presentation13[1]
TRANSCRIPT
Layout design IV.Layout design IV.Ch 6Chapter 6
Layout generationCORELAPALDEPALDEPMULTIPLEMIPMIP
CORELAPCORELAP:: Computerized Relationship Layout PlanningComputerized Relationship Layout PlanningCORELAPCORELAP: : Computerized Relationship Layout PlanningComputerized Relationship Layout Planning
Developed for main frame computersAdjacency‐based method◦ CORELAP uses A=4, E=3, I=2, O=1, U=0 and X=‐1
lvalues
Selection of the departments to enter the layout is based on Total Closeness Ratinglayout is based on Total Closeness Rating◦ Total Closeness Rating (TCR) for a department is the sum of the numerical values assigned to the closeness relationships between the department and all other departments.
m
∑≠=
=m
jijijwTCR
,1
CORELAPCORELAPDepartment selectionDepartment selection
1 The first department placed in the layout is the one with the1. The first department placed in the layout is the one with the greatest TCR value. If there is a tie, then choose the one with more A’s (E’s, etc.).
2. If a department has an X relationship with the first one, it is l d l t i th l t If ti i t h th ithplaced last in the layout. If a tie exists, choose the one with
the smallest TCR value.3. The second department is the one with an A relationship with
the first one (or E, I, etc.). If a tie exists, choose the one with ( )the greatest TCR value.
4. If a department has an X relationship with the second one, it is placed next‐to‐the‐last or last in the layout. If a tie exists, choose the one with the smallest TCR valuechoose the one with the smallest TCR value.
5. The third department is the one with most A (E, I, etc.) relationships with the already placed departments. If a tie exists, choose the one with the greatest TCR value.
6. The procedure continues until all departments have been placed.
CORELAPDepartment placement
• Department neighbors 8 7 6Department neighborso Fully adjacent (in position 1, 3, 5 or 7) with department 0
o Touching (in position 2, 4, 6 or 8) department 0
5
432
1 0
• Placing rating (PR) is the sum of the weighted closeness ratings between the department to enter the layout and its neighbors.
placed}alreadysdepartment{where ==∑ kwPR k
• The placement of departments is based on the following steps:
1. The first department selected is placed in the middle.
placed}already sdepartment{ where∑ kwPRk
ik
p p
2. The placement of a department is determined by evaluating PR for all possible
locations around the current layout in counterclockwise order beginning at the
“ t d ”“western edge”.
3. The new department is located based on the greatest PR value.
CORELAP – Example 1
Given the relationship chart and the departmental dimensionsGiven the relationship chart and the departmental dimensions below determine the sequence of the placement of the departments in the layout based on the CORELAP algorithm. Place the departments in the layout while evaluating each placement.the departments in the layout while evaluating each placement.
Department Sizes Sq.ft. Num of Grids
1. Conf Room 100 2
2. President 200 4
3. Sales 300 6
4. Personnel 500 10
5. Plant Mng. 100 2
6. Plant Eng 500 10
7. P. Supervisor 100 2
8. Controller Office 50 1
9. Purchasing Dept 300 6
A=4 E=3 I=2 O=1 U=0 X= 1
CORELAP – Example 1A=4, E=3, I=2, O=1, U=0, X=‐1
