Scribd
Upload
9 views

Pres2

The document presents a 0-1 quadratic model for optimally solving the machine allocation problem in group technology, which clusters machines to process similar part families. The model incorporates various manufacturing factors and is solved using algorithm A*, although it faces challenges with larger problems. Computational experiences indicate that while the algorithm can solve smaller instances efficiently, it struggles with larger datasets.

Uploaded by

chc25a
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
9 views

Pres2

The document presents a 0-1 quadratic model for optimally solving the machine allocation problem in group technology, which clusters machines to process similar part families. The model incorporates various manufacturing factors and is solved using algorithm A*, although it faces challenges with larger problems. Computational experiences indicate that while the algorithm can solve smaller instances efficiently, it struggles with larger datasets.

Uploaded by

chc25a
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

SOLVING THE MACHINE ALLOCATION PROBLEM

OPTIMALLY FOR GROUP TECHNOLOGY

ChunHung Cheng, School of Business, Kentucky State Univ., Frankfort, KY 40601

ABSTRACT hjj'=cost of intercellular movement between


machines j and j'
A 0-1 quadratic model is formulated for
forming machine cells. The model considers Suppose that there are parts using machines j
many manufacturing factors and is solved and j'. The cost of intercellular movement
optimally by algorithm A*. between machines j and j' when either one of
machines is not assigned to the same
INTRODUCTION machine cell is defined as:

Group technology (GT) groups parts into (1)


families based on similarity in their design
features and/or operations. Then it clusters
different machines in machine cells to process
part families. The problem is formulated in the following
model:
There are two approaches to implement GT
[1]. In matrix formulation, algorithms were
(2)
proposed to cluster a binary machine-part
incidence matrix. In integer programming
formulation, models have been built to s.t.
incorporate different design objectives and
constraints.
(3)
A 0-1 quadratic model for forming machine (4)
cells is presented in this paper. Although the
model is similar to the one in [2], it considers (5)
many manufacturing factors. The model is
solved optimally by algorithm A*
The objective function in model (2)-(6)
THE MACHINE ALLOCATION MODEL
(6)
The model is concerned with the allocation of
machines to machine cells. To model the minimizes total cost of intercellular moves.
problem, define:
Constraint (4) ensures that a machine is
k=cell index, k = 1,..,p assigned to exactly one machine cell.
i=part index, i = 1,..,n Constraint (3) specifies the maximum number
j=machine index, j = 1,..,m of machines allowed in a machine cell.
di=demand of part i
qi=size of unit handling load of part i Reference [2] formulated a 0-1 quadratic
gi=cost of handling a unit load of part i out of model but did not consider cost. Our model
a cell for an operation explicitly considers demand of parts, cell
Gk=maximum number of machines in cell k capacity, and cost of intercellular movement
aij=1if part i requires machine j of parts. Hence, our model is more
0otherwise comprehensive.
Xjk=1if machine j is assigned to cell k
0otherwise
The formulated model is a 0-1 quadratic 1.Place the initial node s on a list called OPEN.
model with linear constraints. The model Set g(s) = 0 and f(s) = 0.
cannot be solved in polynomial time. Heuristic 2.Remove from OPEN node q whose f value is
techniques are the only way to solve large smallest and place it on a list called
problems efficiently. However, an optimal CLOSED. (Resolve ties arbitrarily).
solution technique is needed to evaluate 3.If q is a goal node, exit with the optimal
heuristic techniques. solution; otherwise continue.
4.Branch from node q, generating all of its
SOLVING THE MODEL successors. Reject a successor node if it
contains a partial solution violating
In this section, algorithm A* is used to solve constraint (4). For each feasible successor
the problem. qi, compute g(qi), h(qi), and f(qi).
5. Go to Step 2.
The state description of a node in a search
tree is composed of k sets, each of which COMPUTATIONAL EXPERIENCE
corresponds to a machine cell. The possible
elements for the k sets are machine numbers. The developed algorithm was implemented in
The initial state can be represented by k Turbo PASCAL on a IBM PC-286 compatible
empty sets. and was used to solve several problems.

The expansion of a node q in a tree is done by One test problem with 25 parts and 10
inserting an unassigned machine j to any one machines required 37.62 CPU seconds to
of the k sets at a time in node q and therefore solve. However, the algorithm did not
creates k different successor nodes for node terminate within 30 CPU minutes on problems
q. Since a machine is assigned to a machine involving 17 machines. Obviously, the
cell at a node, constraint (4) is always algorithm is not very effective for large
satisfied. Also a node which contains a partial problems.
solution violating constraint (3) will be pruned.
A goal state is reached when all machines CONCLUSIONS
have been assigned.
In this paper, a 0-1 quadratic model for the
Algorithm A* requires the selection of an machine allocation problem was formulated.
evaluation function f(q) = g(q) + h(q). The The model considered many manufacturing
function g(q) is the minimum cost incurred in factors and was solved optimally by algorithm
a path from the initial node to node q. Assume A*.
that q machines have been assigned at node
q. The g(q) is defined as: REFERENCES

(7) 1.Cheng CH (1992) Algorithms for grouping


machine groups. Omega 20(4),
forthcoming.
The heuristic function h(q) is an estimate of 2.Kusiak A, Boe JW, and Cheng CH (1992)
the minimum cost incurred in a path from Designing cellular manufacturing
node q to a goal node. If the heuristic function systems: Branch-and-bound and A*
is an under-estimate, optimal solutions are approaches. IIE Trans., forthcoming.
guaranteed [3]. One such heuristic function is 3.Nilsson NJ (1981) Principles of Artificial
given as follows: Intelligence. Tioga Publishing, California.

(8)

The developed algorithm is given as follows:

You might also like