1

Simulation OptimizationDecision Support System forShip Panel Line Operations

Winter Simulation Conference Case Study TrackWashington, DC December 2004

Charles LaRue Ingalls Operations, Northrop Grumman Ship SystemsDonny Dorsey Avondale Operations, Northrop Grumman Ship SystemsAllen Greenwood Department of Industrial Engineering, Mississippi State UniversityTravis Hill Center for Advanced Vehicular Systems, Mississippi State UniversityJeffrey Miller Center for Advanced Vehicular Systems, Mississippi State University Clay Walden Center for Advanced Vehicular Systems, Mississippi State University

2

Broad objective: Maximize shipyard throughput, subject to customer due date

• Build ships faster and cheaper

• Increase throughput of the yard and sector; increase profit

• Reduce lead time • Improve the use of key

resources• Employ best practices• Effectively deal with

variability

Problem*: U.S. shipyards take twice as long to build comparable ships; 1/3 as productive as the Japanese, 1/2 as productive as the Europeans

*Liker, J. K., and Lamb, T. A Guide to Lean Shipbuilding, Draft Version 0.5, June 26,2000, Maritech ASE project #10 TIA 2000214, p. 9

3

Focus on the shipyard bottleneck: Panel Shop

D42

D41

D31 D21 D11

D32D22

D23

D12

D24

D25

D26

A21A21

A22

A23

A15A24

A16A25

A26

A27 A17

A31

A32A33A43

A42

A41

A34

A35A45

A44

A36A46

B11

B13

B21

B12B22

B23B33

B14

B15B25

B24

B31B41

B42B32

B43

B34

B35B45

B44

D51D61D71

D52

D64D63

D72

D65

C22C32

C34

C42

C52

C43

C44C54C53

C55 C45

C25

C26

C11

C12

P:\Const_Supp\DDG 89 BLOCK

C13

C14

C15

C23C33

C24

6/7/00

C27D42

D41

D31 D21 D11

D32D22

D23

D12

D24

D25

D26

A21A21

A22

A23

A21A21

A22

A23

A15A24

A16A25

A26

A27 A17

A31

A32A33A43

A42

A41 A31

A32A33A43

A42

A41

A34

A35A45

A44

A36A46

A34

A35A45

A44

A36A46

B11

B13

B21

B12B22

B11

B13

B21

B12B22

B23B33 B23B33

B14

B15B25

B24 B14

B15B25

B24

B31B41

B42B32

B31B41

B42B32

B31B41

B42B32

B43B43

B34

B35B45

B44

D51D61D71

D52

D64D63

D72

D65

C22C32

C34

C42

C52

C43

C44C54C53

C55 C45

C42

C52

C43

C44C54C53

C55 C45

C42

C52

C43

C44C54C53

C55 C45

C25

C26

C11

C12

P:\Const_Supp\DDG 89 BLOCK

C13

C14

C15

C23C33

C24

6/7/00

C27

4

Every panel is unique extreme variabilityin work content

FCAW

0500

100015002000250030003500

1 11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161

SSAW

0200400600800

100012001400

1 11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161

Topside SAW

0

500

1000

1500

2000

2500

1 11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161

Backside SAW

0200400600800

10001200140016001800

1 11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161

Layout

0

200

400

600

800

1000

1200

1 11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161

Edge Grind

0100200300400500600700

1 11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161

5

Project overview

• Components:– Discrete-event simulation model of panel shop– Optimizer to determine best sequence for producing

panels– DSS so the simulation model and optimizer could be

used by planners and shop floor supervisors

• Objective: provide a means to understand and assess the impact on shop performance of changes in:– resources,– operations practices,– panel characteristics,– sequence, etc.

6

Overview of Simulation-OptimizationDecision Support System

7

ProModel simulation model captures shop behavior

Model considers:• Panel size and

conveyor capacity• Work content• Resource availability• Work assignments• Operational rules• Downtime• Task variability• Shift schedule• Relevant measures

of performance

Model runtime: approximately 5 secondsto process 154 panels (~13 weeks in real time)

8

Model accurately captures shop behavior

Hours to complete is based on observation; the number of panels that had exitedat a specified time; e.g., at time 697, 52 panels had been competed.