Table of TCR Values
Dept. Department relationships Summary TCR Order
1 2 3 4 5 6 7 8 9 A E I O U X1 ‐ I I U O U U U U 0 0 2 1 5 0 5 712345
I I U O U U U U 0 0 2 1 5 0 5 7I ‐ O U O U U U O 0 0 1 3 4 0 5 8I O ‐ U I O O E U 0 1 2 3 2 0 10 5U U U ‐ O O O O O 0 0 0 5 3 0 5 9O O I O ‐ A A O O 2 0 1 5 0 0 15 15
6789
O O I O A A O O 2 0 1 5 0 0 15 1U U O O A ‐ I O E 1 1 1 3 2 0 12 2U U O O A I ‐ U O 1 0 1 3 3 0 9 3U U E O O O U ‐ I 0 1 1 3 3 0 8 6U O U O O E O I ‐ 0 1 1 4 2 0 9 49 U O U O O E O I ‐ 0 1 1 4 2 0 9 4
The placement sequence: 5‐6‐7‐9‐3‐8‐1‐2‐4
CORELAP – Example 1A=4, E=3, I=2, O=1, U=0, X=‐1
Both options gives the p gsame PR Score
PR = A[5,7]+I[6,7]
4 + 2 6= 4 + 2 = 6
If the location for the department 7 is chosen as pshown, the PR would bePR = A [5,7] =4
The placement sequence: 5‐6‐7‐9‐3‐8‐1‐2‐4
CORELAP – Example 1A=4, E=3, I=2, O=1, U=0, X=‐1
PR = E[6,9] = 3
PR = E[6,9]+O[5,9] = 3 + 1= 4
The placement sequence: 5‐6‐7‐9‐3‐8‐1‐2‐4
CORELAP – Example 1A=4, E=3, I=2, O=1, U=0, X=‐1
PR I O U (3 9) 2 1 0 3PR = I[3,5] + O[3,7] + U (3,9) = 2 + 1 + 0 = 3
PR = + E + I [8 9] = 3 + 2 = 5PR = + E[3,8] + I [8,9] = 3 + 2 = 5
The placement sequence: 5‐6‐7‐9‐3‐8‐1‐2‐4
CORELAP – Example 1A=4, E=3, I=2, O=1, U=0, X=‐1
PR = I[1 3] + U (1,7) = 2 + 0 = 2[1,3] ( , )
TCR = I[1,2] + I[2,3] = 2 +2 = 4
Continue with Department 4.
The placement sequence: 5‐6‐7‐9‐3‐8‐1‐2‐4
CORELAP – Example 2Given the relationship chart below and the sequence of theGiven the relationship chart below and the sequence of the
placement of the departments 7‐5‐9‐3‐1‐4‐2‐6‐8, find the best layout with CORELAP algorithm assuming that all the departments have the same size. Use these closeness values: A=125, E=25, I=5, O=1, U=0, X=‐
1. Receiving
125 and consider half weight if the departments are only touching by one point.
g
2. Shipping
3. Raw Materials Storage
AA
EO
EA
4. Finished Goods Storage
5. Manufacturing
UU
AO
E
E
A
U
U
A
A
O
O
UU
EE
U6. Work-In-Process Storage
7. Assembly
A
A
X
A
O
O
A E
A
E
UA
U
O
8. Offices
9. Maintenance
X
X
OA
CORELAPCORELAP –– Example 2Example 2
A=125, E=25, I=5, O=1, U=0, X=‐125
CORELAPCORELAP Example 2Example 2The placement sequence: 7‐5‐9‐3‐1‐4‐2‐6‐8
7
125
125 125
62.5 62.5
7 125
62.5 62.5
5125
187.5 187.5
125 62.562.5 62.5187.562.5 187.5
7 0
62.5 0
5187 5
62.5125
7 0
125.5 0
5
0.563.5
3125
62.5
7 05
187.5
187.5
9187.5
62 5 125 62 5
0
751.59126.5
0 5 1 0 5
0
362.5
62.5 125 62.5 0.5 1 0.5
CORELAPCORELAP –– Example 2Example 2
A=125, E=25, I=5, O=1, U=0, X=‐125
CORELAPCORELAP Example 2Example 2The placement sequence: 7‐5‐9‐3‐1‐4‐2‐6‐8
7 1255
100
337.5
12.5 62.5137.537.5
75
12.5
387.5
12.5
62.5
0025
137.5925 0
37.5 112.5 12.5
62.5 1259
125
137.5 1
62.5 125
4
125 62.5
1250 188 62.562.5
75
9
31
1
125
421
0.5
63.5
11 1.5 1.5 0.50.5
CORELAPCORELAP –– Example 2Example 2
A=125, E=25, I=5, O=1, U=0, X=‐125
CORELAPCORELAP Example 2Example 2The placement sequence: 7‐5‐9‐3‐1‐4‐2‐6‐8
0.5 10.5 0.5
75
-60.5
3112 5 11212 5
-61.525.5 612.5
753
6
875
9
3112.5
1
-112
4225
12.5
-37.5
75
9
3
1 42
8
7587.5 -62.5 -37.5 12.512.5
CORELAPCORELAP ‐‐ CommentsCommentsCORELAP CORELAP CommentsComments
The final layouts are evaluated by the distance‐based layout score◦ CORELAP uses the shortest rectilinear path pbetween the departments (not always realistic)
