IE 484: Integrated Production Systems II

Spring 2024

Chapter 6

Mohamed Salama, Ph.D.,

Visiting Assistant Professor
School of Industrial Engineering,
Purdue University, West Lafayette, IN

Introduction
How to generate and evaluate alternative layout designs?

Facility layouts can be viewed at two levels:

1. The block layout
which shows the location, shape, and size of each planning department.
concerned primarily with the macro flows in the facility.
2. The detailed layout
which shows the exact location of all equipment, work benches, and storage
areas within each department.
concerned with the micro flows in the facility.

Our focus is primarily on quantitative methods for

developing and evaluating alternative block layouts.
Introduction (cont.)
Why do people tend to focus first on layout, not material handling?
Because of overemphasis on the manufacturing process.

Example:
If process B occurs immediately after process A, it seems perfectly logical to
place department B next to department A.
In such a situation, the handling problem is reduced to the question, What is
the best way to move materials from A to B? .
However, what if
Materials cannot flow directly from department A to department B
o WIP storage is required in A, B, and/or elsewhere.
o Depending on the storage and control requirements, a centralized WIP
storage area might be used.
o Thus, materials do not flow from A to B, and B no longer needs to be placed
next to A.
AGAIN, the layout and the handling system
should be designed simultaneously.
Within an iterative process!

Basic Layout Types

There are four basic types of layouts:
a) Production line product layout.
See Chapter #3
b) Fixed product layout.
slides 39-42
c) Product family layout.
d) Process layout.

Before generating and evaluating the overall facility layout alternatives,

the facilities planner must identify
the number of departments.
the type (among the listed above) of each department.

Example:
The overall facility layout includes
11 departments.
4 (with Process layout) + 5 (with Production line product layout)
+ 2 (with Product family layout).
Layout Procedures
The procedures can be classified into two main categories:
Construction procedures
for developing a new layout from scratch
Improvement procedures
for generating layout alternatives
by seeking improvements in an existing layout.

Which is common in practice?

Both, despite the popularity of the construction-type procedures!
Opinion of an expert:

Thus, both construction-type and improvement-type procedures are important.

Layout Procedures (cont.)

A number of different procedures have been developed (by scholars) to aid the
facilities planner in developing layout alternatives:

Plant Layout Procedure

Read only!
Plant Layout Procedure Pages 296 and 297 in the textbook.

Systematic Layout Planning (SLP) Procedure

It consists of three stages:
o Analysis
o Search
o Selection
 Layout Procedures (cont.)

(i.e., from-to chart) (i.e., activity relationship chart)

(Proximities are used to

reflect the relationship
between pairs of activities.)

Layout Procedures (cont.)

The preferred alternative is then identified and recommended.

Algorithmic Approaches
How to quantitatively develop a layout alternative?
Using an algorithmic approach.

There are many models and algorithms to help the layout analyst in developing or
improving a layout:
Pairwise Exchange Method
Graph-Based Method
CRAFT
BLOCPLAN
Read only!
MIP
Pages 326-332 in the textbook.
LOGIC
MULTIPLE

Algorithmic Approaches (cont.)

Algorithm Classification:
The algorithms can be classified based on
Type of input Objective Layout Construction/
data function representation Improvement

Some algorithms accept only

o qualitative data (such as a relationship chart),
o quantitative data expressed as a from-to chart.
Some algorithms (such as BLOCPLAN) accept both.
The charts are used only one at a time when evaluating a layout.
The common: algorithms that use a from-to chart.
o Flow values can be converted to relationship ratings and vice versa.
o Therefore, most algorithms can be used with either type of data.
o However, if a relationship chart is converted to a from-to chart, then the
values represent only an ordinal scale.
Algorithmic Approaches (cont.)
Algorithm Classification:
The algorithms can be classified based on
Type of input Objective Layout Construction/
data function representation Improvement

The number of all possible department pairs increases significantly with the
number of departments.
Relationship charts are not practical for problems with 20 or more departments.
From-to chart can be constructed using the production routing data, irrespective
of the problem size.

