early planning
stages by allowing assessment of the
implications of the component designs on the
required machining processes and the
virtual cells that need to be configured in
order to efficiently manufacture the product.
The proposed prototyping algorithms and
implementation strategy provide an
integrated decision-making environment for
deciding the most appropriate combination
of machine tools for specific component
requirements.
References and further reading
Aguiar, M.W.C., Murgatroyd, I.S. and Edwards,
J.M. (1996), ``Object-oriented resource models
their role in specifying components of
integrated manufacturing systems'',
Computer Integrated Manufacturing Systems,
Vol. 9, pp. 33-48.
Bezdek, J. (1980), ``A convergence theorem for the
fuzzy ISODATA clustering algorithms'', IEEE
Transactions on Pattern Analysis and
Machine Intelligence, Vol. 2 No. 1, pp. 1-8.
Chatterjee, S. (1992), ``Resourcing in dynamic
manufacturing'', Proceedings of ASME, PED,
Concurrent Engineering, Vol. 59, pp. 361-74.
Drolet, J., Moodie, C.L. and Montreuil, B. (1989),
``Scheduling factories of the future'', Journal
of Mechanical Working Technology, Vol. 20,
pp. 183-94.
Drolet, J., Moodie, C.L. and Montreuil, B. (1990),
``Virtual manufacturing layout planning'',
International Industrial Engineering
Conference Proceedings, IEE, San Francisco,
CA, pp. 236-41.
Gindy, N.N.Z. and Ratchev, S.M. (1997),
``Component grouping for cellular
decomposition of manufacturing facilities
using resource elements'', Integrated
Manufacturing Systems, Vol. 8 No. 3 & 4,
pp. 215-22.
Gindy, N.N.Z., Ratchev, S.M. and Case, K. (1996),
``Component grouping for cell formation
using resource elements'', International
Journal of Production Research, Vol. 34 No. 3,
pp. 729-54.
McLean, C.R., Bloom, H.M. and Hopp, T.H. (1982),
``The virtual manufacturing cell'', Proceedings
of the 4th IFAC/IFIP Conference on
Information Control Problems in
Manufacturing Technology, Gaithersburg,
MD, pp. 1-9.
Mills, J.J., Graham, J.K., Elmasri, R.A. and
Weems, B.P. (1993), ``The virtual
manufacturing workbench representation
and interface issues'', IFIP Transactions B
Application in Technology, Vol. 10, pp. 231-44.
Montreuil, B., Drolet, J. and Lefrancois, P. (1992),
``Design and management of virtual cellular
manufacturing systems'', Annual
International Conference Proceedings,
American Production and Inventory Control
Society, pp. 410-14.
Ratchev, S.M. and Gindy, N.N.Z. (1996),
``Formation of virtual cells using generic
capability patterns'', Proceedings of the 6th
International Conference on Flexible
Automation and Intelligent Manufacturing,
Atlanta, GA, 12-15 May, pp. 508-18.
Rheault, M., Drolet, J. and Abdulnour, G. (1995),
``Physically re-configurable virtual cells: a
dynamic model for a highly dynamic
environment'', Computer and Industrial
Engineering, Vol. 29, pp. 221-5.
Rheault, M., Drolet, J. and Abdulnour, G. (1996),
``Dynamic cellular manufacturing system
(DCMS)'', Computers & Industrial
Engineering, Vol. 31 No. 1-2, pp. 143-6.
Rogers, G.G. and Bottaci, L. (1997), ``Modular
production systems: a new manufacturing
paradigm'', Journal of Intelligent Systems,
Vol. 8 No. 2, pp. 147-56.
Rolstadas, A. (1995), ``Enterprise modelling for
competitive manufacturing'', Control
Engineering Practice, Vol. 3 No. 1, pp. 43-50.
Xie, X.L. and Beni, G. (1991), ``A validity measure
for fuzzy clustering'', IEEE Transactions on
Pattern Analysis and Machine Intelligence,
Vol. 13 No. 8, pp. 841-7.
Appendix. Virtual cell formation algorithm
an illustrative example
Step 1. Component processing
requirements
For details of the component processing
requirements, see Tables AI and AII.
Step 2. Define virtual cell boundaries using
REs
For details of steps 2.1-2.3, see Tables AIII-AV.
Step 2.4. Calculation of RE based validity
measure R
r
= 0.40. (Repetition of steps 2.1- 2.4
until a stable partition is achieved)
[ 314]
Svetan M. Ratchev
Concurrent process and
facility prototyping for
formation of virtual
manufacturing cells
Integrated Manufacturing
Systems
12/4 [2001] 306315
Step 2.5. Definition of RE based cell
boundaries: G
1
(Comp: 1, 4 , 6; REs: 1, 4); G
2
(Comp.: 2, 3, 5, 7; REs:1, 2, 5)
Step 3. Allocation of machine tools to
virtual cell boundaries
Step 3.1. Definition of cell boundaries: G
1
(Machine tool: 4); G
2
(Machine tools: 1, 5)
Step 3.2. Calculation of machine tool based
validity measure R
m
= 0.13
Step 4. Process and virtual cell evaluation
Step 4.1 Process and virtual cell evaluation.
(Repetition of steps 1 - 3.2 applying different
clustering criteria).
Step 4.2. Increase the number of cells: m =
m + 1. (Repetition of steps 1 - 4.1 for the
increased numbers of cells).
Step 4.3. Selection of the best partition:
partition with two groups/cells
Table AI
Components Resource elements
1 RE1
2 RE1, RE2
3 RE2
4 RE1
5 RE1, RE2, RE5
6 RE1, RE4
7 RE2, RE5
Table AII
Machine tools Resource elements
1 RE1, RE2
2 RE1
3 RE1, RE2, RE3
4 RE2, RE4
5 RE1, RE5
Table AIII
Step 2.1 Number of cells m = 2; definition of
initial group centroids (G
1
, G
2
)
Group centroids RE1 RE2 RE3 RE4 RE5
G
1
(Component 6) 1 0 0 1 0
G
2
(Component 7) 0 1 0 0 1
Table AIV
Step 2.2 Definition of component membership
functions
ij
and component clustering
Group/
Comp. 1 2 3 4 5 6 7
1 0.75 0.5 0.25 0.75 0.25 1 0
2 0.25 0.5 0.75 0.25 0.75 0 1
Table AV
Step 2.3 Recalculation of group centroids
(G
1
, G
2
)
Group RE1 RE2 RE3 RE4 RE5
G
1
0.67 0 0 1 0
G
2
0.33 1 0 0 1
[ 315]
Svetan M. Ratchev
Concurrent process and
facility prototyping for
formation of virtual
manufacturing cells
Integrated Manufacturing
Systems
12/4 [2001] 306315