European Journal of Operational Research 81 (1995)115-133

North-Holland

115

Theory and Methodology

Assignment of tools to machines

in a flexible manufacturing system
Thomas H. D'Alfonso and Jose A. Ventura
Department of Industrial and Management Systems Engineering,
University Park, PA 16802, USA

The Pennsylvania State University,

Received July 1992

Abstract: Subgradient optimization is employed to solve the problem of assigning tools to machines in a
flexible manufacturing system (FMS). Machines in the FMS have a limited number of slots in the tool
magazine, and tools may require multiple slots. Tools are grouped based upon a pairwise similarity
coefficient that indicates the frequency of successive operations that require both tools. Typical solution
strategies have included graph theoretic heuristics. Lagrangian relaxation is utilized in the algorithm
developed in this article. The problem is formulated as a linear integer program. After dualizing two sets
of constraints, two integer subproblems are formed in which the first is further decomposed into several
knapsack subproblems. The second subproblem can be solved by a linear network code. A subgradient
algorithm is developed to solve the dual problem. The algorithm is compared to a graph theoretic
heuristic that utilizes cluster analysis. For most test problems, the subgradient algorithm is superior, but
is sensitive to convergence parameters.
Keywords: Scheduling; Flexible manufacturing systems; Subgradient optimization; Lagrange multipliers;
Heuristics

1. Introduction
This article investigates the application of group technology (GT) to a special manufacturing
environment. GT is the science of grouping objects together based upon similar characteristics. Hyer
(1984) described GT as a way of organizing information about similarity of objects. Gallagher and Knight
(1986) went one step further, and described GT as a way of taking advantage of similarities of parts or
processes in all stages of design and manufacturing systems. GT is a means of organizing both the factory
layout and production flow. These definitions imply that GT has had broad applications in engineering.
Production flow analysis requires a classification scheme for organizing parts in the manufacturing
process. McAuley (1972) published one of the first G T studies in this area. The focus was on analyzing
characteristics of parts to be manufactured. McAuley developed a relationship factor between machines

Correspondence to: Prof. J.A. Ventura, Department of Industrial and Management SystemsEngineering,The PennsylvaniaState
University, UniversityPark, PA 16802,USA.
0377-2217/95/$09.50 1995 - ElsevierScienceB.V. All rights reserved
SSDI 0377-2217(93)E0135-K

116

T.H. D'Alfonso, J,4. Ventura /Assignment o f tools to machines

that manufactured components. The number of components that require both or either machine in a pair
is used to determine the similarity coefficient.
GT is considerably versatile. A successful application requires basically two things: a classification
scheme for objects to be grouped and an objective that can be optimized by examining possible
groupings. Ng (1989) classified both parts and machines into families and cells, respectively. Grouping
efficiency was quantified, and the method can improve machine utilization.
In this paper, GT is applied to a flexible manufacturing system (FMS) environment with special
machine constraints. The objective is to minimize production travel between machines that are capable
of performing specified operations. Various parts are manufactured that require tooling sequences.
When a tool is not included in the tool magazine of the station that is producing a part, the part must
travel to another station. Each FMS machine is restricted to a limited number of tools it may carry.
FMSs are considerably complex, and with each new objective or constraint considered, the model may
change significantly, and so too the solution algorithm. Cluster analysis, graph theory, and dual
algorithms are not new solution strategies to manufacturing models.
Two solution algorithms are used in this study. The first is a subgradient algorithm based upon a
Lagrangian Dual (LD) formulation. The second solution algorithm is a graph theoretic heuristic that
combines techniques supported in separate published papers. Computational results for several FMS
environments are presented. Kusiak (1990) asserted that GT has become one of the most important
methods applied to intelligent manufacturing systems. Designing rows of machines, determining machine
cells, organizing part information, and designing layouts for new factories and renovations of existing
factories, can all benefit from GT. With each new application, new algorithms may need to be developed.
This need has been highlighted by Ng (1989), Kusiak (1990), Mulvey and Beck (1984), and by Ventura
and Hsu (1989) and Ventura et al. (1990).

2. Review of recent optimization techniques for group technology

GT models have been optimized by graph theoretic methods such as cluster analysis, various
heuristics, and nonlinear and integer programming. McCormick et al. (1972) applied a graph theoretic
heuristic to an airport design problem. Similarity coefficients between various services of an airline were
determined, and service stations were established. For example, passenger and baggage check-in are
highly related operations in an airport, and these services almost always occur at the same station. In
1975, Lenstra and Rinnooy Kan showed that the airport design problem could be modelled as a traveling
salesperson problem and solved it with existing graph theoretic algorithms.
Ng (1989) developed a heuristic especially for the machine cell problem. Machines are grouped into
ceils, parts are grouped into families, and a grouping efficiency measure determines how they are joined.
The key to Ng's heuristic is a maximal spanning tree algorithm. Spanning trees, whether minimal or
maximal, are very effective in joining nodes (machines) from a graph based on a simple objective
function. The maximal spanning tree algorithm works as follows. Given a network, repeatedly select the
arc with the highest weight, provided a cycle is not formed by doing so. In a complete network with n
nodes, n - 1 arcs will be selected to form a spanning tree (Gould, 1988).
Lagrangian relaxation is an effective approach to solving complicated mathematical models. In a very
simple description, this approach identifies constraints that are difficult to meet and incorporate them
into the objective function. Also called Lagrangian dual (LD) methods, these techniques use subgradients to optimize over the modified, less complicated, expanded solution space. Part of the objective,
then, is to systematically approach feasibility in the complicated constraints. The term subgradient refers
to the fact that the objective function is nondifferentiable, and subgradients are used in the optimization
algorithm rather than gradients.
In GT modelling, only a few research publications have included Lagrangian relaxation as the solution
algorithm. Mulvey and Crowder (1979) developed one of the first such models. Ventura et al. (1990)
developed a LD formulation of a GT application to FMSs. In this model, production flow data are

117

T.H. D'Alfonso, J.A. Ventura /Assignment of tools to machines

Table 1
Tool sequence and daily demands for production of seven parts
Part

Tool sequence

Daily demand

2
3
4
5
6
7

2
4
5
1
3
4

1
2
2
5
2
2

4
3
1
2
5
3

3
5

25
100
25
50
25
25
100

1
1

considered and tools and parts are grouped so as to minimize the interdependencies among the groups.
U p p e r and lower bounds were placed on the number of tools and parts per FMS station. The
formulation used by Ventura et al. (1990) is similar to the formulation considered in this article. The
complete formulation is given in the next section.

3. P r o b l e m f o r m u l a t i o n

The FMS scenario considered in this article is one in which machines are capable of operating several
tools stored in a tool magazine. Parts to be manufactured require a certain sequence of tooling
operations to be performed, and many different parts can be produced by the system. The problem is to
determine which tools are assigned to which machine, while considering
1) the sequence of tools each part requires,
2) the demand for each part, and
3) the number of tools each machine magazine can hold.
Furthermore,
4) some tools may take up more than one magazine slot, and
5) the number of groups (machines) is predetermined.
Consider the information in Table 1 pertaining to a set of parts that are to be manufactured. Let (i, j)
be defined as operation i immediately preceding operation j. For example, (5, 2) occurs in the
production sequence of Part 4, which has a daily demand of 50. Ordered pair (5, 2) also occurs in the
production sequence of Part 5 with a demand of 25. Ordered pair (5, 2) occurs nowhere else in the
production plan. Because daily demand measures the number of times a particular part will be
manufactured, the total number of times per day that sequence (5, 2) occurs is 75. It is our objective to
minimize the production flow between machines; therefore, the amount of travel between each ordered
pair is an important measurement. Table 2 contains a matrix in which each cell shows daily production
flow between ordered pairs.
Because the objective is to minimize flow between machines, and oij represents the daily production
flow from tool i to tool j, a better performance measure is to consider the total flow between one tool

Table 2
Daily production flow (oij) between ordered pairs of tools
1
2
3
4
5

1
0
175
100
0
25

2
0
0
25
125
75

3
25
125
0
100
0

4
100
0
25
0
0

5
25
25
25
0
0

T.H. D'Alfonso, J.A. Ventura / Assignment of tools to machines

118

Table 3
Daily production flow (aij) between pairs of tools
1

1
2
3
4

175
-

125
150
-

100
125
125

50
100
25
0

and another for a given daily production schedule. That is, consider aij = Oij "[-Off, the flow in both
directions. The matrix A = [ a i ] ] is symmetric, and can be constructed by O + O w, where O = [oil]. The
upper triangular part of such a matrix is given in Table 3. Finally, the number of magazine slots each tool
requires is given in Table 4. This property will constrain the choice of tool assignments.
To formulate the mathematical expression for the objective function, consider the variable Xik, that
indicates the membership of tool i in machine (group) k:
1

Xik =

if t o o l i ls with machine k,
otherwise,

for

i=l
9

n,

k=l

p.

