m. grunow technical university of denmark, copenhagen national tsing hua university, hsinchu, taiwan...
TRANSCRIPT
M. GrunowM. Grunow Technical University of Denmark, Technical University of Denmark,
CopenhagenCopenhagen
National Tsing Hua University, Hsinchu, Taiwan December 13, 2007
Supply Network Planning andSupply Network Planning and
Advanced Planning SystemsAdvanced Planning Systems
H.-O. GüntherH.-O. GüntherTechnical University of Berlin, Technical University of Berlin,
GermanyGermany
OutlineOutline
Introduction
Advanced planning systems
Case 1: Global network design
Case 3: Production planning and scheduling
Case 2: Supply network planning
Case 4: Value chain management
Outlook: Teaching SNP & APS
Supply Chain ManagementSupply Chain Management
www.wikipedia.org:
“The primary objective of supply chain management is to fulfill customer demands through the most efficient use of resources.“
www.wikipedia.org:
“A supply chain, logistics network, or supply network is a coordinated system ...
… of entities, activities, information and resources involved in moving a product or service from supplier to customer.“
Supply Chain ManagementSupply Chain Management
Introduction
Advanced planning systems
Case 1: Global network design
Case 3: Production planning and scheduling
Case 2: Supply network planning
Case 4: Value chain management
OutlineOutline
Outlook: Teaching SNP & APS
Generations of PPC softwareGenerations of PPC software
1960s
1970s
1980s
1990s
2000s
• Predecessors of PPC systems Focus on inventory control Basic order processing
• Material requirements planning (MRP) Bill of material files Calculation of net requirements
• Manufacturing resources planning (MRP II) Enhanced planning functions Integration of financial accounting and management functions
• Integrated systems CIM: Integration of manufacturing ERP systems covering the whole enterprise
• Advanced Planning and Scheduling systems Integration into supply chain management concept Use of true optimization techniques
Advanced Planning Systems (APS)Advanced Planning Systems (APS)
APS modulesAPS modules
Strategic Network Design
Supply Network Planning
DemandPlanning
External Procurement
Production Planning / Detailed
Scheduling
Transportation Planning /
Vehicle Scheduling
Order Fulfilment ATP / CTP
Strategic Network Design
Production Planning / Detailed
Scheduling
Supply Network Planning
long-term
mid-term
short-term
Industrial case studiesIndustrial case studies
Case 1: Global network design for the production of
electrical components
Case 2: Multi-site production-distribution
planning in the chemical industry
Case 3: Characteristics
dependent planning in the
consumer goods industry
Case 4: Value chain management
OutlineOutline
Introduction
Advanced planning systems
Case 1: Global network design
Case 3: Production planning and scheduling
Case 2: Supply network planning
Case 4: Value chain management
Outlook: Teaching SNP & APS
Case 1: Global network designCase 1: Global network design
Strategic Network Design
DemandPlanning
External Procurement
Production Planning / Detailed
Scheduling
Transportation Planning /
Vehicle Scheduling
Supply Network Planning
long-term
mid-term
short-term Order FulfilmentATP / CTP
Case 1: Global network design for the production of
electrical components
Grunow, M., Günther, H.O., Burdenik, H., Alting, L.:Evolving production network structures.CIRP Annals (to appear)
Strategic network designStrategic network design
Strategic Network Design
Mathematical methods Heuristics, MILP, Clustering techniques
Determination of transportation links
Assignment of locations to each other e.g. customers to DCs
Decisions
Number of plants and DCs
Assignment of products to plants
Locations and capacities
APS based on generic model
formulations provide limited support for strategic network
design.
Case 1: InsightsCase 1: Insights
Redesign of global network also affects the organizational structure and
business processes.Potential cost
savings achieved through quantitative
modeling underestimated by decision makers
(engineers).
ProductsProducts
3 Levels Vacuum interrupters
Kits (for local content and taxation reasons)
Switches (usually containing 3 vacuum interrupters)
Supply networkSupply network
Current centralized network Production sites at
Head production centers (produce vacuum interrupters): GER (with R&D) and CHINA 1
USA, INDIA, CHINA 2
MEX, BRA (from kits only)
Supply networkSupply network
Future network
Reduced production complexity at head production centers Increased flexibility through networked supply and distribution
Reduced production and logistics costs
Supply networkSupply network
Product relocation depends on Product maturity
new products at sites with R&D department
Personnel
required for product relocation (to qualify the new production facility for the production of the new products).
