You are on page 1of 6

COMPARISON OF MACHINE ALLOCATION INTER-RELATIONSHIP

ALGORITHM AND DELAYED PRODUCT DIFFERENTIATION IN LAYOUT


DESIGNS
LIJO JOHN SHIBIN K T Dr. R.SRIDHARAN,
Graduate student Graduate student Professor

Department of Mechanical Engg


National Institute of Technology Calicut

Abstract— Group Technology has been applied for responds to the changes in the demand, whereas efficiency
generating the machine groups and corresponding part is a measure of the cost effectiveness of the manufacturing
families for the cellular manufacturing. There is a close system. Responsiveness is a measure of demand flexibility
relationship between the products to be manufactured and efficiency a measure of the design flexibility.
and the layout for its manufacturing. In the present day
volatile market, the demand variation is a big challenge In a manufacturing system, during the layout
faced by the manufacturing companies. Therefore, the design, there is always a trade-off between volume and
companies are constantly trying to improvise their variety. If the layout is designed for the manufacturing
layouts to cope up with the demand variation. There is system which can handle a wide range of demand
a continuous effort to design better layouts which are fluctuation, then the system will not be able to incorporate
technologically feasible and economically viable. But, much changes in the design, whereas the layout which can
one cannot forget the amount of design variation handle a wide variety of design changes will not respond
demanded by the customers. In such cases, it becomes effectively to changes in the demand. Under these
difficult to design a layout which will give larger circumstances, it becomes extremely important to generate
product variation. In this research, two cellular layouts new algorithms for the layout design which can handle both
that include the demand and design variation are the design and the demand flexibility.
compared. The first layout is generated by using the
Machine Allocation Interrelationship (MAIN) A comparative study between two layout design
algorithm. The MAIN algorithm considers the demand procedures has been done here. The MAIN layout design
variation over the given period to generate a layout (Chan et.al,2004) deals with the quantitative demand
which minimizes the cost. In the second layout, a variation. The demand variability in the given multi-period
biological taxonomical technique called cladistics is planning horizon is being tried to capture. Both Static
used for generating the layout. The layout thus obtained Machine Cellular Layout (SMCL) and Dynamic Machine
incorporates delayed product differentiation to optimize Cellular layout (DMCL) has been considered. In SMCL, if
the layout for design variations by considering the n machines are grouped into a cell, there are as many as n!
similarity between the parts. Further, based on the potential layouts. In case of DMCL, the same machine
inferences made, suggestions have been provided to group may have to be rearranged to suit for m periods in a
develop layouts which will have both demand and planning horizon. As the result, the probable combinations
design flexibility. are as large as (n!)m and it is difficult to obtain the optimal
solution based on QAP approach (Rosenblatt 1986,
Keywords Cellular manufacturing, Group technology, Balakrishnan and Chun 1998). Rosenblatt (1986) first
Delayed Product Differentiation, layout design, Cladistics introduced a layout problem-solving technique for multiple
planning periods. His dynamic programming model
attempts to find out the best layout for each predetermined
1. Introduction period. Balakrishnan et al. (1992) added the budget
Group technology is a manufacturing philosophy that constraint for machine rearrangements and solved the
advocates identifying similar parts and grouping them into problem by using the shortest path algorithm.
families and taking advantage of the similarities in their
design or the manufacturing methods. Cellular Delaying Product Differentiation (DPD) can
manufacturing is the practical application to this philosophy increase flexibility, reduce uncertainties and decrease the
where the functionally dissimilar machines are grouped cost of complexity (Shao and Ji 2008). The benefits of
together to make parts that are similar. The cellular postponement strategies include achieving economies of
manufacturing is an important technological innovation scale and mitigating risks by placing buffers of work-in-
which provides both responsiveness and efficiency in the process (WIP) at points of products differentiation (Aviv
manufacturing (Wicks and Reasor, 1999). Responsiveness and Federgruen 2001). In order to attain production
is effectiveness with which the manufacturing system
flexibility without sacrificing efficiency, some design
principles were recommended, ranging in complexity from
convincing the customer to accept less product variety, to
increasing parts standardization, building the capability for
variety into the product rather than the production system by
using different combinations of the same parts, and
identifying the sources of variety changeover time/cost and
seeking ways to eliminate them through product design.
Modularity is the main enabler for production in a DPD
environment. Modularity is a relative property that depends
on the degree of similarity between the physical and
functional architecture of the design and how much
interactions between components are minimized.
Standardized components and interchangeable units are two
facets that relate to modularity in design and manufacturing, Figure 1. A cladogram commonality representation of a five-
taxa example
when creating product variants (Ulrich and Tung 1991).
ElMaraghy (2008) introduced a hierarchy of variations from
products features to parts, modules, sub-assemblies and The rest of the paper is organized as follows. The
parts and products families and platforms to illustrate the problem defenition is defined in section 2. The detail of the
propagation of variety at all levels and discussed techniques solution methodology is described in the section 3. The
and methods for coping with these variations in products section 4 provides the illustration of the algorithm suggested.
and systems design. Section 5 gives a discussion on the results whereas section 6
provides the conclusion

