04_meier_b_snp.pdf

Optimizing the Supply Network in
mySAP Supply Chain Management
Dr. Dirk Meier-Barthold
GBU SCM
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 1
 AG
SAP
20.10.2000 / 1
Agenda
Integrated Supply Network Planning
and Optimization with APO
1
2
3
4
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 2
 AG
SAP
20.10.2000 / 2
Modeling the Supply Network
Optimizing the Supply Network
Selected planning scenario
Supply Network Planning
Decision support for supply network planner:
Decisions to be made:
Global sourcing decisions
Global load-balancing decisions
Global lot-sizing decisions
Supply Chain
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 3
 AG
SAP
20.10.2000 / 3
Planning Horizon
Level of Detail
Integrated Supply Network Planning and Optimization
Network Design
Demand
Planning
Supply Network Planning
Procurement
Planning
Production
Planning
Distribution
Planning
Purchasing
Workbench
Detailed
Scheduling
Vehicle
Scheduling
Available
to Promise
Supply Network Planning
Optimizer
Sourcing
LP
Heuristics
Balancing
MILP
Propagation
Lot-Sizing
CTM
DRP/MRP
Result: Network wide supply decisions
products, locations, periods and quantities
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 4
 AG
SAP
20.10.2000 / 4
Agenda
Integrated Supply Network Planning
and Optimization with APO
1
2
3
4
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 5
 AG
SAP
20.10.2000 / 5
Modeling the Supply Network
Optimizing the Supply Network
Selected planning scenario
Modeling the Supply Network
External
Procurement
T
GR
O
Production
R
I
S
P
R
R
I
Transportation
I
GI
T
GR
R
R
R
O
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 6
 AG
SAP
20.10.2000 / 6
O
Decision Variables
External
Procurement
T GR O
Production
R
I
S
P
R
R
O
I
Production Quantity
Transportation
I GI T GR O
External Procurement
R R R
Additional Capacity
Transportation Quantity
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 7
 AG
SAP
20.10.2000 / 7
Decision Constraints
External
Procurement
T GR O
Production
R
I
S
P
R
R
O
I
Product Constraints
- Consumption (fix, variable)
- Minimal lot size
- Fixed lot size
- Shelf life
Transportation
I GI T GR O
R R R
Customer Constraints
Resource Constraints
(Production, Transport, Handling, Storage)
- Capacity (normal, additional, calendar)
- Consumption (set up, variable)
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 8
 AG
SAP
20.10.2000 / 8
- Back order
- Lost sales
- Safety stock
Cost of Decisions
External
Procurement
T GR O
Cost of Procurement
Production
R
I
S
P
R
R
O
I
(piecewise linear cost function)
- Production quantity
- Transportation quantity
- External Procurement
Transportation
I GI T GR O
Cost of Product Constraints
R R R
- Cost of violating Shelf Life
Cost of Customer Constraints
Cost of Resource Constraints
(Production, Transport, Handling, Storage)
- Cost of additional capacity
- Cost of Inventory consumption
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 9
 AG
SAP
20.10.2000 / 9
(Demand classes)
- Cost of back order
- Cost of lost sales
- Cost of using safety stock
Problem Complexity
3 classes of problem complexity:
-> linear program (LP)
-> all decision variables are proportional
-> mixed integer linear program (MILP)
a) yes/ no decisions
-> set up
-> minimal lot size
-> piecewise linear cost function
b) integer decisions
-> fixed lot size
-> full truck loads
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 10
 AG
SAP
20.10.2000 / 10
Reduction of Problem Complexity (1)
Are mixed integer really necessary?
- Set up
-> not reasonable, if a lot of products are on resource per bucket
- Minimal lot size
-> not reasonable, if minimal lot size is small to average lot size
- Piecewise linear cost function
-> only reasonable, if few pieces are modeled
- Discrete lot size/ rounding
-> not reasonable, if lot size is very high (e.g. 97.5)
-> not necessary, if production over buckets is allowed
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 11
 AG
SAP
20.10.2000 / 11
Reduction of Problem Complexity (2)
How can we create a reasonable SNP model?