Complexity and flexibility Diversity of products at facility increases scheduling complexity Products at multiple facilities increase network complexity Production close to customer markets Exchange rate risks and taxation
Production capacities
test runs and ramp-up of new products
Supply networkSupply network
Product relocation depends on Product maturity
new products at sites with R&D department
Personnel
required for product relocation (to qualify the new production facility for the production of the new products).
Complexity and flexibility Diversity of products at facility increases scheduling complexity Products at multiple facilities increases network complexity Production close to customer markets Exchange rate risks and taxation
Production capacities
test runs and ramp-up of new products
Error !Not covered !
Standard commercial APS
MILP optimization modelMILP optimization model
1
,,,,,,
,,,,,,,,
,,,,,,,,
,,,,,,,,,,,,
)1/(1
t
t
f gtfgtfg
ftftf
k hpctkpchpck
kpctkpchpckkpchpck
i hpctpchpci
pctpchpcipchpci
p f ctcfptcfptcfp
ptfp
ffp
ir
zFCwIC
DCCS
DCCS
CDSCVC
Min
Minimization of the relevant production and logistics costs variable manufacturing costs (including material and personnel) costs for shipments to customers (+ customs duty) costs for shipments of intermediates and kits from head production
centers to other facilities in the network investment costs required for the expansion of the production
capacities for final products costs associated with the complexity in a facility net present value calculation
MILP optimization modelMILP optimization model
Demand
Production
(final products)
tcpf
tcfp D ,,,,,
c
tcfstfs ,,,,,
kpc
tkpchpckc
tchpckthpck ,,,,,,,,
KSp
thpcppipc
tpchpcic
tchpcithpci A ,,,,,,,,,,,
(kits)
(intermediates)
Shipment of intermediates
tkpcshpc
tkpchpck ,,,,,
s
tpcssihpc
tpchpci A ,,,,,,
(kits)
(intermediates)
MILP optimization modelMILP optimization model
Capacities
(final product capacity incl. expansions, here: testing equipment)
(intermediates capacity)
t
ttfftf
stfs wECS
1',,,,
ft
tf FEw ,
tf
tf NEw ,
thpcithpci CI ,,,,
thpck
thpck CK ,,,
(no. of expansions per facility)
(no. of expansions in network per perioddue to personnel requirements, budget constraints)
(kit capacity)
MILP optimization modelMILP optimization model
Relocation
(continuous-binary production variables)
(number of relocations per facility and period, due to capacity requirements for test runs and ramp-ups)
(relocation-production)
(number of total relocations per period, due to personnel requirements)
(ban on relocation of products in R&D phase)
tfsfftftfs xEVFECS ,,,,,
1
0',,,,,,
t
ttfstfstfs xxy
Nys f
tfs ,,
fs
tfs By ,,
01
,, f
TR
ttfs
s
y
M
MILP optimization modelMILP optimization model
Complexity
(production of group)
(flexibility, limitation of exchange risks, demonstration of goodwill, marketing)
(network complexity)
(facility complexity)
gSs
tfstfg xMz ,,,,
tg f
tfg CNz ,,
tfg
tfg CFz ,,,
p p Cc
tcptptrFf
tfptp
rr
DLVLSLV ,,,,,,,
Local production
1
,,
,,,,,,,
,,,
,,,,,
,,,,
,,,,,
,,,,,,,,,
,,
)1/(1)_
)___(
(
tf
ff g
tfg
ftftfpchpcpfp
pchpcptpchpcp
cfpfpcfp
cfppchpcp
tpchpcppchpcp
cfpcfptcfptfp
fpfp
t
itcoefficiencomplexity
initgroupgrproducefacility
ECIIDSMCTR
DSMCTRISSC
DSSCpVMCMin
MILP optimization modelMILP optimization model
Advantages of optimization models Consideration of interdependencies between entities in the SC.
