Outline
GE Energy Wind Turbine Supply Chain Modeling
Abhinav Agarwal Sanjay G. Charles Sam HolsterRatnakar N. Pawar Sunil Ravichandran
Richa Rastogi Michael Y. Thelen
H. Milton School of Industrial and Systems EngineeringGeorgia Institute of Technology
April 8th, 2008
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
Outline
Outline
1 Introduction
2 Optimization ModelObjective FunctionConstraints
3 Results
4 ConclusionsSummary of FindingsFuture Possibilities
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
Outline
Outline
1 Introduction
2 Optimization ModelObjective FunctionConstraints
3 Results
4 ConclusionsSummary of FindingsFuture Possibilities
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
Outline
Outline
1 Introduction
2 Optimization ModelObjective FunctionConstraints
3 Results
4 ConclusionsSummary of FindingsFuture Possibilities
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
Outline
Outline
1 Introduction
2 Optimization ModelObjective FunctionConstraints
3 Results
4 ConclusionsSummary of FindingsFuture Possibilities
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Introduction
Project Scope
GE is one of the world’s leading wind turbine suppliers withover 8,400 worldwide wind turbine installations comprisingmore than 11,300 MW of capacity.
GE assembles 1.5 to 3.6 MW complexes worldwide.
Customers are almost exclusively utility companies.
As of last year, GE Wind Energy is a $2-2.5B enterprise peryear.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Introduction
Project Scope
GE is one of the world’s leading wind turbine suppliers withover 8,400 worldwide wind turbine installations comprisingmore than 11,300 MW of capacity.
GE assembles 1.5 to 3.6 MW complexes worldwide.
Customers are almost exclusively utility companies.
As of last year, GE Wind Energy is a $2-2.5B enterprise peryear.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Introduction
Project Scope
GE is one of the world’s leading wind turbine suppliers withover 8,400 worldwide wind turbine installations comprisingmore than 11,300 MW of capacity.
GE assembles 1.5 to 3.6 MW complexes worldwide.
Customers are almost exclusively utility companies.
As of last year, GE Wind Energy is a $2-2.5B enterprise peryear.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Introduction
Project Scope
GE is one of the world’s leading wind turbine suppliers withover 8,400 worldwide wind turbine installations comprisingmore than 11,300 MW of capacity.
GE assembles 1.5 to 3.6 MW complexes worldwide.
Customers are almost exclusively utility companies.
As of last year, GE Wind Energy is a $2-2.5B enterprise peryear.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Introduction
Goals
Need a strategic tool to bridge the gap between wind turbinesupply and demand.
Optimization engine to automate assignments.
Implications: database and GUI design to integrate businessintelligence and optimal resource allocation.
Adaptive interface to contend supply chain networkuncertainties.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Introduction
Goals
Need a strategic tool to bridge the gap between wind turbinesupply and demand.
Optimization engine to automate assignments.
Implications: database and GUI design to integrate businessintelligence and optimal resource allocation.
Adaptive interface to contend supply chain networkuncertainties.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Introduction
Goals
Need a strategic tool to bridge the gap between wind turbinesupply and demand.
Optimization engine to automate assignments.
Implications: database and GUI design to integrate businessintelligence and optimal resource allocation.
Adaptive interface to contend supply chain networkuncertainties.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Introduction
Goals
Need a strategic tool to bridge the gap between wind turbinesupply and demand.
Optimization engine to automate assignments.
Implications: database and GUI design to integrate businessintelligence and optimal resource allocation.