Now, if tool i and tool j are not in the same group, the daily production flow will be aij. A new
~

variable, Xijk, can be introduced so that

1

Xijk=
for

if tools i and j are with machine k,

otherwise,

i=1 .... ,n-l,

]=i+l,...,n,

k = l . . . . . p.

With this variable, it is easy to derive the objective function.

First, production flow of aij between machines occurs when EkXijk = 0 and aij ~ O. It is therefore the
\
objective to minimize the aggregate flow; namely,
n

min ~_.

aij

--

xij k .

(1)

i=1 j = i + l

Expression (1) simplifies to the objective function that is used in this article:
n--1

max E
i=1 j = i + l

aijXijk"

(2)

k=l

The objective, (2), is interpreted as determining tool groups that maximize aggregate daily production

Table 4
N u m b e r of magazine slots required by each tool
tool (i)

magazine slots (r i)

1
2
3
4
5

2
1
1
2
1

T.H. D'Alfonso, ZA. Ventura /Assignment of tools to machines

119

flow among m e m b e r s within the groups is equivalent to minimizing aggregate production flow between
groups (machines). The objective function can also be expressed in a nonlinear form:
n--1

max E

~_,

i=1 j=i+l

aijXikXjk.

(3)

k=l

In this study, function (2) is used as the objective, but the Xik-variables are also u s e d in the constraint
set that will be developed next.
Because each tool can only be assigned to one machine, the following constraint is necessary:
P

E Xik = 1

for i = 1 , . . . , n.

(4)

k=l

Because the objective (2) and constraint set (4) differ in variable types, constraint sets (5) and (6) are
introduced:

xij k ~ Xik

for i = 1 . . . . , n - 1,

j = i + 1 . . . . . n,

Xiik<_Xjk

fori=l

j=i+l

.... ,n--l,

. . . . . n,

k = 1.... , p,

(5)

k=l,...,p.

(6)

A special case of the FMS scenario is one in which tools may take up m o r e than one slot in the
magazine. Magazine capacity may be limited, however, a minimum n u m b e r of slots for each machine may
need to be filled to justify the additional machine. These properties arc accommodated by
n

L< Erixik~U

fork=l

.... ,p,

(7)

i=1

where r; is the n u m b e r of magazine slots tool i requires.

Finally, the property that variables are indicative is expressed by constraints (8) and (9):

Xijk~{O, 1}
Xik~{O, 1}

fori=l

....

,n-l,

fori=l,...,n,

j=l+l
k=l

. . . . . n,

k--1,...,p,

.... ,p.

(8)
(9)

A linear integer p r o g r a m (LIP) is formed when (2) is optimized subject to constraint sets (4)-(9).
In order to implement a network algorithm to solve LIP, it is advantageous to change riXik to Yik, and
to substitute

Xik

=Yik/ri

for i = 1 , . . . , n ,

k = 1. . . . . p.

(10)

T h e problem formulation in its entirety is now

(LIP)

maxZ(x, y)

n--1

E aUXijk

(11)

i=1 j=i+l k=l

subject to
P

~-, Yik = ri for i = 1 . . . . , n,

(12)

k=l

riXijk<Yik

fori=l

.... ,n--l,

j=i+l

. . . . ,n,

rjXiik<Yik

fori=l

..... n--l,

j=i+l,...,n,

k=l

.... ,p,

k=l,...,p,

(13)
(14)

L<

~Yik<_U
i=1

fork=l

..... p,

(15)

120

T.H.

D'Alfonso,J.A.Ventura/Assignmentoftoolstomachines

Xijk~{0,1 }

fori=l

..... n--l,

Y i k ~ { O , ri}

fori=l

. . . . . n,

j=i+l,...,n,

k=l,...,p,

(16)

k=l,...,p.

(17)

Two algorithms will be examined to solve this problem. The first uses subgradient optimization, and
takes advantage of the structure of LIP. The second is a graph theoretic heuristic that follows the
recommendations given in research papers on similar G T problems.
4. Lagrangian dual formulation
Because problem LIP is difficult, a strategy t o simplify the solution process is employed. Lagrangian
relaxation is used to dualize constraints (13) and (14). The Lagrangian function becomes
n--1

ZD(W,U)=max E ~

~-~ [aijXijk+Wijk(Yik--riXijk)+Uijk(Yjk--FjXijk)].

(18)

i=1 j = i + l k = l

The variables wiyk and viyk are the Lagrange multipliers that correspond to constraints (13) and (14),
respectively.
The Lagrangian dual problem then becomes
(CDP)
Min Z ~ ( w , ~)

(19)

subject to w, v > 0.

(20)

Equation (18) can also be expressed in the form

n--1 n

Z D ( W , v) = max ~

~,

~, ( % -- riWijk -- rjVijk)Xijk +

i=1 j = i + l k = l

~ ~ [( ~
k=1i=1

) (i--j~ll

Wisk +

Viik

)]

Yik"

j=i+l

(21)
In this form, it is easy to determine a decomposition of (21) into functions of xij k and Yik, and to assign
constraints with the corresponding variables. Such a decomposition forms two subproblems, SP1 and
SP2. A comprehensive development of each subproblem follows.
(SP1)
n--1

Max E

E (aiy--riWijk--ryViyk)Xijk

(22)

i=1 j = i + l k = l

subject to
P

~_,Xijk<_l

fori=l

.... ,n--l,

j=i+l,...,n,

fori=l

..... n--l,

j=i+l

(23)

k=l

Xijk~{0,1 }

. . . . . n,

k=l,...,p.

(24)

Constraint (23) is added to the model to reduce the size of the expanded solution space. The
constraint would be redundant in LIP due to constraints (12)-(14). SP1 can be further decomposed into
n(n - 1) 0-1 knapsack (KSP) subproblems. This follows from the fact that 1 + 2 + + n - 1 (i, j)
pairs exist in SP1. Each KSP subproblem (i, j ) can be solved by the following algorithm:
Step 1. L e t k ' ~ arg max{aij - riwij x - rjviiklk = 1 . . . . . p}.
Step 2. If aij - riwij k, - rjvij k, >_ O, then xi*k, = 1 and xi~k = 0 for all k :/: k'. Otherwise, set x * k = 0 for
k = 1. . . . . p.

T.H.D'Alfonso,J.4. Ventura/ Assignmentof toolsto machines

121

This algorithm may be employed because, for each (i,j), the sum of xij k over each k must be less than
or equal to 1, and the objective is maximized over each (i,j) by choosing the largest aij - r i w i y k - r j v i j k.
Constraint (23) is implicitly met by this method.
The second subproblem is as follows.
(SP2)

Max ~P
k=l

~ [( =~i+lWijk)
i=1

(i--j~=l)]
Vjik Yik

(25)

j
P

subject to

Y i k : ri

for i = 1 . . . . , n,

(26)

k=l

L< ~y~k<U

fork=l

..... p,

(27)

i=1

Yik~{O, ri}

fori=l

.... ,n,

k=l .... ,p.