Quick and easy evaluation of scenarios
Provide understandable quantitative results (e.g. cost savings) as basis for final management decisions.
Manager: „Sometimes we make strategic
SND decisions very intuitively.“
Exemplary resultsExemplary results
Scenarios Current business plan Scenario 1: Actual configuration with optimized production/distribution Scenario 2: Reassignment of products with fixed capacity Scenario 3: Reassignment of products with capacity adjustments
Results Savings in production costs
overcompensate additional logistics costs.
Shift of production volumes and products from Germany to China
Increased capacity in Mexico and China
Direct manufacturing costs
Plan Szenario 1 Szenario 2 Szenario 3 Plan Szenario 1 Szenario 2 Szenario 3 Plan Szenario 1 Szenario 2 Szenario 3
2008 2009 2010
Plan Szenario 1 Szenario 2 Szenario 3 Plan Szenario 1 Szenario 2 Szenario 3 Plan Szenario 1 Szenario 2 Szenario 3
2008 2009 2010
OutlineOutline
Introduction
Advanced planning systems
Case 1: Global network design
Case 3: Production planning and scheduling
Case 2: Supply network planning
Case 4: Value chain management
Outlook: Teaching SNP & APS
Case 2: Case 2: Supply network planningSupply network planning
Strategic Network Design
DemandPlanning
External Procurement
Production Planning / Detailed
Scheduling
Transportation Planning /
Vehicle Scheduling
Supply Network Planning
long-term
mid-term
short-term Order FulfilmentATP / CTP
Case 2: Multi-site production-distribution
planning in the chemical industry
Supply Network Planning
Supply network planningSupply network planning
Consideration of production, transportation, and handling capacities as hard constraints
Decisions
Allocation of production quantities between plants
Smoothing out seasonal cycles in demand
Supply from the plants to DCs and from the DCs to customers
Mathematical methods LP and MILP, heuristics
Consideration of demand, due dates, and safety stocks as soft constraints
Considerable cost savings achieved
through application of optimization
model.
Case 2: InsightsCase 2: Insights
Generic model formulations embedded in SNP modules of APS
do not consider application-specific
features.
High degree of acceptance by management.
DEGUSSA AG, Düsseldorf, Germany
Company profileCompany profile
World largest producer of special chemicals
Subsidiaries in all continents
46.000 employees / 300 plants worldwide
Turnover in 2005: 11,800 billion €
Production process Carbon BlackProduction process Carbon Black
Production volume: several 100.000 t per annum
100 product specifications
Continuous production process
Feed A
Feed B
Feed C
Process Mainproduct Silo
Gas
Trans-formation
Energy
Silo Silo
Implementation of an APSImplementation of an APS
Company negotiates annual volumes with key customers.
Forecasting the Forecasting the period distribution period distribution of annual demandof annual demand
Forecasting the Forecasting the period distribution period distribution of annual demandof annual demand
Allocation of Allocation of customer demands customer demands to production sitesto production sites Customers request
deliveries upon short notice.
Demand planningDemand planning
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
20000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
Month
t
Observed demand
Forecast incl. seasonality
1997 1998 1999
Adaptation of Winters forecasting technique
Considerably increased forecast accuracy
Forecast represents network-wide commitment
Forecasting the Forecasting the period distribution period distribution of annual demandof annual demand
Allocation of Allocation of customer demands customer demands to production sitesto production sites
Allocation of Allocation of customer demands customer demands to production sitesto production sites
Decisions in supply network planningDecisions in supply network planning
Transportation volumes between production sites and customers
Production volume in the production sites and at each production train
Generation of energy from side-products
Objectives in Supply Network PlanningObjectives in Supply Network Planning
Transportation costsTransportation costs(Distance, carrier)(Distance, carrier)
Inventory costsInventory costs
t,1Train,GER,pPC
t,2Train,GER,pPC
t,3Train,GER,pPC
t,1Train,RSA,pPC
t,2Train,RSA,pPC
Production costsProduction costs(site, train)(site, train)
(negative) energy refund(negative) energy refund
Minimization ofMinimization ofMinimization ofMinimization of
Customer DemandCustomer Demand
Limited substitutability of products Delivery only from sites, which have a
customer approval for the product
A customer demand may only be covered by deliveries from a limited number of sites
vpctvpcASss
tvpcs CDAT ,,,,,}p) c,,s(|{
,,,,
p,c,sv,p,ct,v,p,c,s BCDT
Selected constraints: DistributionSelected constraints: Distribution
Selected constraints: EnergySelected constraints: Energy
Energy Transformation capacity of the sites
Transformation
tss
sTRstrPp
tpstrspstrs ATMaxCPPVCP ,
)()(,
,),(,),(,
Graphical user interface
User and scenario administration
Solver(CPLEX)
MILPmodels(OPL)
CO
M-
Op
tio
n
SQL
DataSQL
Data base OPL Studio
Optimization software architectureOptimization software architecture
Application of the optimization modelApplication of the optimization model
Currently numerous managers are using the tool for operative planning.