In this research we have compared two layouts, and have 2. Problem Definition
combined the concepts to come out with a new algorithm for This research aims on developing a layout considering both
incorporating the design and the demand flexibility in the the demand and design variations. The manufacturing cell,
system. The first layout that we have considered is the where the machine location sites are regular grids as
Machine Allocation Interrelationship (MAIN). In MAIN considered by Chan et. al 2004. Here in this research we
algorithm the layout has been generated with the help of
have restricted ourselves to consider the layout as a 3 x 2
heuristics. The MAIN algorithm tries to investigate the
Dynamic Machine Cellular Layout in the multiple periods grid with six potential machine location sites where five
planning horizon. Initially, a layout is generated with fixed machines has to located and the empty cell can be either
quantitative demand profile referred to as Static Machine used if we have to opt for machine duplication or for
Cellular Layout (SMCL). This layout is the best for the give material storage. We also assume that, a machine can be
period and the given demand. Then the SMCL’s for all the assigned to any location and the part traveling distance is
periods are considered and a dynamic layout is being estimated by measuring the rectilinear distance from the
generated which allows machine relocation between periods centre of the source machine to the centre of the destination
(Chang et. al 2004). In the second layout the concept of machine. The size of the machine location zone should be
delayed product differentiation is being used. By delaying able to accommodate the largest machine in the group. The
the product differentiation point we can incorporate the machine with larger number of in flows must be place
design flexibility in the layout. The precedence constraints together. The MAIN algorithm aims in generating a layout
are being incorporated into the system by using a biological which will be best suited for the demand variation. The
taxonomical technique called as the cladistics. The DMCL which is an extension of SMCL gives the best
precedence constraints are represented diagrammatically layout. The generated layout can be either single DMCL
with the help of the cladograms. Each node in the which has a unique layout throughout the periods or
cladograms represents the point of differentiation, each multiple-DMCL which has a changing configuration.
character represents the machines and the end taxa represent
the products, as shown in figure (AlGeddawy et. al. 2010)
In layout generation with DPD, the layout is being
generated by considering the precedence constraints. The
precedence constraints can be best depicted with the help of
the cladograms. By selecting the best cladogram we can get
a rough idea about the layout which can further developed
to reduce the cost. Nevertheless the cladograms prove to be
The determination of the basic and the merged part flow
weights are intimately related to the quantitative demand of
a part (i) in a family in a period (Qp,i ) and the part
transportation quantity per move (Hi,jk). Where λi,jk gives the
Figure 2: six zone location site part handling factor which is a measure of the ease with
which the parts can be transported between the machines
a very important tool for generating an initial layout. after the operations.
3. Solution Methodology In dynamic stage the total cell layout cost (α) over multiple
In this section a brief description of the procedure for periods is:
generating the layout by considering the both demand and
design variation is presented.