The layouts often result in irregular building shapes
ALDEPALDEP –– Automated Layout Design ProgramAutomated Layout Design ProgramALDEP ALDEP Automated Layout Design ProgramAutomated Layout Design Program
Similar to CORELAP (objectives, requirements)The main differences:◦ Randomness◦ Multi‐floor capabilityMulti floor capability◦ CORELAP attempts to produce the best layout, ALDEP produces many layoutsy , p y y
ALDEPALDEP ‐‐ ProcedureProcedureALDEP ALDEP ProcedureProcedure
Department selection◦ Randomly selects the first departmentRandomly selects the first department◦ Out of those departments which have “A” relationship with the first one
(or “E”, “I”, etc. – min level of importance is determined by the user) it selects randomly the second department
◦ If no such department exists it selects the second one completely◦ If no such department exists it selects the second one completely randomly
◦ The selection procedure is repeated until all the departments are selected
D t t l tDepartment placement◦ Starts from upper left corner and extends it downward◦ Vertical sweep pattern◦ Sweep width is determined by the user◦ Sweep width is determined by the userAdjacency‐based evaluation◦ If minimum requirements met, it prints out the layout and the scoresRepeats the procedure (max 20 layouts per run)Repeats the procedure (max 20 layouts per run)User evaluation
ALDEPALDEPVertical sweep pattern
Sweep widthDept size = 8 gridsDept size = 14 grids
1 grid
Dept. size = 8 gridsDept. size = 14 grids
2 grids
Dept. size = 14 grids
2 grids
Dept. size = 8 grids
3 grids
ALDEPALDEP ExampleExampleALDEP ALDEP ExampleExample
Use ALDEP procedure to determine the layout vector construct and evaluate the layout for thevector, construct and evaluate the layout for the facility based on the relationship chart and the departmental dimensions given below. Use the sweep width of 2 and the minimum acceptablesweep width of 2 and the minimum acceptable level of importance “E”. The closeness values: A=64, E=16, I=4, O=1, U=0, X=‐1024
Dept. Area # of unit area templates1 12,000 302 8000 203 6000 154 12,000 305 8000 206 12 000 306 12,000 307 12,000 30
ALDEPALDEP ExampleExampleALDEP ALDEP ExampleExample
Department selection
Step Department selected Reason for selection1 4 random2 2 “E” i h 42 2 “E” with 43 1 “E” with 24 6 random5 5 “A” with 66 7 random7 3 remaining
Layout vector◦ 4‐2‐1‐6‐5‐7‐3
ALDEP ALDEP ExampleExampleDept. # of unit area
templates1 30
2 20
Layout construction◦ Layout vector: 4‐2‐1‐6‐5‐7‐3S id h 2
3 15
4 30
5 20
6 30◦ Sweep width: 2 6 30
7 30
• Final layout
ALDEPALDEP ExampleExampleALDEP ALDEP ExampleExample
Adjacency score A=64, E=16, I=4, O=1, U=0, X=‐1024
MULTIPLEMULTIPLE ––Multi‐floor Plant Layout EvaluationMULTIPLE MULTIPLE Multi floor Plant Layout Evaluation
Construction and improvement algorithmDistance‐based algorithmSimilar to CRAFT (departments notSimilar to CRAFT (departments not restricted to rectangular shapes, discrete representation two‐way exchanges)representation, two way exchanges)But MULTIPLE can exchange non‐adjacent departmentsdepartmentsUses spacefilling curves to reconstruct a
l t ft h it tinew layout after each iteration
MULTIPLEMULTIPLE ‐‐ SpacefillingSpacefilling Curves (SFC)Curves (SFC)MULTIPLE MULTIPLE SpacefillingSpacefilling Curves (SFC)Curves (SFC)