Algorithmic Approaches (cont.)

Algorithm Classification:
The algorithms can be classified based on
Type of input Objective Layout Construction/
data function representation Improvement

Two basic objectives:

1) minimizing the sum of flows distances
distance-based objective, suitable when the input data is expressed as a
from-to chart.

o : number of departments
o : flow from department to department
o : cost (or relative weight) of moving a unit load one distance unit from
department to department .
o : distance from department to department . (based on aisle structure)
Algorithmic Approaches (cont.)
Algorithm Classification:
The algorithms can be classified based on
Type of input Objective Layout Construction/
data function representation Improvement

2) maximizing an adjacency score.

adjacency-based objective, suitable for a relationship chart.

o : flow from department to department

o : 1 if departments and are adjacent (i.e., if they share a border) in the
layout, and 0 otherwise.

Algorithmic Approaches (cont.)

Algorithm Classification:
The algorithms can be classified based on
Type of input Objective Layout Construction/
data function representation Improvement

Example:

Discrete representation Continuous representation

A grid is used. No grid structure
Computational tradeoff Computationally difficult!
between small/large grid size.
Algorithmic Approaches (cont.)
Algorithm Classification:
The algorithms can be classified based on
Type of input Objective Layout Construction/
data function representation Improvement

Split departments:
a layout algorithm should not split a
department into two or more pieces.
human eye can easily identify split
departments, but computers need to
apply some measures to recognize
them.

Algorithmic Approaches (cont.)

Algorithm Classification:
The algorithms can be classified based on
Type of input Objective Layout Construction/
data function representation Improvement

Similar to our discussion on slide 4,

Construction algorithms
for developing a new layout from scratch
Improvement algorithms
for improving an existing layout.
Algorithmic Approaches (cont.)
Approach 1 (of 6): Exchange
An improvement-type layout algorithm.
i.e., Given an existing layout, it is required to develop an improved layout by
making feasible or realistic changes to the existing layout.
It can be used with both an adjacency-based and (more commonly) distance-
based objective.

Example:
An existing facility layout with four departments of equal sizes.
The existing layout is

The material flow data are as follows.

Algorithmic Approaches (cont.)

Solution:
A distance matrix can be obtained based on the existing layout:

The distance-based objective function value (or for the existing

layout is computed as follows:

The order of the departments in the initial layout.

Algorithmic Approaches (cont.)
Iteration 1:
The feasible exchanges are
1-2, 1-3, 1-4, 2-3, 2-4, and 3-4.
The distance matrix is recomputed each time an exchange is performed.
The layout costs resulting from the above exchanges are

The exchanged pair that results in the largest reduction in total cost is selected.

Algorithmic Approaches (cont.)

Iteration 2:
The feasible exchanges are the same.
i.e., 1-2, 1-3, 1-4, 2-3, 2-4, and 3-4.
The distance matrix is recomputed each time an exchange is performed.
The layout costs resulting from the above exchanges are

The exchanged pair that results in the largest reduction in total cost is selected.
Algorithmic Approaches (cont.)
Iteration 3:
The layout costs resulting from the feasible exchanges are

The lowest layout cost for the third iteration, 95, is worse than the current layout
cost of 90 (obtained in the second iteration).
Thus, the procedure is terminated.
The final layout arrangement is 2-3-1-4.

Note that the pairwise exchange method is

NOT guaranteed to yield the optimal layout.
Because the final layout is dependent on the initial layout.

Algorithmic Approaches (cont.)

Approach 2 (of 6): -Based
A construction-type layout algorithm.
It is often used with an adjacency-based objective.

Example:
A new facility consisting of 5 departments.
The relationship chart can be used to draw a relationship diagram.
In both charts, numerical rather than alphabetic closeness ratings.
Algorithmic Approaches (cont.)
For each of the following alternative block layouts:
o an adjacency graph is constructed.
o each node represents a department
o a connecting arc between two nodes indicating that two departments share a
common border.

Better
alternative
Maximally weighted block layout. layout!