Model Complete is the time a panel left the system in the model; e.g. Panel 52was completed at time 681.

0

200

400

600

800

1000

1200

1400

1600

1800

2000

11 23 28 33 37 42 49 54 61 70 76 82 86 95 102

111

120

123

128

135

141

Panel #

Hou

rs

Hours to Complete

Model Complete

Panel 150: Model 1729 –Actual 1734

Hours to complete is based on observation; the number of panels that had exitedat a specified time; e.g., at time 697, 52 panels had been competed.

Model Complete is the time a panel left the system in the model; e.g. Panel 52was completed at time 681.

0

200

400

600

800

1000

1200

1400

1600

1800

2000

11 23 28 33 37 42 49 54 61 70 76 82 86 95 102

111

120

123

128

135

141

Panel #

Hou

rs

Hours to Complete

Model Complete

Panel 150: Model 1729 –Actual 1734

9

Workstation processing times based on work standards and panel characteristics

min/seam min/ft seams feet TimeSweep 3 5 15Flux 5 5 25Wire 12 5 60Align 30 5 150Console 13 5 65Weld 0.83 143.2 119Traverse return Console 6 4 24 Traverse 0.054 143.2 8Remove ram 2 5 10Remove plate 6 5 30Slag chips 2 5 10Defect repair 0.72 143.2 103

619

Standards Panel DDG 356

10

Simulation model incorporates dynamic resource assignments

Panel Weld Time (min.)

Assign 4 welders(based on work in

process ahead)

Production flow

Panelsequence

11

Optimal sequence based on genetic algorithm

NGSS fitness over generations

0.531 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89 93 97 101

Generations

Fitn

ess Best

Average

• Modified evolutionary strategy• Fitness function

– based on total weld feet, make span, days late for each job– value is evaluated for each combination using the simulation

model..• DSS manages optimization process, including evaluation of each

solution by the discrete-event simulation model• Sample run for a set of 50 panels

12

Example analysesPercent change in makespan (time to complete panel set)

Machine Utilization

100 90 80 70 60

100 6.0% 5.8% 5.7% 5.2% 4.4%

95 3.9% 3.9% 3.8% 3.2% 2.7%

85 0.0% -0.6% -0.6% -1.6% -1.1%

70 -7.7% -7.4% -8.3% -8.5% -9.0%Base Case: Personnel Utilization = 85%, Machine Utilization = 100%

Process Variability

none -5/+10 -10/+20 -25/+50 -25/+100

100 6.5% 6.0% 5.4% 3.0% -4.8%

85 -0.1% 0.0% -1.2% -3.4% -11.7%

70 -7.1% -7.7% -8.4% -11.6% -20.0%Base Case: Personnel Utilization = 85%, Process Variability = -5% / +10%

Pers

onne

l U

tiliz

atio

nPe

rson

nel

Util

izat

ion

13

Basic DSS architecture

• Support planner-level and shop-floor-level decisions– Easy-to-use interfaces– Intuitive and relevant output– Model operations transparent to

users• Driven by NGSS data; responsive to

changes in data• Sequence:

– based on shop-floor behavior, capabilities, and constraints

– performance assessed using simulation model

– generated by genetic algorithm• Provides work assignments required

to meet optimal sequence

Planner Shop Floor Admin

PerformanceMeasures

GenerateSchedule

ProductData

ForecastedPanels

WorkContent

Standards

OperatingConditions /Parameters

SatisfactorySolution ?

OptimizerSystemController

Yes

No

SequencingStrategies

User Interfaces

ShopOperations

Model(Simulation)

14

DSS interface

Panelselection Optimal

sequence

Operationsparameters

*run in ProModel or QUESTModel*Selection

15

DSS output

Schedule

ResourceAllocation

Performance

16

Future directions: application across a sector

Decision Makers

30

Real System

ModelingDecision SupportSystem

Panel Line 225FabricationSuppliersCSA,

Outfitting,etc.

Customers

Models