INE 333

Facility Design and Operations

Lecture 5: Computerized Layout

Procedures
Computerized Layout Procedures

 Once a site is selected, the preparation of data for the

layout planning process is initiated.
 From this initial step, one can begin to proceed into
developing detailed facility layout schematic alternatives.
 Working with the overall plan and the schematic layout
alternatives, one can develop an evaluation process to
compare and choose the best alternative.
 After performing this evaluation, one can use computer
software for generating detailed layouts of the plant and
operations.
 Finally, given this detailed layout plan development, one
continues with the implementation of the best alternative
Facilities Planning and Design, by Alberto Garcia-Diaz and J. MacGregor Smith, Pearson Education Limited 2014
A facility design project may be
viewed as a task in network
design comprised of two
fundamental problems:
(1) the generation or selection of
the topological configuration
of the finite set of nodes (V) and
finite set of arcs (A) which define
the planar graph G(V, A);
(2) the optimal assignment of the
activities to the nodes of G(V, A);
Layout software
(1) an optimal seeking
branch-and-bound
program, called
MAFLAD; and
(2) A network-based
heuristic steepest-descent
procedure, called STEP.
(3) An improvement
heuristic based on another
steepest-descent procedure,
FLAP, is related to CRAFT.
Facility layout software issues
1. Optimal vs. Suboptimal Solutions?
 one must carefully determine whether or not optimal solutions are desired
 QAP problem is NP complete - computing time
2. Equal vs. Unequal Areas?
 QAPs are equal-area problems. If unequal-area problems- complexity of
achieving optimal solutions skyrockets
3. Existing vs. New Facilities?
 Existing facilities often have fixed locations of activities, immovable walls
and columns, and fixed-position equipment, which makes the activity
layout much more difficult.
4. Deterministic vs. Stochastic Flows?
 From–to chart and material handling flows are deterministic. When they
are stochastic, different model assumptions must be accounted for

Facilities Planning and Design, by Alberto Garcia-Diaz and J. MacGregor Smith, Pearson Education Limited 2014
Facility layout software issues
5. Quantitative Flow Data vs. Qualitative REL-Data?
 QAP-type Approaches Vs graph-theoretic models
6. Grid Decomposition vs. Continuous Space Decomposition?
 Grid or that the activities can be in any location within the perimeter of
the building
 Continuous space decomposition will require nonlinear programming
approaches rather than combinatorial optimization approaches.
7. Explicit Shape vs. Rectangular Activity Shape?
 Controlling the shape of an activity in the layout normally requires a
discrete or combinatorial optimization approach.
8. Layout and Material Handling Integration?
 very desirable feature in layout planning but makes the problem even
more difficult
9. Multistory Layouts?
 generalized quadratic assignment (GQAP) type problems
Facilities Planning and Design, by Alberto Garcia-Diaz and J. MacGregor Smith, Pearson Education Limited 2014
 Layout software
(1) an optimal seeking
branch-and-bound
program, called
MAFLAD; and
(2) A network-based
heuristic steepest-descent
procedure, called STEP.
(3) An improvement
heuristic based on another
steepest-descent procedure,
Facilities Planning and Design, by Alberto Garcia-Diaz and J. MacGregor Smith, Pearson Education Limited 2014
FLAP, is related to CRAFT.
Exact QAP

a flow placement term. a linear placement cost term

for locating activity i in
location k

Facilities Planning and Design, by Alberto Garcia-Diaz and J. MacGregor Smith, Pearson Education Limited 2014
Exact Graph-Theoretic Procedures

 Graph-theoretic layout and network-based procedures

(GTLN) are a natural methodology for developing facility
layout plans.
 given ordinal scale data such as a relationship (REL)
matrix,

NP complete –
as in the QAP, exact
solutions are
difficult to generate.

Facilities Planning and Design, by Alberto Garcia-Diaz and J. MacGregor Smith, Pearson Education Limited 2014
Key differences between the QAP and the
graph-theoretic (GTLN) models

 The GTLN model is more suitable for new than for existing
layouts, whereas the QAP approaches can account for
fixed area and the shape of activities and predetermined
locations.
 The GTLN model ignores the linear placement term in the
QAP model, so it is somewhat less general than QAP
models.
 GTLN approaches include only directly adjacent activities
in the objective function, whereas QAP approaches
include nonadjacent pairs.

Facilities Planning and Design, by Alberto Garcia-Diaz and J. MacGregor Smith, Pearson Education Limited 2014
 Block plan layout

planar graph solution

Facilities Planning and Design, by Alberto Garcia-Diaz and J. MacGregor Smith, Pearson Education Limited 2014
Heuristic Procedures

 Information need generally consist of :

 from–to chart, flow matrices, relationship charts;
 area requirements for departments;
 in some cases, cost information on material handling
 Heuristic procedures
 improvement heuristics
require a feasible layout as input and continue to modify the
layout by swapping areas and scoring the revised layouts until no
further improvement can be found
 construction heuristics
 graph-theoretic and network-based procedures.