By making the y/k-variables continuous this model becomes

lower and upper bounds on the requirements. The original nodes
required number of slots for each tool, and the destination nodes
at the destination nodes are bounded by L and U. By making
solution to SP2 does not necessarily satisfy constraint (28). This
next section.

(28)
a linear transportation problem with
represent tools, the input flows are the
represent machines. The requirements
Yik continuous, however, the optimal
and other issues are addressed in the

5. Subgradient optimization
Because the objective function of LDP is nondifferentiable, and generally contains many variables,
subgradient optimization is a good choice for a solution algorithm. The technique either produces an
optimal or provides a good bound on the optimal solution. The subgradient algorithm is an iterative
procedure that produces a sequence of Lagrangian multipliers (Held et al., 1974). This sequence contains
a subsequence that converges to the optimal multipliers (Goffin, 1977). The following subgradient
algorithm was employed to solve LDP.

Subgradient algorithm.
Step 1. Initialization. Pick the stopping criterion, e > 0. Set w 1 =

v 1 = 0,

LB = 0,

n--1

UB =

i=1

j=i+l

aiy,

and r = 1. The vectors w and v are dual multipliers for constraint sets (13) and (14), respectively. LB is
the lower bound, associated with the primal problem, and UB is the upper bound, associated with the
dual problem.
Step 2. Solve SP1 and SP2. Using the current Lagrangian multipl!ers w r and v r, let x r be an optimal
solution to SP1 with objective function value SP r, and let y~ be an optimal solution to SP2 with objective
function value SP2 r. Set ZD(Wr, Vr) = SPY + SP2 r.
Step 3. Update upper bound. Let UB r = min{(UW -1, ZD(W ~, V~)}. UB r is the upper bound, namely the
best current dual solution, Z D-

Step 4. Generate primal feasible solution.

Step 4.1. If (X r, y~) is feasible to LIP, then set xx ~ = x r, yy~ = y r , and go to Step 5. Otherwise,
Step 4.2. For each i, let Kr(i) = arg max{Y[klk = 1. . . . . p}.
Step 4.3. For each i, if I g r ( i ) l = 1, let k <i) be the only member of g r ( i ) . If [ g r ( i ) [ > 1, randomly
select k <i)~ gr(i).

122

T.H.

D'Alfonso, J.A. Ventura /Assignment o f tools to machines

Step 4.4. Let

yyrik = { riO
r __
XXijk --

{~i

otherwise,ifk=k(i) for i = 1 . . . . , n ,
i f b o t h y y / ~ = r i a n d y y jkr =ry,
otherwise,

k = 1.... ,p,

for i = 1 , . . . , n -

1,

j=i + l,...,n,

k = l . . . . . p.

(The vector (XXr, yyr) may not be primal feasible).

Step 5. Update lower bound. If (xx r, yy~) is primal feasible, LB ~ = max{(LB r - l , Z ( x x r, yy~)}; otherwise,
LB r = L W -1. LB r is the lower bound, namely the current best primal solution, Z.
Step 6. Check stopping criterion. If ( U B r - - L B r ) / L B r < e, then stop. Otherwise, go to Step 7.
Step 7. Calculate subgradients.
G~k = rixryk -- yirk,
r _

H i j k -- ryxiik - Yjk,

i = 1 . . . . . n - 1,

j=i+l

. . . . . n,

k=l,...,p,

i = 1 . . . . , n -- 1,

j=i+l

. . . . . n,

k = 1,...,p.

Step 8. Compute step size.

[ f r ( U B r - LBr)]
tr:

v~n-l~n
vi~p [ [ G r ~2
(nirl.k)2]"
z-'i~l z-'j=i+lZ"k=l[[, ijk} 4-

Step 9. Update Lagrange multipliers.

r I
wi~.~1 = max{0, wijr k + t r G u~J,
r
viyr +kl =max{0, vijr k + t r Hijk},

i = 1 , . . . , n - 1,
i= l .... ,n-l,

j=i+l

.... ,n,

j=i+l,...,n,

k= 1,...,p,
k = l . . . . . p.

Step 10. Increment r by 1 and go back to Step 2.

It should be pointed out that Step 4 can produce a primal feasible solution, which is not always at
hand for Lagrangian algorithms. Step 4.3 calls for the k-th group to be selected randomly if more than
one group is a candidate. This feature helps to prevent cycling in the algorithm.
Subgradients are similar to gradients, which are directions of steepest ascent for a vector function.
Subgradients exist when the function is nondifferentiable. Since the primal problem is a linear integer
program, the dual function is nondifferentiable. The step size determines how far along a subgradient
the next variable value will be. The numerator coefficient, f r , is a value that may be changed each
iteration. It will be shown in a later section that the choice of the step size parameter, f r , for each
iteration is very important to finding a good solution. In this study an initial f r is chosen, and is
periodically halved (every 1, 2, 3, or 4 iterations).
SP2 was solved with a network flow computer program called N E T F L O (Kennington and Helgason,
1980). A constraint was added to the network flow problem requiring tool 1 to be assigned to group 1,
tool 2 to be assigned to either group 1 or group 2, and so on, until the p-th tool and above, that can b e
assigned to any group. This modification reduced the number of existing optimal solutions and greatly
improved the performance of the subgradient algorithm.

6. Graph theoretic heuristic

The subgradient algorithm described in the preceding section has its drawbacks. As the number of
tools and machines increases, the problem size gets significantly large. Furthermore, the subgradient
algorithm requires careful selection of the step size for each iteration. This will be demonstrated in the
next section. Finally, Lagrangian dual techniques do not guarantee a primal feasible solution. Because

T.H. D'Alfonso, J.A. Ventura / Assignment of tools to machines

123

constraints are dualized, solutions generated by the algorithm may be infeasible in these constraints. The
subgradient algorithm has the objective of minimizing infeasibility, but not necessarily eliminating
infeasibility. In this case, however, we have b e e n able to develop a heuristic algorithm that may produce
a primal feasible solution at each iteration.
Many heuristics have b e e n developed and applied to problems similar to LIP. Many of these heuristics
are simple to program, and find good feasible solutions. T h e heuristic described in this section is called
Chop the Maximal Spanning T r e e (CMST). It combines a technique suggested by Ng (1989) to form a
maximal spanning tree (MST) of the tool production flow matrix, and an algorithm for determining a
solution that meets the machine capacity constraints. The latter part of the heuristic is similar to
modifications m a d e to the d e n d r o g r a m technique of Auguston and Minker (1970) by Seifoddini (1988)
and Shtub (1988). Shrub suggested to start with one group, and form additional groups by removing
m e m b e r s so as to m e e t feasibility constraints while reducing dissimilarity. The C M S T algorithm
described here follows the advice of Shtub, but the feasibility heuristic is unique.

Chop the Maximal Scanning Tree Algorithm.

Step 1. Form a maximal spanning tree. T h e tool production flow matrix is symmetric. It can be
considered a complete graph, in which the nodes are tools, and arcs are weighted by aij. W h e n a tool
requires m o r e than one slot, say r of them, then r nodes are used to represent this tool. The weight
between these nodes is infinite, ensuring their selection in the maximal spanning tree (MST). An MST
can be formed from this arrangement. Gould (1988) provided an algorithm for accomplishing this.
Step 1.1. R e p e a t e d l y select the largest arc t h a t does not form a cycle until a single tree is formed.
Step 2. Create a forest maintaining lower bound feasibility. T h e M S T must have p - 1 arcs removed to
form p subgraphs. Removing arcs from a tree results in what is called a forest. Each tree in the forest
created by C M S T must be b e t w e e n a minimum and a maximum size. W h e n choosing an arc to be
removed, two things must be considered. First, the forest must m e e t size requirements, and second, the
objective function must be kept in mind. These steps are followed:
Step 2.1. Sort the arcs by nondecreasing weight.
Step 2.2. R e m o v e arcs from the top of the list, maintaining lower bound (LB) feasibility. If removing an
arc violates LB feasibility consider the next arc in the list.
Step 2.3. Continue until p - 1 arcs have b e e n removed. It may be necessary to backtrack, replace an arc
that has b e e n removed, and continue. The worst-case scenario, complete enumeration, has iterations
totalling (p21).