Rolled out in Europe, US and Asia.
Estimated financial benefitper year from supply network planning exceeded project costs by far.
Scenario mode is used extensively, e.g., for capacity decisions, evaluation of approvals and of the profitability of energy transformation.
Further benefits arise from improved supply network design.
Performance improvement in optimizationPerformance improvement in optimization
Average algorithmic improvement 1988 2001
Estimated machine improvement 1988 2003: 800 x Net improvement for LP problems 1988 today: ~2,000,000 x
Bixby, B., Solving real-world linear programs: a decade and more of progress. Operations Research, 50 (2002), pp. 3-15
OutlineOutline
Introduction
Advanced planning systems
Case 1: Global network design
Case 3: Production planning and scheduling
Case 2: Supply network planning
Case 4: Value chain management
Outlook: Teaching SNP & APS
Case 3: Case 3: Production planning / Production planning / detailed schedulingdetailed scheduling
Strategic Network Design
DemandPlanning
External Procurement
Production Planning / Detailed
Scheduling
Transportation Planning /
Vehicle Scheduling
Supply Network Planning
long-term
mid-term
short-term Order FulfilmentATP / CTP
Case 3: Characteristics
dependent planning in the consumer goods industry
Günther H.O., Grunow, M., Neuhaus, U.:Realizing block planning concepts in make-and-pack production using MILP modelling and SAP APO. IJPR (2006), 3711-3726
Lütke Entrup, M., Günther, H.O., van Beek, P., Grunow, M., Seiler, T., MILP approaches to shelf life integrated planning and scheduling in yogurt production. IJPR (2005), 5071-5100
Production planning / detailed schedulingProduction planning / detailed scheduling
Lot-sizing, sequencing, and procurement proposals
Decisions Generation of production orders
Allocation of resources according to a finite scheduling policy
Mathematical methods Genetic algorithms
Consideration of the availability of resources as hard constraints
Production Planning / Detailed
Scheduling
Application specific algorithms
Consideration of due dates, time windows etc. as soft constraints
Constraint programming
Rules and heuristics
Development of a mathematical
modelling approach based on human expertise.
Case 3: InsightsCase 3: Insights
Efficient support by characteristic based planning tool of APS
module.
High performance of MILP model enables
reactive planning mode.
Block planning conceptBlock planning concept
Given a natural sequence of set-ups (e.g. light to dark)
Decisions on cyclic production patterns (i.e. lot sizes)
Minimizing inventory and set-up costs
Block 1 Block 4Block 3Block 2
Production cycle
1
4 weeks
Production cycle
2 3 4
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
Setup families
time
Production order Major setup
Production of hair dyes
Numerical experimentsNumerical experiments
time4 weeks 8 weeks 12 weeks
...
Period 1 Period 3Period 2 ...