Spacefilling curve connects all the grids in a layoutE h id i i it d tl◦ Each grid is visited exactly once
◦ Next grid visited is always adjacentto the current grid (only horizontal orvertical moves)vertical moves)
SFC is generated by the computerSFC allows MULTIPLE to
Map a layout vector into aMap a layout vector into a two‐dimensional layoutProcedure: ◦ The departments are placed based on the layout vector (similar as MCRAFT)
◦ The SFC is followed until the required number of grid for h d h deach department is reached
MULTIPLEMULTIPLE ‐‐ ExampleExampleMULTIPLE MULTIPLE ExampleExample
Create a MULTIPLE layout for the d b l b d h ldepartments below based on the layout vector 1‐2‐3‐4‐5‐6. Then build a new layout by exchanging the departments 1 and 5 Theby exchanging the departments 1 and 5. The facility and SFC are given below.
Department Area (m^2)1 162 82 83 44 165 85 86 12
MULTIPLEMULTIPLE ‐‐ ExampleExampleMULTIPLE MULTIPLE ExampleExample
Layout vector 1‐2‐3‐4‐5‐6Dep. Area1 162 8
2 3 4
2 83 44 165 82 3 46 12
511
6
MULTIPLEMULTIPLE ‐‐ ExampleExampleMULTIPLE MULTIPLE ExampleExample
Layout vector 1‐2‐3‐4‐5‐6Dep. Area1 16
2 3 4
2 83 44 165 82 3 4 5 86 12
511
6
MULTIPLEMULTIPLE ‐‐ ExampleExampleMULTIPLE MULTIPLE ExampleExample
Exchange 1 and 5 ‐ Layout vector 5‐2‐3‐4‐1‐6
MULTIPLEMULTIPLE ‐‐ ExampleExampleMULTIPLE MULTIPLE ExampleExample
Layout vector 5‐2‐3‐4‐1‐6 Dep. Area1 16
4 1
2 83 44 165 84 1
3
5 86 12
3
225 6
MULTIPLEMULTIPLE ‐‐ ExampleExampleMULTIPLE MULTIPLE ExampleExample
Initial layout
L t ftLayout afterthe exchange
MULTIPLEMULTIPLE ‐‐ Conforming CurvesConforming CurvesMULTIPLE MULTIPLE Conforming CurvesConforming Curves
Conforming curves are hand‐generated curvesThey are used:◦ If the building shape is irregular◦ If we want to capture the initial layout exactly◦ If we want to capture the initial layout exactly◦ If there are numerous obstacles (walls)◦ If there are fixed departments
Procedure:◦ May start and end at any grid◦ The curve visits all the grids assigned to a particular department before visiting other department◦ The fixed departments and obstacles are not visitedp
MULTIPLEMULTIPLE ‐‐ Conforming CurvesConforming CurvesMULTIPLE MULTIPLE Conforming CurvesConforming Curves
MULTIPLEMULTIPLEMULTIPLEMULTIPLEFinal MULTIPLE layout for the CRAFT exampleexample
The cost is lower than for the final layout f d b !found by CRAFT!◦ MULTIPLE is very likely to obtain lower‐cost solutions than CRAFT, since it considers a larger so ut o s t a C , s ce t co s de s a a geset of possible solutions at each iteration
MULTIPLEMULTIPLEMULTIPLEMULTIPLE
Final MULTIPLE layout for the CRAFT example may also need massaging to smooth the department borders
MULTIPLEMULTIPLE ‐‐ Construction algorithmConstruction algorithmMULTIPLE MULTIPLE Construction algorithmConstruction algorithm
Any SFC or conforming curves could be used to fill the vacant buildingAny vector can be used as the initial ylayout vectorAlternative layouts can be generated byAlternative layouts can be generated by trying different SFC◦ The cost may not be much different◦ The cost may not be much different
Optimal layout vector: D‐B‐H‐C‐F‐ED B H C F E
Cost z=54,200
Optimal layout vector:D‐E‐F‐B‐C‐H
Cost z=54 900Cost z=54,900