Table 5
Computational results for five-tool test problem
Number of Tools (n)
Number of Groups (p)
Lower Bound (L)
Upper Bound (U)
Subgradient Algorithm:
Solution
Groupl
Group2
Dual Lower Bound
Iterations
CMST Algorithm:
Solution
Groupl
Group2
L iterations
U iterations

Test 1
5
2
2
5

Test 2
5
2
2
4

625
{1,2,3,5}
{4}
975
3

450
{1,2,3}
{4,5}
975
3

625
{1,2,3,5}
{4}
2
1

450
{1,2,5}
{3,4}
2
2

T.H. D'Alfonso, J.A. Ventura / Assignment of tools to machines

124

Table 6
Description of control variables for airport design problem
Item

Space

Description

01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27

1
1
2
1
2
2
2
3
2
2
1
2
2
2
2
2
2
1
1
3
2
1
2
3
2
2
2

Passenger check-in
Baggage check-in
Baggage claim
Baggage moving system
Intra-airport transportation
Cargo terminal
Close-in parking lot
Remote parking lot
Main access road
Circulation roads
Rental car service area
Rental car parking lot
Curb space for unloading
Curb space for loading
Waiting areas at gates
Transportation stations
Aircraft loading system
Concessions
Rental car desk
Runway capacity
Number of gates
Passenger information
Cargo transfer
Air-traffic control
Refuse removal
Flight crew facilities
Aircraft service on the apron

Table 7
Dependency coefficients for airport design problem
0
1
0 1
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21

0
2
-

0
3
3

0
4
0

0
5
2

3
3

0
6

2 0
0 0
1 0
0 0
0

0
7

0
8

0
9

1
0

1
1

1
2

1
3

1
4

1
5

1
6

1
7

1
8

1
9

2
0

2
1

2
2

2
3

2
4

2
5

2
6

1
1
1
0
2
0

1
1
1
0
2
0
2

0
0
0
0
1
2
2
2

0
0
0
0
1
1
3
2
3

0 0 2 0 0 0 0 1 0 0 0 3 0 0 0 0
0 0 2 0 0 1 0 0 1 0 0 3 0 0 0 0
0 0 0 3 0 1 0 0 1 0 0 0 0 0 0 0
0 0 1 1 0 0 1 0 0 0 0 0 0 0 0 0
0 0 0 0 0 3 0 0 1 0 3 1 0 0 0 0
0 0 0 0 0 0 1 0 0 1 0 0 3 0 0 1
1 0 1 1 0 1 0 0 0 0 0 1 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 1 2 2 0 0 0 0 1 0 0 0 0 0 0 0
3 0 0 0 0 0 0 0 1 0 0 0 0 1 0
1 1 0 0 0 0 1 0 0 0 0 0 0 0
0 0 1 0 0 1 0 0 2 0 0 0 0
0 1 0 0 1 0 0 0 0 0 0 0
1 1 1 0 0 3 1 0 0 0 0
1 1 1 0 3 2 0 0 0 0
0 0 1 1 0 1 0 0 0
0 0 0 1 0 0 2 0
0 0 1 0 0 0 0
2 0 0 3 0 1

2
7
0
0
0
0
0
2
0
0
0
0
0
0
0
0
1
0
3
0
0
0

T.H. D'Alfonso, J.A. Ventura / Assignment of tools to machines

175

125

('~

Figure 1. G r a p h theoretic representation of test problem

Step 3. Redistribution o f nodes to obtain upper bound feasibility, p s u b t r e e s

successful.
bound

Nodes

are

redistributed

among

groups

in the

remainder

of the

are

formed

heuristic

if Step

2 is

to satisfy upper

constraints.

Table 8
Computational results for airport design problem a

fr

II

0.50
0.50
0.50
0.50
0.75
0.75
0.75
0.75
1.00
1.00
1.00
1.00
2.00
2.00
2.00
2.00
Z~TER
Z~

1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4

ZDITE R

CMST *
LITER
UITEr

p = 3

p = 4

p = 5

U = 20

U = 30

U = 15

U=20

U = 15

U = 20

94
96
106
91
106+
90
95
96
96
96
93
91
98-- c
98
91
93
69
143
93
INF
6
252

120
119
120
118
114
119
120
123+
123
119
120
119
119-118
119
118
68
146
95
104
6
3

81
73
79
84
70
73
74
83
74
71
77
78
71-67
81
85+
56
140
91
INF
8
1749

86
85
81
88
87
93
99+
83
88
83
83
81
78-91
86
88
70
143
100
91
8
1

61
77+ b
74
55
57
67
69
62
69
65
64
68
69-65
64
52
47
140
90
86
10
1

84
78
77
71
81
85
77
91 +
72
76
81
81
83-82
86
67
69
143
101
86
10
1

a p = n u m b e r of groups, U = m a g a z i n e tool capacity, f r = initial step-s~e coefficient, II = m u l t i p l e of iterations to halve step-size

coefficient, Z * I T E R = smallest n u m b e r of iterations required to obtain the best primal solution for subgradient algorithm,
ZDrrE R smallest n u m b e r of iterations required to obtain best dual solution for subgradient algorithm, CMST * = best solution
determined by Chop the Maximal Spanning Tree heuristic, LITER = n u m b e r of lower b o u n d iterations for heuristic, UITER =
n u m b e r of u p p e r b o u n d iterations for heuristic, I N F = no feasible solution was determined.
b + : This solution produces largest primal objective value ( Z * ) .
c _ : This solution produces smallest Z * .
- -

T.H. D'Alfonso, J.A. Ventura / Assignment of tools to machines

126

Step 3.1. Check feasibility. If all subtrees have no more than UB nodes, stop, the heuristic is successful.
Otherwise, examine the subtree with the largest number of nodes. Select the lightest-weighted arc that
disconnects a single node. UB is the upper bound on the group size.
Step 3.2. Reconnect the node to a different subtree with less than UB nodes. The node is connected by
the largest weighted arc between it and a candidate subtree. Go to Step 3.1.

7. Computational results
Five-tool test problem
The 5-tool problem that was described earlier provides the first test results. Recall from Table 4 that
tools # 1 and # 4 require 2 slots of the FMS machine magazine. Figure 1 shows the graph representing
the tool production flow matrix obtained from Tables 3 and 4. The problem is to group these five tools
into 2 groups with at least 2 slots filled, but no more than 5 slots filled. With these bounds, both solution
algorithms found the optimal solution. When the upper bound on magazine capacity is lowered from 5 to
4, the current solution is infeasible. B o t h algorithms again found the optimal solution. Results of both

Table 9
R e o r d e r e d matrix for airport design test problem
Tool

Tool
1

1
2
3
4
5
7
8
9
13
16
19
22
6
15
17
20
21
23
24
26
27
10
11
12
14
18
25

1
3
0
2
2
1
1
0
2
0
0
3
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0

3
0
3
0
1
1
0
2
1
1
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

0
1
3
1
1
1
0
0
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
3
0
0

2
0
1
0
0
0
0
1
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
1
0
0

2
3
3
1
2
2
1
0
3
1
1
0
0
0
0
3
0
0
0
0
1
0
0
0
0
0

0
1
0
1
2
2
1
1
0
1
0
0
0
0
0
0
0
0
0
3
1
0
1
0
0

1 1 0
1 1 0
1 1 0
0 0 0
2 2 1
1 2 2
- 1 2
2 - 1
0 0 1 0 1
0 0 1
0 0 2
0 2 0
0 0 0
0 0 0
0 0 0
0 0 0
0 0 0
0 0 0
0 0 0
0 0 0
2 3 2
0 0 0
0 0 1
0 0 0
0 0 0
0 0 0