SolutionSolution
MILP model: ILOG OPL Studio and CPLEX 7.0MILP model: ILOG OPL Studio and CPLEX 7.0
Characteristic dependent planning of SAP APO 3.1Characteristic dependent planning of SAP APO 3.1
Setup families
1
4 weeks
Production cycle
2 3 4
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
Production order Major setup
Production of hair dyes
Numerical experimentsNumerical experiments
Master data
No. of master data records
With CDP Without CDP
Locations 1 1
PPMs 30 56
Resources 14 14
Products 37 140
Total 84 213
No
. of
Products 26
Periods 6
Production orders 156
Variables 780
Constraints 786
CPU time 0.25
APO implementation MILP block planning model
Recuced data volume Small CPU times
• Combination of characteristics dependent planning and MILP modelling seems appealing
Numerical experimentsNumerical experiments
OutlineOutline
Introduction
Advanced planning systems
Case 1: Global network design
Case 3: Production planning and scheduling
Case 2: Supply network planning
Case 4: Value chain management
Outlook: Teaching SNP & APS
Case 4: Value chain management
Case 4: Value chain managementCase 4: Value chain management
Strategic Network Design
DemandPlanning
External Procurement
Production Planning / Detailed
Scheduling
Transportation Planning /
Vehicle Scheduling
Supply Network Planning
long-term
mid-term
short-term Order FulfilmentATP / CTP
SCM is usually understood as a “coordinated system … involved in
moving a product or service from supplier to customer.“
Dynamic spot pricing for sales and procurement makes a business unit
highly profitable.
Coordinated decision making along the
entire supply chain appears to be very
effective.
Case 4: InsightsCase 4: Insights
Change of paradigm from “supply chain
management” to “value chain management”?
OutlineOutline
Industrial case study
Sales model
Distribution model
Numerical experiments
Production model
Procurement model
Supply chainSupply chain
Europe
China
Planning problemPlanning problem
• Standard products with defined quality
• Price is the key buying criterion
• Considerable sales price and demand volatility
• Increasing raw material costs due to higher crude oil prices
Chemical commodities
Planning problem
• Operational planning for 6-12 months
• Coordination of sales, production, procurement and distribution activities for the entire company-internal value chain
Value chain networkValue chain network
Procurement locations
Production location
Production location
Sales locations
suppliers’value chain
Sales
company-internal value chain network
Procure-ment
Prod-uction
Distribu-tion
Sales
customers’ value chain
Procure-ment
financial flowmaterial flow information exchange
Sales locations
Procurement locations
Distribution locations
Transportation lanes
Spot sales opportunity
Decision dependenciesDecision dependenciesScenario: Spot sales opportunitiesScenario: Spot sales opportunities
Pipeline inventory?
Stocks?
Production capacity?
Production mode?
Costs?
Raw material availability?
Costs?Transportation capacity?
Sales locationsSales
locations
Productionlocations
Procurementlocations
Distribution locations
Oversupply
Decision dependenciesDecision dependenciesScenario: Oversupply at production locationScenario: Oversupply at production location
Sales locationsSales
locations
Productionlocations
Procurementlocations
Distribution locations
Push spot sales
Production mode?
Inventory allocation?
Delay contract procurement?
OutlineOutline
Industrial case study
Sales model
Distribution model
Numerical experiments
Production model
Procurement model
Sales available total supply
Demand
€/t €/t €/t
€/t
€/t €/t
push
€/t
€/t-spot
sales price
spot sales quantity
€/t
cut
contractdemand price
contractdemand quantity
spot demand quantity
spot demand price
1 2 3 Periods
contractsales quantity
contract sales price €/t €/t€/t
- - -
1 2 3 Periods
Contract demand needs to be fulfilled, spot demand has bid character and is fulfilled depending on supply and profitability.
• Average spot sales price increases
• Reduced production volume
• Savings in raw material consumption• Reduced spot procurement
• Average spot sales price decreases
• Reduced inventories
• Efficient use of contract procurement quantities
• Change of production mode (recipe)
Contract and spot demandContract and spot demand
Price-quantity function
100
100
300
93.3
400
90
600
83.3
0
100
0 300 600Cum. quantity [t]
Price [€/t]
Price-quantity-share function
17%
120%
50%
112%
67%
108%
100%
100%
0%
100%
0% 50% 100%
Cum. quantity share [%]
[%]