Adaptive interface to contend supply chain networkuncertainties.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Wind Turbine
Turbine
Hub
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
ObjectiveConstraints
Outline
1 Introduction
2 Optimization ModelObjective FunctionConstraints
3 Results
4 ConclusionsSummary of FindingsFuture Possibilities
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
ObjectiveConstraints
Objective Function
Min: z = TotalTransportationCostss,w ,p + TotalLDCostsp
where
TotalTransportationCostss,w ,p =∑s∈ Supplier
∑w∈ Week
∑p∈ Project
Supplys,w ,p · TransportationCosts,p
andTotalLDCostp =
∑p∈ Project
LDCostp
for supplier s, week w , and project p.Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
ObjectiveConstraints
Outline
1 Introduction
2 Optimization ModelObjective FunctionConstraints
3 Results
4 ConclusionsSummary of FindingsFuture Possibilities
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
ObjectiveConstraints
Constraints
Net Inventory Constraint
ExcessNetInventoryp,w ,c − NetInventoryShortagep,w ,c =
ExcessNetInventoryp,w−1,c − NetInventoryShortagep,w−1,c+
ArrivingInventoryp,w ,c − Demandp,w
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
ObjectiveConstraints
Quantity Arriving at Customer Base p at Week w
ArrivingInventoryp,w ,c =∑
s∈ Supplier
τs,p,w
where
τs,p,w =
Supplys,w− LeadTimes,p ,p, if w − LeadTimes,p ≥ 1
0, otherwise
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
ObjectiveConstraints
Constraints
For each customer, the total inventory arriving must exceed thetotal demand.∑
w∈Week
ArrivingInventoryp,w ,c ≥∑
w∈Week
Demandp,w
Production must be kept within production capacity specifications.
Productions,w ≤ ProductionCapacitys,w
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
ObjectiveConstraints
Constraints
For each customer, the total inventory arriving must exceed thetotal demand.∑
w∈Week
ArrivingInventoryp,w ,c ≥∑
w∈Week
Demandp,w
Production must be kept within production capacity specifications.
Productions,w ≤ ProductionCapacitys,w
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
ObjectiveConstraints
Constraints
SupplierInventorys,w = SupplierInventorys,w−1 + Productions,w−∑p∈ Project
Supplys,w ,p
∑p∈Project
Supplys,w ,p ≤ Productions,w + SupplierInventorys,w−1
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
ObjectiveConstraints
Constraints
SupplierInventorys,w = SupplierInventorys,w−1 + Productions,w−∑p∈ Project
Supplys,w ,p
∑p∈Project
Supplys,w ,p ≤ Productions,w + SupplierInventorys,w−1
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
ObjectiveConstraints
Constraints
Maximum Early Delivery
ExcessNetInventoryp,w ,c ≤Ω∑
w ′=w+1
Demandp,w ′
where Ω = minw + MaximumEarlyDelivery, ω and ω = theweek number of the GE’s horizona.
aFor our project, GE’s horizon is defined as week #104
Supplier Stock Inventory Capacity
SupplierInventorys,w ≤ StorageCapacitys
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
ObjectiveConstraints
Constraints
Maximum Early Delivery
ExcessNetInventoryp,w ,c ≤Ω∑
w ′=w+1
Demandp,w ′
where Ω = minw + MaximumEarlyDelivery, ω and ω = theweek number of the GE’s horizona.
aFor our project, GE’s horizon is defined as week #104
Supplier Stock Inventory Capacity
SupplierInventorys,w ≤ StorageCapacitys
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
ObjectiveConstraints
Constraints
The Switchp would capture the liquidated damages costs perproject into the portion below the cap and the portion above.
LDCapp+ Switchp = LDRatep×∑
w∈ Week
maxcNetInventoryShortagep,w ,c
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
ObjectiveConstraints
Constraints
LDCostPerProjectp = BinarySwitch · LDCapp+
(1− BinarySwitch) · LDCostPerProjectp
where BinarySwitchp = ISwitch(p)≥0
LDCostPerProjectp ≤ 4 · LDCapp
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
ObjectiveConstraints
Constraints
LDCostPerProjectp = BinarySwitch · LDCapp+
(1− BinarySwitch) · LDCostPerProjectp
where BinarySwitchp = ISwitch(p)≥0
LDCostPerProjectp ≤ 4 · LDCapp
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
User Interface Demonstration
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Breakdown of Costs
Proportion of Liquidated Damages
297170032, 99%
2140850, 1%
Total Transportation Cost
Total Liquidated Damages
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Breakdown of Costs–After Random Supplier Removed
Proportion of Liquidated Damages
295078678, 98%
5421322, 2%
Total Transportation Cost
Total Liquidated Damages
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Supplier Shortages
Comparison of LD Costs
0
100000
200000
300000
400000
500000
600000
1-HAGHL
1-EN
K06
1-MQ1D
T_Q1
1-9F
U6L
1-P9
MU5-Q2 5
5024
85
5044
27
1-C4Z
S4 (2
)
5036
64
1-CH8X
E
5039
07
5043
82
5031
61
Projects
LD Costs
All Suppliers
Random Sample
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Breakdown of Costs–After Major Suppliers Removed in CA and FL
Proportion of Liquidated Damages
35120000, 8%
3.92E+08, 92%
Total Transportation Cost
Total Liquidated Damages
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Capacity Utilization Adjustments
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Summary of FindingsFuture Possibilities
Outline
1 Introduction
2 Optimization ModelObjective FunctionConstraints
3 Results
4 ConclusionsSummary of FindingsFuture Possibilities
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Summary of FindingsFuture Possibilities
Conclusions
Summary of Findings
Optimization module complete. Able to effectively allocateresources to minimize costs.