61517202123242627101112141825

913161922

2
2
0
1
0
1
0
0
1
1
2
0
1
.
0
3
0
0
0
0
0
0
0
1
1
0

0
1
1
0
3
1
1
0
1
1
1
0
0
1
0
0
0
0
0
0
1
0
1
1
0
0

0
1

3
3

1 0
0 0
1 1
0 1
0 0
0 0
1 2
1 2
1 1
- 1
0 1 0
0 0
0 1
0 0
0 3
0 0
0 1
0 2
0 1
0 0
0 0
0 0
1 0
0 0

0
0

0 0 0 0 0 0
0 0 0 0 0 . 0
0 0 0 0 0 0 0
0 0 1 0 0 0 0
0 0 0 0 3 0 0
0 0 0 0 0 0 0
0 0 0 0 0 0 0
2 0 0 0 0 0 0
0 0 0 0 0 0 0
0 1 1 0 3 0 0
0 0 0 0 0 0 0
0 1 0 0 0 0 0
1 0 1 1 0 3 0
- 1 1 0 3 0 0
1 1 - 1 1 1 1 0
0 1 - 1 2 0 3
3 1 2 - 1 0 1
0 1 0 0 - 1 0
0 0 3 1 0 - 1
0 0 1 1 0 2 1 3 0 1 1 1 0
0 0 0 0 0 0 0
0 0 1 0 0 0 0
0 0 0 0 0 0 0
0 0 0 0 0 0 0
1 0 0 0 0 0 0
0 0 0 0 0 0 0

a Subgradient algorithm, 3 machines, magazine capacity = 20 slots.

0
0
0
0
0
0
0
0
0
0
0
1
0
0
1
1
0
2
1
0
0
0
0
0
3

0
0
0
0
0
0
0
0
0
0
0
2
1
3
0
1
1
1
0
1
1
1
2
0
0

0
0
0
1
3
2
3
2
0
1
0
1
0
0
0
0
0
0
0
0
1
3
0
0
1

0
0
0
0
1
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
1
1
1
0
0

0
0
0
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
1
3
1
0
0

0
0
3
1
0
1
0
0
0
1
1
0
0
0
0
0
0
0
0
0
0
2
0
1
1
2

1
0
0
0
0
0
0
0
0
1
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
1
-

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
3
0
1
0
0
2
1

T.It. D'Alfonso, J.4. Ventura / Assignment of tools to machines

127

t e s t p r o b l e m s a r e s u m m a r i z e d i n T a b l e 5. T w o o p t i m a l s o l u t i o n s e x i s t t o t h i s v e r s i o n o f t h e p r o b l e m a n d
both are represented in the table.

A i r p o r t layout test problem

M c C o r m i c k e t al. (1972) u s e d d a t a r e l a t i n g s e r v i c e s a t a n a i r p o r t . T h i s e x a c t t e s t p r o b l e m w a s a g a i n
u s e d b y L e n s t r a a n d R i n n o o y K a n (1975) t o d e m o n s t r a t e t r a v e l i n g s a l e s m a n p r o b l e m m o d e l s . T h e
o r i g i n a l p r o b l e m w a s t o g r o u p a i r p o r t s e r v i c e s i n t o c l a s s e s b a s e d o n a d e p e n d e n c y f a c t o r r a n g i n g f r o m 0,
f o r i n d e p e n d e n t s e r v i c e s , t o 3, f o r h i g h l y d e p e n d e n t s e r v i c e s . A n e x a m p l e o f r e l a t i v e l y i n d e p e n d e n t
services are passenger check-in and the rental car repair shop. An example of highly dependent services
is p a s s e n g e r c h e c k - i n a n d b a g g a g e c h e c k - i n . I n m o d e r n a i r p o r t s r e l a t i v e l y s m a l l s t a t i o n s m a y b e u t i l i z e d if
n e c e s s a r y t o o p t i m i z e c u s t o m e r f l o w . F o r t h i s r e a s o n , g r o u p s i z e s w e r e p e r m i t t e d t o b e as l o w as o n e
unit.
T h i s t e s t p r o b l e m p r o v i d e s d a t a t h a t will c o r r e s p o n d t o a n i m a g i n a r y t o o l d e p e n d e n c y m a t r i x . I t is
a d a p t e d b y a s s i g n i n g t o e a c h s e r v i c e a n i n t e g e r s i z e f a c t o r . T h a t is, t h e p r o b l e m c o n s i d e r e d i n t h i s
research requires tools to be grouped into FMS machines with limited space. Each tool takes up an
i n t e g e r a m o u n t o f slots. I f e a c h a i r p o r t s e r v i c e is a s s i g n e d a n i n t e g e r s i z e f a c t o r , a n d l i m i t s a r e p l a c e d o n
the number and size of groups, then the problems are equivalent. Data for the airport design problem
a r e d e s c r i b e d i n T a b l e 6 a n d c o n t a i n e d i n T a b l e 7.

Table 10
Reordered matrix for airport design test problem a
Tool

Tool
1 22

1
22
2
13
19
4
12
3
11
14
5
21
15
16
6
9
23
10
7
8
17
27
25
18
20
24
26

3
3
2
0
2
0
0
0
0
2
0
0
0
0
0
0
0
1
1
0
0
0
1
0
0
0

3
3
2
1
0
0
0
0
0
1
0
1
2
0
0
0
0
1
0
0
0
0
1
0
0
0

2
3
3
2
1
3
0
0
0
0
0
0
0
1
0
0
0
0
1
1
0
0
0
0
0
0
0

13
2
2
2
1
1
1
0
0
0
0
0
0
1
0
0
0
2
1
0
0
0
0
0
0
0
0

19
0
1
1
1
0
1
1
0
1
1
0
0
1
0
0
0
1
0
0
0
0
0
0
0
0
0

4
2
0
3
1
0
0
3
0
1
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0

12
0
0
0
1
1
0
0
3
1
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0

3
0
0
0
0
1
3
0
0
3
1
0
0
1
0
0
0
0
1
1
0
0
0
0
0
0
0

11

14

0
0
0
0
0
0
3
0
0
0
0
0
0
0
0
0
1
1
0
0
0
1
0
1
0
0

0
0
0
0
1
1
1
3
0
0
0
0
1
0
0
0
2
1
0
0
0
0
0
0
0
0

21

15

16

2
1
0
0
1
0
0
1
0
0
3
0
3
0
1
0
1
2
2
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0
0
3
3
3
0
0
0
0
0
0
1
1
0
0
2
1
1

0
1
0
0
0
0
0
0
0
0
0
3
1
0
0
0
0
0
0
1
1
0
1
0
0
0

0
2
1
1
1
0
0
1
0
1
3
3
1
0
0
0
0
1
1
1
0
0
1
0
0
0

6
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2
3
1
0
0
1
2
0
0
1
0
1

23

10

17

27

25

18 20

24

26

0
0
0
0
0
0
0
0
0
0
1
0
0
0
2
0
3
2
2
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0
0
0
0
0
0
3
0
0
0
0
1
1
0
0
0
0
0

0
0
0
2
1
0
1
0
1
2
1
0
0
0
1
3
0
3
2
0
0
0
0
0
0
0

1
1
1
1
0
0
0
1
1
1
2
0
0
1
0
2
0
3
2
0
0
0
0
0
0
0

1
0
1
0
0
0
0
1
0
0
2
0
0
1
0
2
0
2
2
0
0
0
0
0
0
0

0
0
0
0
0
1
0
0
0
0
0
1
1
1
1
0
1
0
0
0
3
0
0
1
0
0

0
0
0
0
0
0
0
0
0
0
0
1
1
0
2
0
1
0
0
0
3
3
0
0
1
0

0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
3
2
0
0
0

1
1
0
0
0
0
0
0
0
0
0
0
1
1
0
0
0
0
0
0
0
0
2
0
0
0

0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
1
0
0
3
2

0
0
0
0
0
0
0
0
0
0
0
1
0
0
1
0
0
0
0
0
0
0
0
0
1
2
-

a CMST algorithm, 5 machines, same solution for magazine capacity = 15 or magazine capacity = 20.