Priceshare
Algorithm steps Example
Customer [Unit] A B C D
Quantity [t] 100 200 100 200
1. List individual customer price and quantity
forecasts for product, sales location and
period
→
Price [€/t] 100 90 80 70
2. Sort forecasts by price and determine rank → Rank 1. 2. 3. 4.
3. Determine cumulated quantity for rank → Quantity [t] 100 300 400 600
4. Determine average price for rank → Ø Price [€/t] 100 93.3 90 83.3
5. Determine cumulated quantity share of rank → Quantity [%] 17 50 67 100
6. Determine average price share of rank → Price [%] 120 112 108 100
7. Conduct linear regression for shares → Regression y = -0.2407 x + 1.2408
8. Determine elasticity from regression → Elasticity 2407.0Ssplt
Algorithm steps Example
Customer [Unit] A B C D
Quantity [t] 100 200 100 200
1. List individual customer price and quantity
forecasts for product, sales location and
period
→
Price [€/t] 100 90 80 70
2. Sort forecasts by price and determine rank → Rank 1. 2. 3. 4.
3. Determine cumulated quantity for rank → Quantity [t] 100 300 400 600
4. Determine average price for rank → Ø Price [€/t] 100 93.3 90 83.3
5. Determine cumulated quantity share of rank → Quantity [%] 17 50 67 100
6. Determine average price share of rank → Price [%] 120 112 108 100
7. Conduct linear regression for shares → Regression y = -0.2407 x + 1.2408
8. Determine elasticity from regression → Elasticity 2407.0Ssplt
Linear price-quantity functionLinear price-quantity functionfor each product, sales location and periodfor each product, sales location and period
Quadratic turnover functionQuadratic turnover function
sp
ox
( )p x
Price-quantity function
Average spot price
Cumulated spot sales quantity
x
minx maxx
Turnover function
x
ox
( )p x x
Cumulated spot sales quantity
Spot sales turnover
minxmaxx
Cumulated spot sales quantity
Price-quantity function
ox
( )p x
Average spot sales price
sp
Spot sales quantity
Spot sales
price
x
minx maxx
Piecewise linear approximationPiecewise linear approximation
Turnover approximation
actual turnover curve
Cumulated spot sales quantity
x
( )p x x
Spot sales turnover
Spot sales quantity
minx maxx
ox
Sales locationsDecisions for p,l and t
spot sales quantitySspltx
total sales quantitySpltx
Ssplty spot sales turnover
Input for p,l and t
contract demand quantityScpltq
Distribution locations,SsSspltpltX X min. and max. spot sales
Distribution locations
Constraints of the sales model (1)Constraints of the sales model (1)
• Total sales quantity equals the sum of spot and contract sales quantity.
• Boundaries for spot sales quantities
Scplt
Ssplt
Splt qxx { , } Sp l PL Tt;
Ssplt
Ssplt
Ssplt XxX { , } Sp l PL Tt;
Product-sales location combination
, Sp l PL
DSpltx
• Linear approximation of spot sales quantities and spot sales turnover
Constraints of the sales model (2)Constraints of the sales model (2)
1
1
NSs Ssplt jplt
j
y y
{ , } Sp l PL Tt;
Ss Ss Ssjplt jplt jplty x { , } Sp l PL 1.. 1j N Tt; ;
1
1
NSs Ssplt jplt
j
x x
{ , } Sp l PL Tt;
1,Ss Ss Ssjplt iplt i pltx q q { , } Sp l PL Tt; ;1ijNi ..j 1 N 1 1i; ; ;
OutlineOutline
Industrial case study
Sales model
Distribution model
Numerical experiments
Production model
Procurement model
Decisions for p and t
transportation quantity received via e
Dpetx
Input for p,l and e
,D Dpe peLT UT min. and max. transportation
quantity for p on e
Constraints of the distribution model (1)Constraints of the distribution model (1)
• Transportation quantities are limited for each lane.
• Transportation quantities are limited for each lane and product.