- Use aggregated time-buckets
- Focus on Supply Chain relationships
- Use key products and bottleneck resources, only
- Design easy PPM’s
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 12
 AG
SAP
20.10.2000 / 12
Agenda
Integrated Supply Network Planning
and Optimization with APO
1
2
3
4
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 13
 AG
SAP
20.10.2000 / 13
Modeling the Supply Network
Optimizing the Supply Network
Selected planning scenario
Optimizing the Supply Network
Objectives for the SNP-Optimizer:
- good performance of planning result
- good performance of planning runtime
Planning Runtime
Runtime
Planning
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 14
 AG
SAP
20.10.2000 / 14
Planning Result
Result
Planning
Incremental Optimization
Can be necessary due to:
- Problem size
- User experiences
-> Important: Focusing on strongest constraints
- with Selection
-> horizontal aggregation
-> vertical aggregation
-> DRP/ MRP-like planning
- within Selection
-> product decomposition
-> time decomposition
-> priority decomposition
-> activate constraints
-> restrict runtime
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 15
 AG
SAP
20.10.2000 / 15
Horizontal Aggregation
Aggregation of demands by classes
Aggregation of shortage costs
Advantage: Customer information are taken into account
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 16
 AG
SAP
20.10.2000 / 16
Vertical Aggregation
Production
2.
3.
1.
1. Aggregation by product-location hierarchy
-> supply, demand, stock, costs
2. Optimization on aggregated level
3. Disaggregation by deployment algorithm push fair share A
-> product-location hierarchy, ppm hierarchy
-> Vertical aggregation for special production structure, only !
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 17
 AG
SAP
20.10.2000 / 17
DRP/ MRP-like Planning
3.
4.
1.
2.
Step by Step Planning over the Supply Chain
To get feasible solution:
Set secondary and distribution demand as
soft constraint (demand class)
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 18
 AG
SAP
20.10.2000 / 18
Optimization with Decomposition
Decomposition via Product
Product 1
:
Product n
Decomposition via Priority
Demand class 1
Decomposition via Time
:
Demand class n
Time
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 19
 AG
SAP
20.10.2000 / 19
Activate Constraints
External
Procurement
T GR O
Variable Constraints
Production
R
I
S
P
R
R
O
I
Transportation
I GI T GR O
R R R
Resource Constraints
- production capacity
- transportation capacity
- handling capacity
- inventory capacity
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 20
 AG
SAP
20.10.2000 / 20
(end date, bucket oriented)
- set up
-> production
- minimal lot-size
-> production
- piecewise linear cost function
-> production, transport,
external procurement
- fixed lot-size/ rounding
-> production, transport
Agenda
Integrated Supply Network Planning
and Optimization with APO
1
2
3
4
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 21
 AG
SAP
20.10.2000 / 21
Modeling the Supply Network
Optimizing the Supply Network
Selected planning scenario
Combination of Vertical and Horizontal Aggregation
2b.
Production
3.
2a.
1.
1. Selection of bottleneck part of supply chain
2. Optimization with
2a. Horizontal Aggregation
2b. Vertical Aggregation
3. Optimization of non bottleneck part
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 22
 AG
SAP
20.10.2000 / 22
SNP Optimizer - Customer Problems (1)
●
Discrete industry
Model: 13 Buckets, 15012 Locations-Products, 4887 Arc-Materials,
7581 PPMs
■ Solution: optimal after 10 minutes
■
●
Consumer industry
Model: 30 Buckets, 19.000 Locations-Products, 23.000 Arc-Materials,
8.500 PPMs
■ LP: 2.600.000 Variables, 600.000 Constraints
■ Solution: optimal after 30 minutes
■
●
Chemical industry
Model: 3 Buckets, 2131 Locations-Products, 1461 Arc-Materials, 356
PPMs
■ MIP: 20.300 Variables (1.050 discrete, 1.050 binare), 10.500 Constraints
■ Solution: < 1% optimality-gap after 1 minute
■
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 23
 AG
SAP
20.10.2000 / 23
SNP Optimizer - Customer Problems (2)
●
Consumer industry
Model: 22 Buckets, 916 Location-Products, 333 Arc-Materials, 741
PPMs
■ MIP: 104.000 Variables (14.000 discrete), 46.000 Constraints
■ Solution:
◆ < 5% optimality-gap after 5 minutes
◆ < 3% optimality-gap after 80 minutes
■
●
Financial sector
Model: 23 Buckets, 1 Product, 22 Locations, 30 Lanes
■ MIP: 3000 Variables (300 binare), 1600 Constraints
■ Solution: < 1% optimality-gap after 1 minutes
■
 SAP
2001,Optimizing the SN, Dr. Dirk Meier-Barthold page 24
 AG
SAP
20.10.2000 / 24