0
0
0
0
0
0
0
0
1
0
0
2
0
0
1
0
0
0
0
0
1
0
0
0
3
1

T.H. D'Alfonso, J.A. Ventura /Assignment of tools to machines

128

Table 11
Reordered matrix for airport design problem a
Tool

Tool
1

10 22

11

12

13

14

16

17

19 20 21

24

26

15 23

18 25

27

2
4
5
7

3
2
2
1

3
0
1

3
0
0

0
0
2

1
0
2
-

1
0
2
2

0
0
1
2

0
0
1
3

3
0
1
1

0
3
1
1

0
0
0
1

0
0
0
0

2
1
0
1

0
1
0
1

1
0
3
1

0
1
0
0

1
0
1
0

0
0
0
0

0
0
3
0

0
0
0
0

0
0
0
0

0
0
0
0

0
0
0
0

0
0
0
0

0
0
0
0

0
0
0
0

0
0
0
0

9
10
22
3
11
12
13
14
16
17
19
20
21
24
26
6
15
23
18
25
27

0
0
3
0
0
0
2
0
0
0
0
0
0
0
0
0
0
0
1
0
0

0
0
3
0
0
0
2
0
1
0
1
0
0
0
0
0
0
0
0
0
0

0
0
0
3
0
0
1
1
0
1
0
0
0
0
0
0
0
0
0
0
0

1
1
1
1
0
0
0
0
3
0
1
0
3
0
0
0
0
0
0
0
0

2
3
1
1
1
0
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0

2
2
0
1
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0

3
0
0
0
0
0
0
0
0
0
0
0
0
0
2
0
0
0
0
0

3
0
0
1
1
2
2
0
0
1
0
0
0
0
1
0
0
0
0
0

0
0
0
0
0
2
0
2
0
1
0
0
0
0
0
1
0
1
0
0

0
0
0
0
0
0
3
1
0
1
0
0
0
0
0
0
0
0
0
0

0
1
0
0
3
0
0
0
0
0
1
0
0
0
0
0
0
0
1
0

0
1
0
0
3
1
1
0
0
1
0
0
0
0
0
0
0
0
0
0

0
2
2
0
0
1
0
1
0
1
0
0
0
0
0
0
0
0
0
0

0
2
0
3
0
1
0
1
0
1
0
0
0
0
0
0
0
0
0
0

0
0
2
1
0
0
1
1
1
1
0
3
0
0
0
1
0
1
0
0

0
0
0
0
0
0
0
0
1
0
1
1
0
0
I
1
1
0
0
3

0
1
1
1
0
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0

0
0
0
0
1
0
0
0
0
1
0
2
3
1
1
0
0
0
0
0

0
0
0
0
0
0
0
0
3
1
0
2
1
1
0
3
0
0
0
1

0
0
0
0
0
0
0
0
0
0
0
3
1
2
0
0
0
0
0
1

0
0
0
0
0
0
0
0
0
0
0
1
1
2
1
0
0
0
0
0

2
1
0
0
0
0
0
0
0
1
0
1
0
0
1
0
3
0
0
2

0
0
1
0
0
0
0
0
1
1
0
0
3
0
0
0
0
1
0
1

0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
3
0
0
0
1

0
0
1
0
0
0
0
0
1
0
0
0
0
0
0
0
1
0
2
0

0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
2
3

0
0
0
0
0
0
0
0
0
3
0
0
1
1
0
2
1
1
0
3
-

a Subgradient algorithm, 5 machines, magazine capacity = 15.

W h e n t e s t i n g t h e s u b g r a d i e n t a l g o r i t h m o n p r o b l e m s as l a r g e as t h i s , it w a s d i s c o v e r e d t h a t t h e s t e p
size value significantly affects the convergence of the algorithm. Table 8 includes primal feasible
s o l u t i o n s f o r s t e p s i z e n u m e r a t o r v a l u e s r a n g i n g f r o m 0.5 t o 2, a n d h a l v i n g f r o m e v e r y t o e v e r y f o u r t h
i t e r a t i o n , f r is t h e n u m e r a t o r c o e f f i c i e n t a n d I I is t h e m u l t i p l e o f i t e r a t i o n s t o w a i t b e f o r e h a l v i n g f t . A n
i n t e r e s t i n g r e s u l t w a s t h a t t h e b e s t p r i m a l s o l u t i o n s o c c u r r e d o v e r v a r i o u s v a l u e s o f f~, w h i l e t h e b e s t
dual u p p e r b o u n d s o c c u r r e d w h e n f ~ = 2 a n d t h e s t e p size was h a l v e d every iteration. This result was
r e p e a t e d for t h e o t h e r test p r o b l e m s as well.
Occasionally the heuristic CMST found a better primal feasible solution than the subgradient
a l g o r i t h m f o r c e r t a i n v a l u e s o f f r a n d II. T h e h e u r i s t i c C M S T f a i l e d t o f i n d a f e a s i b l e s o l u t i o n f o r t w o o f
t h e six t e s t p r o b l e m s . A f e w r e o r d e r e d m a t r i c e s a r e p r e s e n t e d i n T a b l e s 9, 10, a n d 11.

Miscellaneous test problem

An additional test problem was created with various tool sizes and restrictions on the number of
m a g a z i n e s l o t s o c c u p i e d ( L f o r l o w e r a n d U f o r u p p e r b o u n d s ) . T a b l e 12 c o n t a i n s t h e i n p u t m a t r i x a n d
t w o s e t s o f t o o l sizes, R 1 a n d R 2 . T a b l e 13 c o n t a i n s c o m p u t a t i o n a l r e s u l t s f o r t o o l s i z e s e t R 1 , w i t h
v a r i o u s g r o u p s i z e s a n d m a g a z i n e r e s t r i c t i o n s . I t is i n t e r e s t i n g t o n o t e t h a t , w h i l e t h e b e s t p r i m a l s o l u t i o n

T.H. D'Alfonso, J.A. Ventura /Assignment of tools to machines

129

T a b l e 12
Input matrix for miscellaneous test problem
01

02

03

04

05

06

01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19

15
100
-

0
15
50

100
10
5
100
-

50
20
75
0
10

2 0

10
.
.
.
.
.
.

.
.
.
.
.
.

07

08

10
15
10
0
50
5
5
5
50
5
5
5
0
-.
. .
-.
.
..
- . - .
.

09

10

11

12

13

14

15

16

17

18

19

20

R1

R2

100
100
50
5
5
0
0
0
-

5
75
10
10
0
0
0
50
0

10
0
0
10
100
100
5
10
5
20
-

5
0
50
0
5
75
0
0
5
50
25

5
0
0
0
0
0
25
75
0
100
0
0
-

0
0
0
100
10
5
50
50
5
0

10
5
0
5
50
0
0
100
20
0
15
10
10
5

20
0
0
5
100
5
0
5
100
100
100
10
100
10
5

50
0
50
10
15
15
5
0
100
75
25
5
10
100
5
10

0
0
0
100
0
15
10
10
5
100

0
100
0
5
15
50
0
10
0
0

5
0
0
0
100
75
50
10
0
5

2
3
2
1
2
1
2
1
1
3

1
1
1
3
2
2
1
2
1
1

0
100
0
100
75
25
0

100
75
5
50
100
0
75

5
25
25
75
10
5
50

1
3
2
3
1
1
2

1
1
2
2
1
1
2

10

0
5

3
1
2

1
1
1

15
0
10
-

.
.

.
.

.
.

.
.