Spltx sales quantity
Product-distribution location combination
, Dp l PL
Transportation lane: e EProduct-transportation
lane combination
, Dp e PE
transportation quantity sent via e with lead time zero
Dpetx
transportation quantity sent via e with lead time one
Dpetx
inventory level at location lDplty
min. and max. transportation quantity on e,D D
e eLT UT
,D Dplt pltLI UI min. and max. inventory
level for p at l in t
Sales locations
, D
D D De pet e
p e PE
LT x UT
,e E t T
D D Dpe pet peLT x UT , ,Dp e PE t T
Constraints of the distribution model (2)Constraints of the distribution model (2)
Decisions for p and t
transportation quantity received via e
Dpetx
Input for p,l and e
,D Dpe peLT UT min. and max. transportation
quantity for p on e
Spltx sales quantity
Product-distribution location combination
, Dp l PL
Transportation lane: e EProduct-transportation
lane combination
, Dp e PE
transportation quantity sent via e with lead time zero
Dpetx
transportation quantity sent via e with lead time one
Dpetx
inventory level at location lDplty
min. and max. transportation quantity on e,D D
e eLT UT
,D Dplt pltLI UI min. and max. inventory level
for p at l in t
Sales locations
• Limited inventory for each product and location
• Balancing incoming and outgoing material flows
D D Dpl plt plLI y UI , ,Dp l PL t T
, 1 , 1
0 1D D Dpl pl pl
D D D D Dplt pl t pet pet pe t
e R e S e S
y y x x x
, ,Dp l PL t T
• Sales and distribution balance
Dpl
D Spet plt
e S
x x
, ,Sp l PL t T
OutlineOutline
Industrial case study
Sales model
Distribution model
Numerical experiments
Production model
Procurement model
Production siteProduction site
P4
process group 1 process group 2
Line 1
Period 1 Period 2
Line 2
carry-over process group
change-over processsequence-dependent
processes withinprocess group
P1
P5
P2 P3
P4 P6P7
P1
Idle
Operational conditions
• Equipment continuously operated except for planned maintenance
• Minimum utilization rate
• Raw material consumption depends on processing mode (throughput speed)
• Processes may have multiple outputs with flexible output rates
• At most one change-over between process groups per period
Production locations Decisions for r,l and t
quantity of process sPrlstx
operating time of group gPrlgtY
Prlsty operating time of process s
Input for r,l
operating hours in tPrltH
Distribution locations,P Prls rlsLQ UQ min. and max. output rate
Constraints of the production model (1)Constraints of the production model (1)related to resourcesrelated to resources
• Allocation of operating time among processes in a group
• Output variation of process groups
Pg
P Prlgt rlst
s S
Y y
, , , ,P Plr l RL r g RG t T
• Continuous operation of equipment
' '
','
Pl
P P P Prlgt rlg g rlg gt rlt
g Gr g RGg g
Y C H
, ,Pr l RL t T
P P P P Prls rlst rlst rls rlstLQ y x UQ y , , , ,P P
lr l RL r g RG t T
, Pr l RL
Resources , Plr g RG
Production locations Decisions for r,l,g and t
quantity of process groupPrlgtX
change-over from g to g’'Prlg gt
Prlgt setup state at the end of t
Input for r,l,g
Change-over time from g to g’'Prlg gC
Constraints of the production model (2)Constraints of the production model (2)related to change-oversrelated to change-overs
• At most one change-over between groups per period
• Start of campaign depends on setup state
• Setup state at the end of a period
, Pr l RL
Change-over , Plr g RG
'
, , ''
1P Pl l
rlg gt
r g RG r g RGg g
, ,Pr l RL t T
,
1Pl
Prlgt
r g RG
, ,Pr l RL t T
'
, ''
( )Pl
P P Prlgt rlgt rlg gt
r g RGg g
X B
, , , ,P Plr l RL r g RG t T
Raw material consumptionRaw material consumption
Raw material recipe function
Production utilization
Inputquantity
Raw material
Outputproduct
Outputquantity
Resource
RecipefactorMaximum capacity
Minimum utilization Linear recipe
Variable prod. costs
0
0.5
1.0
1.5
0% 20% 40% 60% 80% 100%Utilization
Production locations
Decisions for p,l and t
input quantityPinpltx
output quantityPoutpltx
Input for p,r,l,s