An adjacency graph with the maximum sum of arc weights.

Algorithmic Approaches (cont.)

In general, for any adjacency graph,
The adjacency score does NOT account for
o distance
o relationships other than those between adjacent departments.
Dimensional specifications of departments are not considered.
The length of common boundaries between adjacent departments is not
considered.
The arcs do not intersect.
The total score is very sensitive to the assignment of numerical weights in the
relationship chart.

How to develop a maximally weighted adjacency graph?

Two alternative strategies:
Starting with the graph from the relationship diagram and iteratively developing
the maximally weighted adjacency graph.
The node insertion algorithm.
Algorithmic Approaches (cont.)
Step 1 (of 5)
o From the relationship chart, select the department pair with the largest
weight.
o Ties, if any, are broken arbitrarily.
o Thus, departments 3 and 4 are selected to enter the graph.

Step 2 (of 5)
o Select the third department to enter.
o The selection is based on the sum of the weights with respect to departments
3 and 4.
o Thus, department 2 is chosen with a value of 12+13=25.

Algorithmic Approaches (cont.)

Step 3 (of 5)
o Select the fourth department to enter.
o The selection is based on the value of adding one of the unassigned
departments on a face of the graph.
o The value of adding departments 1 and 5 is 27 and 9, respectively.
o Thus, department 1 is selected.
Algorithmic Approaches (cont.)
Step 4 (of 5)
o Determine on which face to insert department 5.
o For this step, department 5 can be inserted on faces
1-2-3 yields 0+7+0=7
1-2-4 yields 0+7+2=9
1-3-4 yields 0+0+2=2
2-3-4 yields 7+0+2=9
o Inserting 5 on faces 1-2-4 and 2-3-4 yields identical values of 9.
o We arbitrarily select 1-2-4.
o The final adjacency graph with a total sum of arc weights equals 81.

Algorithmic Approaches (cont.)

Step 5 (of 5)
o Using the adjacency graph to construct a block layout.
o The logic of the SLP method can be used.
Algorithmic Approaches (cont.)
Approach 3 (of 6):
Computerized Relative Allocation of Facilities Technique (CRAFT)

An improvement-type layout algorithm.

It uses a from-to chart as input data for the flow.
Layout is measured by the distance-based objective function:

Departments are not restricted to rectangular shapes.

The layout is represented in a discrete fashion (i.e., a grid is used).
CRAFT is flexible with respect to department shapes, as long as the department
is not split.

CRAFT is not guaranteed to yield the optimal layout.

How to potentially improve the solution quality?
By trying different initial layouts.

Algorithmic Approaches (cont.)

Procedure:
1) Determine the centroids of the departments in the initial layout.
2) Calculate the distance (typically rectilinear) between pairs of department
centroids.
3) Store the values in a distance matrix.
4) The initial layout cost is determined by summing the multiplication of
each entry in the from-to chart

the corresponding entry in the unit cost matrix

the corresponding entry in the distance matrix.

5) Consider all possible two-way (pairwise) or three-way department exchanges.
6) Identify the best exchange (i.e., the one that yields the largest reduction in the
layout cost.).
7) Repeat steps 1-6 for the new layout.

The process continues until no further reduction in layout cost can be obtained.
Algorithmic Approaches (cont.)
Simplification assumption for exchange :
When CRAFT considers exchanging two departments,
instead of examining all possible exchanges as we stated above,
it considers exchanging only those departments that are either
adjacent (i.e., that share a border) or equal in area.

What is the rationale?

Because the departments cannot be split, it would be impossible to exchange two
departments without the location of the other departments in the layout,
unless the two departments are either adjacent or equal in area.

However, if two departments are not equal in

area, then adjacency is a necessary but not
sufficient condition for being able to exchange
them without shifting the other departments.

Example: Think about exchanging departments

2 and 4 without splitting the departments.
(This is uncommon in practice)

Algorithmic Approaches (cont.)

Example:
A manufacturing facility with seven departments.
The department names, their areas, and the from-to chart:

Assume that all the values are set equal to 1.