Optimal layout vector: D‐E‐F‐H‐B‐C
Cost z=54,540
MIP‐Mixed Integer Programming
◦ Generally construction type models◦ Could be used to improve a layout as welli b d bj i◦ Distance –based objective
◦ All the departments are assumed to be rectangularrectangular◦ Centroid and the length (width) of the department will fully define its location and p yshape
◦ Continuous representation
Obj i
MIP‐Mixed Integer Programming
Objective◦ Minimize the total cost/total distance travelledtravelled
Subject to◦ Rectangular department shapesg p p◦ The minimum and maximum sizes for the width and heights of departments◦ The minimum area requirement of◦ The minimum area requirement of departments◦ Departments are completely located inside the building◦ Departments do not overlap◦ All the departments are located on the floor◦ All the departments are located on the floor
MIP‐Mixed Integer Programming
• The notations used in formulations are:• The notations used in formulations are:
Parameters:
Decision variables:Decision variables:
MIP‐Mixed Integer ProgrammingY
),( '''ii yx ),( ''''
ii yx
• Mathematical Program Formulation * (αi, βi)),( ''ii yx ),( '''
ii yxBy
X
(1)
(2)
(3) XBX
Makes the model nonlinear
(4)
(5)
(6)
Make sure that departments do not
( )
(7)
(8)
(9)overlap
(9)
(10)
(11)
(12)(12)
(13)
MIP MIP ‐‐ ExampleExampleFormulate MIP model for the facility with the required dimension and coordinates ofdimension and coordinates of the departments below. Given is also Flow‐between chart and the costs of transfer among the
Flow-between Chart
the costs of transfer among the departments. The available space for the facility is 39 X 43 = 1677m^2.
Total cost• Solution:o Total cost calculation:
MIP MIP Example Example ‐‐ ResultsResults
X Y
A
Upper 35 7.14
Lower 0 0
B
Upper 27.49 12.14
Lower 7.49 7.14
C
Upper 27.49 22.14
Lower 7 49 12 14
22.14
C Lower 7.49 12.14
C
12.14
22.14
B
7.14
A
3527.497.49
MIPMIP‐‐ CommentsCommentsMIPMIP CommentsComments
Benefits of MIP Model◦ Department shapes as well as their areas can be modeled throughDepartment shapes as well as their areas can be modeled through
individually specified lower and upper limits◦ MIP might be able to control length‐width ratio as well
(xi’’ – xi’ ) <= R (yi’’‐yi’) or (yi’’ – yi’ ) <= R (xi’’‐xi’)(yi yi ) < R (xi xi )
◦ However, to use the MIP model effectively, some flexibility in the departmental specifications is required
May be used to improve a given layout as wellSolving the problem exactly (optimal solution) is hard◦ 7‐8 departments is the typical size solvable in a reasonable amount of
time◦ In practice, the number of departments easily exceeds 15‐18In practice, the number of departments easily exceeds 15 18Heuristically, CRAFT could be combined with MIP. ◦ Get an initial layout using CRAFT, use MIP to find the best rectangular
layout designA rectangle is not always an ideal shape for the department
Conclusion Conclusion Layout generation algorithms Layout generation algorithms
There is no commercial package that will suit all the needs, partly due to thesuit all the needs, partly due to the difficulty of the problem, but more due to the fact that facility layout is athe fact that facility layout is a combination of Science and Art.
Next lectureNext lectureNext lectureNext lecture
Quantitative methods