Developed a GUI to incorporate business strategy to makebetter decisions from optimal results.
Google Earth application to provide spatial visualization to aidin manual assignment of resources.
Application is robust to business growth.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Summary of FindingsFuture Possibilities
Conclusions
Summary of Findings
Optimization module complete. Able to effectively allocateresources to minimize costs.
Developed a GUI to incorporate business strategy to makebetter decisions from optimal results.
Google Earth application to provide spatial visualization to aidin manual assignment of resources.
Application is robust to business growth.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Summary of FindingsFuture Possibilities
Conclusions
Summary of Findings
Optimization module complete. Able to effectively allocateresources to minimize costs.
Developed a GUI to incorporate business strategy to makebetter decisions from optimal results.
Google Earth application to provide spatial visualization to aidin manual assignment of resources.
Application is robust to business growth.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Summary of FindingsFuture Possibilities
Conclusions
Summary of Findings
Optimization module complete. Able to effectively allocateresources to minimize costs.
Developed a GUI to incorporate business strategy to makebetter decisions from optimal results.
Google Earth application to provide spatial visualization to aidin manual assignment of resources.
Application is robust to business growth.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Summary of FindingsFuture Possibilities
Outline
1 Introduction
2 Optimization ModelObjective FunctionConstraints
3 Results
4 ConclusionsSummary of FindingsFuture Possibilities
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Summary of FindingsFuture Possibilities
Conclusions
Future of User Interface
Continue to develop Google Earth interface and other spatialvisualization tools.
Tracking function to store manual changes made over time.
Manually override optimization results.
Building function to maintain customer and supplierrelationships, in terms of a long term horizon.
Completely integrate Access VB for mass customization.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Summary of FindingsFuture Possibilities
Conclusions
Future of User Interface
Continue to develop Google Earth interface and other spatialvisualization tools.
Tracking function to store manual changes made over time.
Manually override optimization results.
Building function to maintain customer and supplierrelationships, in terms of a long term horizon.
Completely integrate Access VB for mass customization.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Summary of FindingsFuture Possibilities
Conclusions
Future of User Interface
Continue to develop Google Earth interface and other spatialvisualization tools.
Tracking function to store manual changes made over time.
Manually override optimization results.
Building function to maintain customer and supplierrelationships, in terms of a long term horizon.
Completely integrate Access VB for mass customization.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Summary of FindingsFuture Possibilities
Conclusions
Future of User Interface
Continue to develop Google Earth interface and other spatialvisualization tools.
Tracking function to store manual changes made over time.
Manually override optimization results.
Building function to maintain customer and supplierrelationships, in terms of a long term horizon.
Completely integrate Access VB for mass customization.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Summary of FindingsFuture Possibilities
Conclusions
Future of User Interface
Continue to develop Google Earth interface and other spatialvisualization tools.
Tracking function to store manual changes made over time.
Manually override optimization results.
Building function to maintain customer and supplierrelationships, in terms of a long term horizon.
Completely integrate Access VB for mass customization.
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Summary of FindingsFuture Possibilities
Questions?
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling
IntroductionModel
ResultsConclusions
Summary of FindingsFuture Possibilities
Thank You
Agarwal, Charles, Holster, Pawar, Ravichandran, Rastogi, Thelen GE Energy Wind Turbine Supply Chain Modeling