The building and the current layout:
layout
o Each grid = 20 20 = 400 square feet.
o Total area = 72,000 square feet.
o Required for departments A-G =
70,000 square feet.
Algorithmic Approaches (cont.)
A common practice:
Depending on the amount of excess space available,
it is highly desirable to use two or more dummy departments
with more or less evenly distributed space requirements.
For practical reasons, we assume that the locations of the receiving (A) and
shipping (G) departments are fixed.
Initial layout cost (in units) = 2,974
o Example:
Rectilinear distance between
the centroids of departments A
and B = 6 units
Flow from A to B = 45.
Cost = 6 * 45 = 270 units
(we can compute actual cost = 6
20 45 1 = 5,400)
Repeat the calculation for all
department pairs.

Algorithmic Approaches (cont.)

Iteration 1
o Exchange departments E and F (after evaluating all two-way exchanges).
o Note that there are several ways in which departments E and F may be
exchanged.
o The reduction in the layout cost is equal to 21 units. (from 2,974 to 2,953)
Algorithmic Approaches (cont.)
Iteration 2
o Exchange departments B and C (after evaluating all two-way exchanges).
o The reduction in the layout cost is equal to 119.5 units.
(from 2,953 to 2,833.5)

Iteration 3
o No other (equal-area or adjacent) two-way exchange can further reduce the
cost of the layout
o Terminate with the obtained final solution.

Algorithmic Approaches (cont.)

Even if all the departments in the initial layout are regular,
the department shapes deteriorate with the number of iterations.
Thus, the final layout generated by the computer should
not be presented to the decision maker before the layout
planner molds or massages it into a practical layout.

Disregard the grids and use

a continuous representation!
Algorithmic Approaches (cont.)
Approach 4 (of 6):
BLOCPLAN can be used as
o a construction algorithm
o an improvement algorithm
It uses the continuous representation.
All the departments are rectangular in shape.
Improvements in the layout are through two-way department exchanges.
It uses either the following as input data:
o a relationship chart
o a from-to chart
The two charts can be used only one at a time when evaluating a layout.
Layout can be measured by
o the distance-based objective
o the adjacency-based objective

For the procedure (with an example),

see pages 323 and 324 in the textbook.
Read-only (Not required for IE 484)!

Algorithmic Approaches (cont.)

Approach 5 (of 6):
Layout Optimization with Guillotine Induced Cuts (LOGIC)

LOGIC can be used as

o a construction algorithm
o an improvement algorithm
It uses the continuous representation.
The departments generated by LOGIC are rectangular, provided that the
building is rectangular.
LOGIC can exchange two unequal-area and nonadjacent departments.
It typically uses a from-to chart as input data for the flow.
The layout is measured by the distance-based objective function.

How it works?
LOGIC is based on dividing the building into smaller and smaller portions by
executing successive guillotine cuts .

Straight lines that run from one

end of the building to the other.
Algorithmic Approaches (cont.)
Each guillotine cut is either a vertical cut or a horizontal cut.

i.e., a department is assigned either to i.e., a department is assigned either to

the east side of the cut or to the west. the north side of the cut or to the south.
How to determine the exact x-coordinate How to determine the exact y-coordinate
of the cut? of the cut?
Given Given
the building width (or a portion of it) the building length (or a portion of it)
the total area of all the departments the total area of all the departments
assigned to a side of the vertical cut. assigned to a side of the horizontal cut.

LOGIC executes a series of horizontal and vertical cuts.

With each cut, an appropriate subset of the departments is assigned to the
east west side of the cut OR north-south side of the cut

cut tree
execute the cuts and keep track of the departments.

Algorithmic Approaches (cont.)

Example:
A vacant building, with length and width = 360' 200', respectively.
It is required to plan a manufacturing facility with seven departments.
The department names, their areas, and the from-to chart:

First Cut:
Suppose the first cut is a vertical cut and Note that the cuts are
o Departments D, F, and G are assigned to its east. performed randomly in
Total area required = 36,000 square feet. this example.
o Remaining departments are assigned to its west.
In practice, a
Thus, the cut divides the building into two halves:
metaheuristic is used!
o Length and width = 180' 200', respectively.
Algorithmic Approaches (cont.)