Constraints of the production model (3)Constraints of the production model (3)related to productsrelated to products
• Sales and distribution balance
• Recipe for consumption of input material
• Aggregation of output from different processes
, Pr l RL
Products , ,Pin Poutp l PL PL
raw material recipe function,P Pprls prlsa b
Distribution locationsPprls output yield in process s
Distribution locations
Dpetx
, Pl
Pout P Pplt prls rlst
r s RS
X x
, ,outp l PL t T
, ,P P Pl g l
Pin P P P Pplt prls rlst prls rlgt
r g RG s S r g RG
X a x b Y
, ,Pinp l PL t T
Ppl
D Poutpet plt
e S
x x
, ,Pp l PL t T
OutlineOutline
Industrial case study
Sales model
Distribution model
Numerical experiments
Production model
Procurement model
Spot and contract procurementSpot and contract procurement
spot offer quantity
Contract offer quantity
800 €
600 €
Contract offer cost rate
spot offer cost rate spot procure-
ment quantity
Totalprocurement
quantity
800 €
600 €
Offer Procurement
750 €
Contract
Spot
Average procure-ment costs
Procurement locations
Decisions for p,l and t
spot procure-ment quantity
Qspotpltx
Input for p,l and t
Constraints of the procurement modelConstraints of the procurement model
• Balance of procurement quantities and production input
, Qp l PL
contract quantityQconpltq
Distribution locations,Q Qplt pltLS US min. and max. spot quantity
Pinpltx
• Boundaries for spot procurement quantities
Q Qspot Qspotplt plt pltLS x US { , } ,Qp l PL t T
, Pin
Qspot QconPinplt plt plt
p l PL
x x q
, ,Qp l PL t T
Production locations
Objective functionObjective function
Procurement locations
Production location
Production location
Sales locationsDistribution locations
maxt T
S
Ss Ssplt plt
pl PL
x p
, Q
Qspot Qspotplt plt
p l PL
c x
, P Pl g
P Prst rs
r g RG s S
x c
, D
D DIplt pl
p l PL
y c
, D
D DTpet pe
p e PE
x c
OutlineOutline
Industrial case study
Sales model
Distribution model
Numerical experiments
Production model
Procurement model
Basis elements Scale
PeriodsProducts
129
Locations Product-location combinationsTransportation lanesTransportation lane – product
comb.
24713467
Production resourcesProduction processes
1415
ConstraintsVariablesCPU time (MIP gap 0.5%)
14,35317,2113.0 sec
Model implementation using ILOG OPL Studio 3.71 and CPLEX 9.1
Test case and implementation Test case and implementation
Test case: excerpt from a real industrial value chain
Profit
Distribution costs
Production costs
Procurement costs
Turnover
High volatility of profit, sales and procurement costs, but stable production and distribution costs
Optimization run with company dataOptimization run with company data
No full capacity utilization and spot demand coverage
OutlineOutline
Introduction
Advanced planning systems
Case 1: Global network design
Case 3: Production planning and scheduling
Case 2: Supply network planning
Case 4: Value chain management
Outlook: Teaching SNP & APS
Teaching SNP & APSTeaching SNP & APS
Development of teaching software for SCMusing Advanced Planning Systems
Joint project by:B. Fleischmann, Augsburg, Germany,M. Grunow, Copenhagen, Denmark,H.-O. Günther, Berlin, Germany,H. Meyr, Darmstadt, Germany,H. Stadtler, Hamburg, Germany
Supported by:SAP AG, Walldorf, Germany,DATANGO AG, Berlin, Germany
Output:DVD with teaching software (Phase 1)Accompanying text book (Phase 2)
19 products made to stock 3 production sites with two lines each 3 warehouses 60 customers
Ludwigshafen
Augsburg
Magdeburg
Case study: Soft drink manufacturerCase study: Soft drink manufacturer
short-term
medium-term
long-term
PurchasingWorkbench
Procurement Production Distribution Sales
Strategic Network Planning
Modules of mySAP SCMModules of mySAP SCM
Supply Network Planning (SNP)
DeploymentDeploymentProductionPlanning
&Detailed
Scheduling(PP/DS)
ProductionPlanning
&Detailed
Scheduling(PP/DS)
DemandPlanning
(DP)
DemandPlanning
(DP)
TransportationPlanning & VSTransportationPlanning & VS
ATP & CTPATP & CTP
for
Thank you
attention! your