Transcript BA 851: 4 Main Components

DS 851: 4 Main Components
1. Applications
• The more you see, the better
2. Probability & Statistics
• Computer does most of the work
3. A Modeling Environment
• Arena
4. Simulation Methodology
• Issues faced by any simulation modeler
Robert M. Saltzman © 2005
1
1. Applications
• Imitation is a good way to learn
• Focus on business & public sector problems
EXAMPLES:
• Call Centers: How many agents are needed?
• Traffic & Parking: How does behavior affect availability?
• Airports: Can delays & missed flights be minimized?
• Ambulance Deployment: What’s the best dispatch rule?
• Financial Planning: What’s the effect of uncertainty?
Article list: we’ll hear about many applications
Robert M. Saltzman © 2005
2
2. Probability & Statistics: Why?
INPUT
• Most systems contain
uncertainty (randomness)
• Model inputs include
probability distributions
• EX: Service time – how to
represent in the model?
OUTPUT
• Simulation models
generate variable output
• Each model run just gives
1 approximate answer
• EX: Customer flow time –
what’s a 95% CI for m?
Robert M. Saltzman © 2005
3
3. Modeling Environment: Arena
• Professional-quality package at low cost
– Lots of preprogrammed features
• Graphical interface; minimal coding
• Easy to animate
• Integrated environment:
– Input Analyzer; Debugger; Output Analysis
• Supported by one of the market leaders
Robert M. Saltzman © 2005
4
3. Arena - Drawbacks
• Professional => Non-trivial package
– There are a lots of things to learn
• Some academic version limitations
– Can only have 150 entities in system at once
– Can only have 100 modules in model
• Professional version is quite expensive
Robert M. Saltzman © 2005
5
4. Methodology Issues
• Scattered throughout the course
EXAMPLES:
• How big of a model should I build?
• How much input data should I collect?
• How do I analyze the output?
• How do I validate my model?
• What factors affect implementation?
Robert M. Saltzman © 2005
6
Main Types of Simulations
I. PHYSICAL
• SF Bay & Delta Model (Sausalito)
• Shoe tester (Exploratorium, SF)
• Wind tunnels (NASA Ames, Mtn. View)
• Weightlessness training (NASA, Houston)
• Car crash dummies (Mercedes Benz)
Robert M. Saltzman © 2005
7
Types of Simulations - continued
II. COMPUTER-BASED: Substitute a computer model for
real system, and observe its behavior.
• Deterministic: No uncertainty.
– Describe key relationships with equations
– Try a set of parameter values & calculate
– Recalculate for various sets of parameter values
• Probabilistic: Randomness is involved.
– Static (“Monte Carlo”): Repeated trials with sampling
– Dynamic: System evolves over time
Robert M. Saltzman © 2005
8
Why use models, in general?
• Convenience
– e.g., use a map to locate places in the world
• Experimentation & Sensitivity Analysis
• Test alternative policies to make a decision:
– Probably cheaper to use model
– Maybe faster, safer
• Organize your knowledge about the system
• Real system may not yet exist
Robert M. Saltzman © 2005
9
Why use simulation models?
• To estimate … average system behavior;
variation; effect of adding resources …
• To compare … alternative system designs;
various operating policies, scheduling rules …
• To visualize … almost any process, even very
complex ones.
• To educate … get others involved, enthused.
Robert M. Saltzman © 2005
10
What can’t simulation models do?
• Can’t give meaningful results if input data are
inaccurate (GIGO)
• Can’t reflect structures that are not modeled
• Can’t give exact solutions
• Can’t directly tell you the optimal strategy,
though it may be inferred by experimenting
• How does simulation compare to LP?
Robert M. Saltzman © 2005
11
When to use simulation
• For analyzing systems with
–
–
–
–
–
several sources of randomness, or
complicated structure, or
unusual operating rules, or
nonlinear relationships, or
entities that interact with one another
• When other analytical frameworks are too limiting,
e.g., queuing model assumptions
• Simulation models have few limitations,
but they may take longer to develop
Robert M. Saltzman © 2005
12