Process Improvement via Simulation Optimization

1
Process Improvement via Simulation Optimization

• Manuel Laguna
• University of Colorado

2
References

• Laguna, M. and J. Marklund (2005) Business
  Process Modeling, Simulation and Design, Pearson
  Prentice Hall, ISBN 0-13-091519-X
• Laguna, M. and R. Martí (2002) Neural Network
  Prediction in a System for Optimizing
  Simulations, IIE Transactions, vol. 34, no. 3,
  pp. 273-282
• April, J., F. Glover, J. Kelly and M. Laguna
  (2003) Practical Introduction to Simulation
  Optimization, Proceedings of the 2003 Winter
  Simulation Conference, S. Chick, P. J. Sánchez,
  D. Ferrin, and D. J. Morrice (Eds.), New Orleans,
  pp. 71-78
• OptQuest Engine Documentation (http//www.opttek.c
  om)

3
Process View
Policies Resources
Inputs
Outputs
Uncertainty
4
Classification

• Production and Manufacturing
• Business

5
Examples of Generic Business Processes
6
Process Orientation

• Focus on business processes is largely due to
  Reengineering the Corporation by Michael Hammer
  and Jim Champy (1993)
• The main ideas for designing (or redesigning)
  efficient business processes come from industrial
  engineering

7
Application of Technology

• What can we do now (with this new technology)
  that we couldnt do before?
• Automation is not innovation!

8
Eliminate Waste
9
Radical vs. Incremental Improvement
10
Improving Performance
Policies Resources
Inputs
Outputs
Uncertainty
11
Inventory Management and Order Fulfillment Example
PC Supplier
Receive order from MassPC
Fulfill order to MassPC
MassPC Warehouse
Store
Release inventory
Distributors
Order fulfilled
Distributor order
Adapted from Managing Business Process Flows by
Anupindi, et al., Prentice Hall, 2006
12
Order Fulfillment Process View
Resources Production capacity and warehouse
space Policies Reorder point, order quantity,
order priority, mode of transportation
Order placed
Order fulfilled

• Sources of Uncertainty
• Demand
• Production time
• Transportation time

13
Simulation Optimization View
Process Simulation
Decisions Production capacity Warehouse
space Reorder point Order quantity Order
priority Mode of transportation
Performance Holding cost Ordering cost Service
level

• Sources of Uncertainty
• Demand
• Production time
• Transportation time

14
Black-box Simulation Optimization
Optimizer
Decisions
Performance
Simulation
15
Metaheuristic Optimization
Maximize Performance
16
Evolutionary Methods

• Generate a set (population) of solutions
• Combine subsets of solutions and/or modify single
  solutions
• Decide which solutions will be part of the next
  population
• Repeat

17
Generating an Initial Set of Solutions

• Totally random
• Deterministic procedures
• Greedy randomized constructions

18
Combining Solutions

• How many solutions will be combined?
• Which solutions will be combined?
• How are the selected solutions combined?
• How many solutions are generated from each
  combination?

19
Combination Methods

• Binary strings
• Permutation vectors
• Integer variables
• Continuous variables

20
Evolving the Population

• How are solutions selected?
• How many solutions will survive?
• How do we know that we have a good population?
• Do these decisions depend on the solution
  representation?

21
Discarding Solutions Before Simulating
Optimizer
Decisions
Performance
Good?
Simulation
Yes
Predicted performance
No
Metamodel
Discard
22
OptQuest

• General-purpose optimizer originally designed for
  simulation-optimization
• First version completed in mid 1990s
• The main engine is based on scatter search (an
  evolutionary method)
• More information can be found at www.optquest.com

23
Commercial Products
24
Conclusions

• Metaheuristic optimization technology has enable
  black-box optimization for simulated systems
• There are still some limitations (e.g., dealing
  with large number of variables, expensive
  simulations, constraints and statistical
  significance) and additional research and
  development efforts are still undergoing