Facilities Planning and Design, by Alberto Garcia-Diaz and J. MacGregor Smith, Pearson Education Limited 2014
Improvement Algorithm –
Computerized Relative Allocation of Facilities Technique (CRAFT)

 The goal of the CRAFT program is to minimize the total transportation

cost.
 The transportation cost for a particular move between two
departments is defined as the product of the number of trips by the
corresponding distance and then by a specified cost per unit distance.
 Pros
1. Widely used in practice, even though somewhat dated, and forms the
basis of many other layout approaches.
2. Can deal with fixed activity locations and thus incorporate considerations
of obstacles and impediments in the layout.
3. Flexible with different activity shapes and also can accommodate a
variety of plant shapes [32].
4. Utilizes dummy activities to deal with unrealistic activity alignments [28].

Facilities Planning and Design, by Alberto Garcia-Diaz and J. MacGregor Smith, Pearson Education Limited 2014
Improvement Algorithm –
Computerized Relative Allocation of Facilities Technique (CRAFT)

 Cons
1. Final solution is very dependent on the initial starting solution, so multiple starting
solutions are recommended.
2. Results are not as effective when activities have different sizes, and, furthermore,
activities can be split into nonadjacent cells [32].
3. Considers exchanging only those activities that are either adjacent (share a
boundary wall) or equal in area [64].
4. A “steepest-descent” heuristic and thus can overlook optimal exchanges [64].
 Outline of Procedure
1. Compute centroids for departments in the initial layout.
2. Create a distance matrix between centroids.
3. Compute transportation costs of initial layout.
4. Consider interchanges of department with equal area or with common borders.
5. Select the interchange with the greatest cost reduction.
6. Compute cost and repeat the procedure until no further reductions in cost are
obtained.

Facilities Planning and Design, by Alberto Garcia-Diaz and J. MacGregor Smith, Pearson Education Limited 2014
Quadratic Set Packing Models–
Multi-Attribute Facility Layout and Design (MAFLAD)

 QSP assumes that the system of interest has the form of a grid layout, where clusters
of grid cells associated with the different activity locations are known in advance.
This way, different shapes and sizes of activities can be taken into account as long as
they can be described by a set of cells within the grid layout.
 Program interacts with the user, allowing the user to request a graphics
display of the solution and to select one of three heuristics to be used in the
branch-and-bound process.
 the total area required to arrange all activities is divided into q blocks.

Facilities Planning and Design, by Alberto Garcia-Diaz and J. MacGregor Smith, Pearson Education Limited 2014
 Quadratic Set Packing Models–
Multi-Attribute Facility Layout and Design (MAFLAD)

each activity is assigned to exactly one possible

cluster within the grid layout

a cell is not a
member of more than one cluster of a
given activity

With

q number of blocks into which the total area occupied by all facilities is divided into
Ii number of potential locations for facility i
Ji(j) set of blocks occupied by facility i if it is assigned to location j
d(ji, lk) distance between the centroids of locations j and l if facility i is assigned to
location j and facility k is assigned to location l a major disadvantage lies in the sharp
xij equals 1 if facility i is assigned to location j and 0 otherwise increase in the problem size as the total
pijt equals 1 if block t is an element of Ji(j) and 0 otherwise area is divided into smaller blocks.
Facilities Planning and Design, by Alberto Garcia-Diaz and J. MacGregor Smith, Pearson Education Limited 2014
A Relaxed Assignment Formulation of the QSP (see
GMFLAD read me file)

Quadratic Set
Packing (QSP)
branch and bound
algorithm for solving
an exact solution
A Relaxed Assignment Formulation of the
QSP (see GMFLAD read me file)
 QSP formulation of the layout or planning problem resembles the more
general quadratic assignment problem, but differs in several respects
 Predetermined activity alternatives In particular, the QSP formulation
presented above is based on the assumption that the study region is
tesselated by a Cartesian grid and that clusters of grid activities representing
alternate locations for each activity are predetermined.
 Multi-attribute utility objective function The objective function in the QSP
model is comprised of two terms. In the first term, site-placement utility values
occur; in the second, quadratic utility flow terms occur. Maximization of utility is
very different than the minimization of distance or costs that is typical in most
QAP models.
 Constraints The zero–one decision variables and the coefficients of utility in
the linear term of the objective function are based on known clusters of
activities indexed by t for each activity k. Individual activities belonging to a
cluster are designated by the zero-one coefficients in the first set of
constraints; that is, is 1 if activity i is a member of the tth cluster for activities k,
and 0 otherwise. This type of set packing model formulation captures the one-
activity, one-site constraints of the QAP model.
STEP ALGORITHM
 A hybrid of a graph-theoretic and improvement approach using an underlying
Steiner circulation flow graph to estimate the congestion in the layout, and a
sampling design scheme to perturb the arrangement of the activities
 The Sampling Test and Pairwise Exchange Procedure (STEP) is an algorithm
for constructing a heuristic solution where stochastic or deterministic flows can
be modeled.
 It assumes that the material handling system is captured in a network as
depicted on the next slide.
 The Steiner node is termed a circulation passage.
 In this node, the dynamic flow of people, parts, or vehicles is represented.

Facilities Planning and Design, by Alberto Garcia-Diaz and J. MacGregor Smith, Pearson Education Limited 2014