T a b l e 13
Computational results for miscellaneous test problem a

fr

0.50
0.50
0.50
0.50
0.75
0.75
0.75
0.75
1.00
1.00
1.00
1.00
2.00
2.00
2.00
2.00
Z~TER
Z~

ZDITE R
CMST *
LITER
UITER

II

1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4

p = 3

p = 4

p = 5

L=3
U = 20

L=3
U = 32

L=17
U = 20

L=17
U = 32

L=I
U = 9

L=I
U = 12

3195
2945
3095
3195
2920
3130
3140
3235+
3185
3235
2900
3115
3185-- c
3180
2990
3080
56
4749
85
2855

4190
4205
4190
4140
4200
4015
4200
4100
4200
4020
4205 +
4095
3655-3655
4095
4205
43
4213
4
3520

3195+ b
3095
3080
2990
2990
2870-3110
2860
3140
2905
3105
3010
2990
3010
3010
2935
88
4506
18
2855

3195+
3095
3080
3185
2990
3110
3110
2860
3140
2905
3105
3010
2985-3010
3010
2935
88
4440
58
2855

1395
1430
1535
1730
1445
1445
1280
1465
1295
1495
1560
1000
1225-1775
1430
1815+
73
4414
9
1020

1690
1810
1670
1675
1505
1920
1570
1595
1990
1830
1665
1780
1800-1450
2060+
1835
22
4558
97
1415

10
1

10
1

4
83

4
60

8
3

3
1

a p = n u m b e r o f g r o u p s , U = m a g a z i n e t o o l c a p a c i t y , f r = initial s t e p - s i z e c o e f f i c i e n t , II = m u l t i p l e Of i t e r a t i o n s t o h a l v e s t e p - s i z e
c o e f f i c i e n t , ZI*TER = s m a l l e s t n u m b e r o f i t e r a t i o n s r e q u i r e d to o b t a i n t h e b e s t p r i m a l s o l u t i o n f o r s u b g r a d i e n t a l g o r i t h m ,
Z ~ I T E R = s m a l l e s t n u m b e r o f i t e r a t i o n s r e q u i r e d to o b t a i n b e s t d u a l s o l u t i o n f o r s u b g r a d i e n t a l g o r i t h m , C M S T * = b e s t s o l u t i o n
d e t e r m i n e d b y C h o p t h e M a x i m a l s p a n n i n g T r e e h e u r i s t i c , LITER = n u m b e r o f l o w e r b o u n d i t e r a t i o n s f o r h e u r i s t i c , UITER =
n u m b e r o f u p p e r b o u n d i t e r a t i o n s f o r h e u r i s t i c , IN-F = n o f e a s i b l e s o l u t i o n w a s d e t e r m i n e d .
b +: This solution produces largest primal objective value (Z*).
c _ : This solution produces smallest Z~.

130

T.H. D'Alfonso, ZA. Ventura /Assignment of tools to machines

Table 14
Computational results for miscellaneous test problem a

fr
0.50
0.50
0.50
0.50
0.75
0.75
0.75
0.75
1.00
1.00
1.00
1.00
2.00
2.00
2.00
2.00
Z~TER
Z~
ZDITE R

CMST *
LZTER
UttER

II
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4

L=2

L = 13

U=16

U=24

U = 16

U=24

2920
2950
3075
2980
2880
3010
3050
2985
3090+
2730
3010
3045
2845
2940
3085
3030-95
4357
76
2855
2
9

4065
4065
4130
4095
4135
4065+
4135
4065
4065
4130
4215
4110
4215-4065
4130
4065
64
4881
33
3520
2
2

3170
2870
2875
2925
2870
2835
2310
3015
3035
3120
3015
3085- b
3050
2765
2845
3290+
26
4611
31
INF
20
0

3170
2870
2875
2925
2870
2835
3210
3015
3035
3120
3015
3085
3050
2765-2845
3290+
26
4388
30
INF
20
0

a p ~ number of groups, U = magazine tool capacity, fr initial step-size coefficient, II = multiple of iterations to halve step-size
coefficient, Z~-ER= smallest number of iterations required to obtain the best primal solution for subgradient algorithm,
ZD, ITER= smallest number of iterations required to obtain best dual solution for subgradient algorithm, CMST* = best solution
determined by Chop the Maximal spanning Tree heuristic, LITER= number of lower bound iterations for heuristic, UITER=
number of upper bound iterations for heuristic, INF = no feasible solution was determined.
b +: This solution produces largest primal objective value (Z*).
c _ : This solution produces smallest Z*.
=

v a r i e d greatly a m o n g the different step size p a r a m e t e r s , all b u t o n e of the best d u a l u p p e r b o u n d s

o c c u r r e d with a n u m e r a t o r p a r a m e t e r of 2, a n d halving every iteration.
I n T a b l e 14, the s e c o n d tool size set was e x a m i n e d with varying m a g a z i n e restrictions, b u t only two
groups ( p = 2) were f o r m e d in each case. T h e effect of having u p p e r a n d lower b o u n d s closer t o g e t h e r is
evident. F o r these cases, the p r i m a l solutions stayed the s a m e or i n c r e a s e d as the u p p e r b o u n d s
i n c r e a s e d by 50 p e r c e n t .
I n T a b l e 15, g r o u p sizes were i n c r e a s e d from 2, to sizes 3, 4, a n d 5. O n l y 2 slots were r e q u i r e d to b e
filled, as i n d i c a t e d by the L = 2 value. O n the average, p r i m a l solutions d e c r e a s e d by 12 p e r c e n t w h e n
t h e n u m b e r of groups i n c r e a s e d from t h r e e to four, a n d f u r t h e r d e c r e a s e d by 20.5% with a n a d d i t i o n a l
group. It is n o surprise that i m p o s i n g m o r e clusters results i n m o r e travel for parts.
Finally, T a b l e 16 c o n t i n u e d with the test p r o b l e m in T a b l e 14 a n d T a b l e 15, b u t five groups were
c o n s i d e r e d here. This f u r t h e r s u p p o r t e d the assertion that m o r e groups impose g r e a t e r p r o d u c t i o n travel.
It s h o u l d b e p o i n t e d o u t t h a t the s u b g r a d i e n t a l g o r i t h m was s u p e r i o r to the C M S T algorithm in most
cases. O n l y w h e n m a n y groups were to b e formed, a n d b o u n d s o n m a g a z i n e capacity were tight, did the
C M S T heuristic p r o d u c e a b e t t e r solution. It should also b e p o i n t e d o u t that a rule for d e t e r m i n i n g step
sizes to get a good p r i m a l s o l u t i o n is n o t forthcoming. T h e n u m b e r of times the best d u a l u p p e r b o u n d
was f o u n d with a step size n u m e r a t o r coefficient of 2, a n d halving the step size every iteration, should
n o t go u n n o t i c e d . This is a r e c o m m e n d a t i o n for f u t u r e a p p l i c a t i o n a n d testing.

T.H. D'Alfonso, J.A. Ventura /Assignment of tools to machines

131

Table 15
Computational results for miscellaneous test problem a

fr

0.50
0.50
0.50
0.50
0.75
0.75
0.75
0.75
1.00
1.00
1.00
1.00
2.00
2.00
2.00
2.00
Z~TER
Z~

II

1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4

Z D ITER

CMST *
LITER
UttER

p=3

p-~4

p=5

L=2
U = 16

L=2
U = 24

L=2
U = 16

L=2
U = 24

L=2
U = 16

L=2
U = 24

2805
2435
2655
2320
2245
2920+
2780
2390
2515
2620
2435
2480
2745
27452650
2635
43
4695
43
2605
3
7

4015- c
4260+ b
4015
3975
3495
4260
4035
3900
4115
3875
3875
3995
3975
4260
4115
4115
39
4847
55
3045
3
1

2670
2000
2600
2355
2455
2210
2330
2105
2890+
2695
2545
2530
2395
23852630
2390
71
4619
22
2410
4
4

3290
3200
2675
3240
3535
3275
3695
3375
3535
3355
3215
3345
3560
3840*
3265
3340
41
4811
69
2940
4
1

1575
1950
1640
2430+
2130
2230
1665
2155
2020
1860
1575
2310
22952185
2350
2130
27
4702
29
2305
5
3

2705
2180
2355
1935
1965
2990
3000+
2725
2255
2415
2985
2535
3415
23352850
2745
106
4867
61
2685
5
1

a p = number of groups, U = magazine tool capacity, f r = initial step-size coefficient, II = multiple of iterations to halve step-size
coefficient, ZI*TER = smallest number of iterations required to obtain the best primal solution for subgradient algorithm,
Z I ~ ITER = smallest number of iterations required to obtain best dual solution for subgradient algorithm, CMST* = best solution
determined by Chop the Maximal spanning Tree heuristic, LITER= number of lower bound iterations for heuristic, UITER=
number of upper bound iterations for heuristic, INF = no feasible solution was determined.
b +: This solution produces largest primal objective value (Z*).
c _ : This solution produces smallest Z~.