Notations
Zp – Total part traveling score
Dj,k – Distance between two locations ω represents the cost per unit traveling score and it is a
Fp,jk - merged part flow weight for period p constant. In the period-to-period machine rearrangement
Tp,jk ,Tp,kj - the unidirectional basic part flow weights cost (Rp,p +1), it consists of two portions; they are the basic
Qp,i - quantitative demand of a part (i) in a family in a cost for machine relocation such as the setup/installation
period cost (Mj) and the cost per unit machine movement (Cj). The
Hi,jk - transportation quantity per move calculation of this cost is based on:
λi,jk - part handling factor
ω - cost per unit traveling score
α- the total cell layout cost
Rp,p +1 - period-to-period machine rearrangement cost
Mj - setup/installation cost The next step in the proposed algorithm is to draw the
Cj - cost per unit machine movement cladogram, which gives us the information contained in the
machine-part matrix. While drawing the cladogram it should
The static stage of the MAIN algorithm is to determine the be kept in mind that the length of the cladogram should be
total part traveling score (T) in a cell in a period Zp with a tried to reduce as much as possible. Also, while drawing the
particular layout. The purpose is to minimize the total cladogram the precedence constraints should be included in
traveling score and this is analogue to the cost on making such a way that the points of differentiation should be
part movements. The objective function is defined as: pushed as lower as possible. After drawing the cladogram,
depending upon the information contained in it, a
preliminary layout is being generated and by using the
traveling score equation the total score for the layout is
being calculated. In most cases the initial layout generated
Where will not be the optimal one. So a pair wise interchanging of
Dj,k = | Xj - Xk | + | Yj - Yk the machines can be done to bring those machines together
which have large amount of flow between them. For each
The Manhattan method has been chosen for calculating the layout generated the total traveling score is being calculated
distance (Djk) between two points where (Xj,Yj) and (Xk,Yk ) and the layout which yields the lowest score is being
are the coordinates of the measuring points of machines j selected.
and k with reference to the 3x2 grid.
Fp,jk=Tp,j,k+Tp,k,j The same procedure is repeated for all the periods and
the layout with the least score for all the periods put together
where Fp,jk is the merged part flow weight for period p. It is being selected. This is similar to the single DMCL, which
combines the unidirectional basic part flow weights (Tp,jk considers the demand of all the periods in the planning
and Tp,kj) to form a generic correlation of two machines.
horizon. We can also opt for the multiple DMCL wherein static stage of the MAIN when dealing with the period ( p1)
we go for the machine relocation in between periods. are presented in the third column in table 3. After the part
flow weights were ranked, the layout designer can start
pulling the machines to form a 3x2 machine layout grid
according the rankings. Once the pulling operations have
been completed, the distances among machines (Dj,k) can be
obtained by referring to the Manhattan method. Then, the
traveling score for each machine pair can be determined
accordingly and the summation of these traveling scores
give the total part traveling score. The results have been
summarized in the table 3

Table 3: Iteration results for period 1

iteration 1 iteration 2
Fp,jk x Fp,jk x
j,k Fp,jk RANKING Djk Djk Djk Djk
3,4 50 5 2 100 1 50
4,2 50 2 100 2 100
2,1 95 2 1 95 1 95
1,5 50 2 100 2 100
1,4 160 1 1 160 1 160
Figure 3 : Cladogram for the given data 4,5 160 1 160 1 160
5,3 60 3 1 60 2 120
3,2 60 2 120 1 60
4. Numerical establishments 2,5 55 4 3 120 3 165
1,3 55 3 165 2 110
The data used for the research is being randomly generated
total 1180 1120
and the calculations are being done and summarized. In the
process operation side, there were four parts processed by
The layout conforming to iteration 1 is as shown in the
five machines. For each part the machining sequence and table 4. When we consider the Fp,jk x Djk the largest value is
demand over a four period planning horizon is as tabulated 165, so as a rule of thumb, we try to reduce this value by
below. bringing to machines 1 and 3 closer. This is shown in the
Table 1: Machining sequence table 5. Thus in second iteration the value reduces to 110.
PARTS MACHINING SEQUENCE Therefore the total traveling score reduces to 1120 from
1180, which shows that the layout generated by iteration 2 is
A 3 4 2 1 5 better. In most cases the optimal layout can be obtained in
B 2 1 4 5 0 two iterations
C 1 4 5 3 2 Table 4: Results of iteration 1 for all periods
D 2 5 4 1 3 p1 score p2 score p3 score = p4 score
=1180 =1070 1720 =1925
Table 2 : Demand over the planning horizon
1 4 5 1 4 5 1 4 5 1 4 5
DEMAND 2 3 2 3 2 3 2 3
PERIOD A B C D
P1 50 45 60 55
Table 5: Results of iteration 2 for all periods
P2 70 40 50 40
P3 80 90 75 85 p1 score p2 score p3 score = p4 score
=1120 =1180 1720 =1785
P4 85 80 90 70
1 4 5 1 4 5 1 4 3 1 4