LOGIC next treats each portion of the building as a by itself.

The procedure is repeated until each contains only one department.
Second Cut:
Consider first the that contains departments A, B, C, E, and H.
Suppose the next cut is a horizontal cut and
o Departments A and B are assigned to the north of the cut.
Total area required = 20,000 square feet.
o Departments C, E, and H are assigned to its south.
Since the length is 180 feet, the width of the which
contains departments A and B is 20,000/180 =111.11 feet.

Algorithmic Approaches (cont.)

Third Cut:
Consider next the that contains departments D, F, and G.
Suppose the third cut is a horizontal cut and
o Department G is assigned to the north of the cut
Area required = 12,000 square feet.
o Departments D and F are assigned to its south.
Since the length is 180 feet, the width of the which
contains department G is 12,000/180 = 66.67 feet.
Algorithmic Approaches (cont.)

Forth Cut:
Consider next the that contains departments A and B.
Suppose the fourth cut is a horizontal cut and
o Department A is assigned to the north of the cut
Area required = 12,000 square feet.
o Department B is assigned to its south.
Since the length is 180 feet, the width of the which
contains department A is 12,000/180 = 66.67 feet.

Algorithmic Approaches (cont.)

Fifth Cut:
Consider next the that contains departments C, H, and E.
Suppose the fifth cut is a vertical cut and
o Departments C and H are assigned to the west of the cut
Total area required = 8,000 square feet.
o Department E is assigned to its east.
Since the width is 88.89 feet, the length of the which
contains departments C and H is 8,000/88.89 = 90 feet.

Sixth Cut:
Consider next the that contains departments D and F.
Suppose the sixth cut is a horizontal cut and
o Department D is assigned to the north of the cut
Area required = 12,000 square feet.
o Department F is assigned to its south.
Since the length is 180 feet, the width of the which
contains department D is 12,000/180 = 66.67 feet.
Algorithmic Approaches (cont.)

Seventh (and last) Cut:

Consider next the that contains departments C and H.
Suppose the fourth cut is a horizontal cut and
o Department C is assigned to the north of the cut
Area required = 6,000 square feet.
o Department H is assigned to its south.
Since the length is 90 feet, the width of the which
contains department C is 6,000/90 = 66.67 feet.

Algorithmic Approaches (cont.)

1st cut
v: vertical cut

Cut-tree

N: North side 7th cut

S: South side h: horizontal cut
W: West side
E: East side
Algorithmic Approaches (cont.)
How the LOGIC method is used as an algorithm?
Example:
Exchanging departments D and E in the last layout to potentially improve.

Solution:
In the Cut-tree, replace all the Ds with Es and vice versa.
Compute the new x - and y-coordinates of the cuts.

Notice the different

dimensions of all departments.
Compare CRAFT with LOGIC in exchanging adjacent departments!

Algorithmic Approaches (cont.)

Approach 6 (of 6):
Multifloor Plant Layout Algorithm (MULTIPLE)

An improvement-type layout algorithm.

It uses a from-to chart as input data for the flow.
Layout is measured by the distance-based objective function.
It was originally developed for multiple-floor facilities.
However, it can also be used in single-floor facilities simply by setting the
number of floors equal to one. (In IE 484, only a single floor is covered)
Departments are not restricted to rectangular shapes.
The layout is represented in a discrete fashion (i.e., a grid is used).
MULTIPLE is similar to CRAFT except for
o exchange procedure.
o layout formation. (using Spacefilling curves; SFC)
To construct a layout when any
two departments are exchanged.
Algorithmic Approaches (cont.)
Hilbert Curve
Layout vector:
1-2-3-4-5-6

Layout Vector

Fill sequence

How to exchange any two departments?

By
exchanging their locations in the layout vector.
reassigning the grids.

Example: exchanging departments 1 and 5.

Layout vector: 5-2-3-4-1-6