8. Conclusions
I n this article, a m a n u f a c t u r i n g system has b e e n s t u d i e d a n d a p a r t i c u l a r a s p e c t o f t h a t system has
r e c e i v e d special a t t e n t i o n . T h e d e c i s i o n o f l o c a t i n g tools in a flexible m a n u f a c t u r i n g system ( F M S ) can
h a v e t r e m e n d o u s i n f l u e n c e o n t h e efficiency o f t h e factory. P r o d u c t i o n flow is r e l a t e d directly to t h e
l o c a t i o n o f tools n e c e s s a r y to m a n u f a c t u r e a given part. F o r t h e F M S w i t h m a c h i n e s c a p a b l e o f h a n d l i n g
se ve r al tools, p r o d u c t i o n efficiency g r e a t ly d e p e n d s o n t h e p l a c e m e n t o f tools.
T h e m a t h e m a t i c a l m o d e l t h a t has b e e n c r e a t e d to o p t i m i z e t h e d e c i s i o n o f assigning tools in t h e F M S
system can b e a p p l i e d to o t h e r p r o b l e m s . T h e m o d e l is e q u i v a l e n t to g r o u p i n g objects o f varying sizes
into clusters w i t h size s p e c i f i c a t i o n s . C l u s t e r analysis has h a d w i d e a p p l i c a t i o n , an d t h e results o f this
study are a p p l i c a b l e to t h o s e p r o b l e m s t h a t r e s e m b l e t h e F M S p r o b l e m c h o s e n h e r e .
I n O p e r a t i o n s R e s e a r c h , p r o b l e m s a r e o f t e n c o m p l e x e n o u g h t h a t a g l o b a l o p t i m a l s o l u t i o n is
i m p o s s i b l e to d e t e r m i n e . In such cases, a g o o d m a t h e m a t i c a l m o d e l m a k e s all t h e d i f f e r e n c e in solving
t h e p r o b l e m as b e s t as possible. T h e c o m p l e x i t y o f t h e m o d e l d e v e l o p e d in this article is i n d e e d sizable
w h e n m a n y tools are to b e g r o u p e d i n t o s e v e r a l clusters. T h e L a g r a n g i a n d u a l a p p r o a c h an d t h e C h o p
t h e M a x i m a l S p a n n i n g T r e e ( C M S T ) h e u r i s t i c p r o v e d to b e suitable m e t h o d s o f d e t e r m i n i n g g o o d
soluti o n s to t h e p r o b l e m .

T.H. D'Alfonso, J.A. Ventura / Assignment of tools to machines

132

Table 16
Computational results for miscellaneous test problem a

fr
0.50
0.50
0.50
0.50
0.75
0.75
0.75
0.75
1.00
1.00
1.00
1.00
2.00
2.00
2.00
2.00
Z{TER
Z~
ZDIXER
CMST *
LITER
UrrER

II

1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4

L=1

L=4

U=7

U=9

U=7

U=9

1475
1470
1205
1270
1370
1475
1655+
1290
1460
1330
1470
1400
1580- b
1410
1450
1575
53
4557
79
610
4
88

1515
1515
1585
1645
1595
1545
1535
1615
1550
1675
1770
1625
1850-1560
1515
1860+
71
4515
53
1450
4
60

1375
1345
1325
1330
1530+ a
1475
1210
1280
1470
1295
1395
1335
1395-1245
1450
1510
62
4539
31
1700
15
1

1280
1035
985
1140
1530
1205
1445
1505
1210
1120
985
1315
1455
1390-1380
1545+
68
4510
28
1700
15
1

a p = number of groups, U = magazine tool capacity, f r = initial step-size coefficient, II = multiple of iterations to halve step-size
coefficient, Z~rER = smallest number of iterations required to obtain the best primal solution for subgradient algorithm,
.
ZD ~WER= smallest number of iterations required to obtain best dual solution for subgradient algorithm, CMST* = best solution
determined by Chop the Maximal Spanning Tree heuristic, LrrER = number of lower bound iterations for heuristic, UITER=
number of upper bound iterations for heuristic, INF = no feasible solution was determined.
b +: This solution produces largest primal objective value (Z*).
c _ : This solution produces smallest Zx~.

References
Auguston, J.G., and Minker, J. (1970), "An analysis of some graph theoretical clustering techniques", Journal of the Association for
Computing Machinery 17/4, 571-588.
Gallagher, C.C., and Knight, W.A. (1986), Group Technology Production Methods in Manufacture, Ellis Horwood, New York.
Goffin, J.C. (1977), "On the convergence rates of subgradient optimization methods", Mathematical Programming 13, 329-343.
Gould, R. (1988), Graph Theory, Benjamin/Cummings Publishing Company, Menlo Park, CA.
Held, M., Wolfe, P., and Crowder, H.O. (1974), "Validation of subgradient optimization", Mathematical Programming 6, 62-88.
Hyer, N.L. (1984), "Manager's guide to Group Technology", Operations Management Review 1/2.
Kennington, J.L., and Helgason, R.V. (1980), Algorithms for Network Programming, Wiley, New York.
Kusiak, A. (1990), Intelligent Manufacturing Systems, Prentice-Hall, Englewood Cliffs, NJ.
Lenstra, J.K., and Rinnooy Kan, A.H.G. (1975), "Some simple applications of the travelling salesman problem", Operational
Research Quarterly 26/4, 717-733.
McAuley, J. (1972), "Machine grouping for efficient production", The Production Engineer 51, 53-57.
McCormick, W.T., Schweitzer, P.J., and White, T.W. (1972), "Problem decomposition and data reorganization by a clustering
technique", Operations Research 20/5, 993-1009.
Mulvey, J.M., and Beck, M.P. (1984), "Solving capacitated clustering problems", European Journal of Operational Research 18,
339-348.
Mulvey, J.M., and Crowder, H.P. (1979), "Cluster analysis: An application of Lagrangian relaxation", Management Science 25/4,
329-340.
Ng, S.M. (1989), "A new spanning tree algorithm for group technglogy", Working Paper 1989-17, University of Southern
California, Los Angeles, CA.
Seifoddini, H. (1988), "Comparison between single linkage and average linkage clustering techniques in forming machine cells",
Computers and Industrial Engineering 15/1, 210-216.

T.H. D'Alfonso, L4. Ventura / Assignment of tools to machines

133

Shtub, A. (1988), "Capacity allocation and material flow in planning group technology cells", Engineering Costs and Production
Economics 13, 217-228.
Urquhart, R. (1982), "Graph theoretical clustering based on limited neighborhood sets", Pattern Recognition 15/3, 173-187.
Ventura, J.A., Chen, F.F., and Wu, C.H. (1990), "Grouping parts and tools in flexible manufacturing systems production planning",
International Journal of Production Research 28/6, 1039-1056.
Ventura, J.A., and Hsu, Y.H. (1989), "Minimization of part movements in flexible manufacturing systems", in: Proceedings of the
20th Annual Conference on Modeling and Simulations, Pittsburgh, PA May 1989, 447-452.