In the static stage, all the part transportation 2 3 2 3 2 5 2 5 3


quantity per move and the part handling factors were set to
‘1s’. The part flow weights (Fp,j,k) generated by applying the
Next, the cladogram for the given machine-part matrix is During the analysis we considered the both part
being generated. By considering the precedence constraints handling factor and the transportation quantity as 1. Thus
the machines which have maximum interaction is being two important elements were not being used during the
identified. Based on the inferences, layouts are generated and analysis. We have also not considered the machine
the part traveling score for each layout is being calculated. duplication cost. The machine duplication cost will also be
Further, the machine locations are being interchanged and added to the total cost equation. In some cases the layout
the layout which gives the least score is being selected. The generated by both the MAIN and the DPD yielded the same
part traveling score for the selected layout in all the periods layout under such cases we can very well take the layout to
is being calculated. The results are tabulated in the tables
be the optimal for the given period.
below
Table 6: Results for DPD layout
For the single DMCL we can go for the layout which
yields the least score. In this case except for the second
iteration 1 iteration 2
period we can see that the layout generated by the DPD gives
Fp,jk Fp,jk the least traveling cost. In case of the multiple DMCL,
j,k Fp,jk RANKING Djk x Djk Djk x Djk relocation cost of machines between periods is also
3,4 50 5 1 50 2 100 considered. Thus for the each given period the best layout
4,2 50 2 100 2 100 can be selected from the layouts which yield the least
2,1 95 2 1 95 1 95 traveling score.
1,5 50 2 100 2 100 6. Conclusion
1,4 160 1 1 160 1 160
4,5 160 1 160 1 160 Machine layout problems in manufacturing have received
5,3 60 3 2 120 1 60 considerable attention The proposed MAIN-DPD algorithm
addresses the problems in both static and dynamic layouts
3,2 60 3 180 2 120
by incorporating practical factors such as the part-handling
2,5 55 4 1 55 1 55 factor, the basic cost for machine relocation and the cost per
1,3 55 2 110 3 165 unit machine movement, etc. MAIN makes a best
total score 1130 1115 compromise of the above-mentioned factors with the
closeness of machines in order to minimize the total cell
layout cost. This algorithm also caters for the possibility of
Table 7: Layout generated by DPD allowing a single layout for multiple periods with demand
iteration 1 iteration 2 fluctuation. On the other hand the DPD considers the design
similarity of the parts that are being grouped into one cell.
2 5 2 5 3 The proposed methodology borrows the well-known
1 4 3 1 4 Cladistics classification analysis from biology, where
commonalities in different entities are used, and modifies it
The part traveling score for all the periods is for the layout to impose precedence constraints that must be considered in
obtained for DPD in second iteration is being given in the sequencing operations. This technique results in an optimum
table below. layout for DPD, determines the differentiation points,
Table 8: Part traveling score for all periods provides a feasible processes-sequence, and reduces
processes repetition, without using any prior assumptions or
period period 2 period 3 period 4 conditions regarding the number and locations of the
differentiation points.
1115 1100 1700 1725 A combination of these two concepts together in
layout design gives a better control over the factors affecting
it. Thus depending upon demand or design, which ever has a
5. Results and Discussion
larger impact on the system can be identified and can be
captured. For instance if the variation in demand is to be
For the randomly generated data, both MAIN algorithm and
controlled then MAIN will generate a better layout. In cases
the DPD is being used to generate a suitable layout. From
where design has to be controlled DPD yields a better result.
the results tabulated above we can see that the initial layout
By the proposed algorithm, depending upon which factor
obtained by the MAIN algorithm may not be optimal. When
has to be controlled the MAIN or the DPD will yield the
we analyze the results it can be found the layouts obtained
layout.
by the DPD yielded better results on an average basis. But
when we consider the period two the iteration 1 of MAIN
algorithm yielded better results. But in all the other periods
the layout generated by the DPD yielded better results.
Reference
Shao, X. and Ji, J., 2008. Evaluation of postponement
Wicks, E.M and Reasor, R.J, 1999, Designing cellular strategies in mass customization with service guarantees.
manufacturing systems with dynamic part populations, IIE International Journal of Production Research, 46, 153–171
Transactions,31,11-20
Aviv, Y. and Federgruen, A., 2001. Design for
Chan W.M, Chan,C.Y, Kwong,2004, C.K Development of postponement: a comprehensive characterization of its
the MAIN algorithm for a cellular manufacturing machine benefits under unknown demands distributions. Operations
layout ,International Journal of Production Research,42, 51- Research, 49, 578–598.
65
Ulrich, K. and Tung, K., 1991. Fundamentals of product
AlGeddawy, Tarek, ElMaraghy, Hooda, 2010, Assembly modularity. Issues in Design Manufacturing/Integration,
systems layout design model for delayed products ASME, 39, 73–79.
differentiation, International Journal of Production
Research,48, 5281-5305 ElMaraghy, H., 2008. Changing and evolving products and
systems – models and enablers, In: H. ElMaraghy, ed.
Balakrishnan J, Chun H.C,1998, Dynamic layout Changeable and reconfigurable manufacturing systems.
algorithms: a state-of-the-art survey. Omega, International London: Springer-Verlag, 25–45
Journal Management Science, 26, 507–521

Rosenblatt, M. J., 1986, The dynamics of plant layout.


Management Science, 32, 76–85

Balakrishnan, J., Jacaobs, F. R. and Venkataraman, M. A.,


1992, Solution for the constrained dynamic facility layout
problem. European Journal of Operational Research, 57,
280–286

You might also like