I' Introduction

1
I. Introduction

• M. Peter Jurkat
• CS452/Mgt532 Simulation for Managerial Decisions
• The Robert O. Anderson Schools of Management
• University of New Mexico

2
Definitions

• Simulation
• process of experimenting with a model of a
  dynamic systems (e.g., process) to
• study or test the behavior of the system
• improve, problem solve
• design and/or select new systems , and/or
• train operators on a model of an existing systems
• System purposeful, interrelated components with
  interdependencies and complexity
• Behavior purposeful, interrelated sequences of
  activities
• Dynamic time varying (static systems are dull!)

3
Examples

• Service Systems
• Traffic on Networks messages to/from computers,
  cars on roads/rails, airplanes to/from
  airports/gates, ships to/from harbors/piers,
  elevators
• Retail/Service stores selling goods,
  service/repair shops, logistics/inventory/distribu
  tion/MRP
• Manufacturing Systems
• Materials, Chemicals, Biologicals
• Appliances, Automobiles/Trucks, Toys, Clothing
• Electronics, Weapons Systems
• Computations using models from other disciplines
• Macroeconomic taxation/interest rate
  cost/benefits
• Pollution environmental intervention
  cost/benefits
• Project Management completion time vs resources

4
Why Simulate?

• To overcome human limitations in
• Physical capability avoid injury and death be
  able to control systems whose dynamics are not
  yet known,
• Mental capability attention, memory, processing,
  
• Analysis allows us to study systems too complex
  for analytic description and/or too dangerous for
  human safety gain knowledge
• Design attempt changes in IVs to drive one or
  more DVs toward an optimal value or combination
  of values for design, improvement, and/or problem
  solving

5
When not to Simulate!

• When theory can determine sufficient results
• When it will cost more to simulate than the
  return on the knowledge gained
• When there is incomplete information about the
  system (can handle imprecise but not missing
  pieces)
• Need at least inputs and related outputs for
  black boxes
• Can assume missing information and check against
  known results if agreement, support for
  assumptions
• When it is not possible to develop a
  representative, tractable simplification of the
  system

6
Definitions (cont.)

• Model representation of a system three phases
• Verbal always included in any representation
• Graphical see pages 22, 39, 50, 54, 367, and
  536
• Algorithm and/or computer program
• Experimentation purposeful, structured, and
  controlled change of the inputs factors
  (independent variables IVs, exogenous, ) of a
  product and/or process to observe resulting
  changes in outputs (dependent variables - DVs,
  responses, results, outcomes, )
• Both IVs, DVs also called measures or metrics
• In simulation literature a run is one execution
  of the simulation program at one combination of
  input variable values also called a replication

7
Graphical RepresentationLogical Symbols

• BCNN 4th Ed., Figure 2.1, page 22 Single Server
  Queuing System

8
Graphical RepresentationState Variable Tracking

• BCNN 4th Ed., Example 2.2, Figure 2.11, page 39

9
Graphical RepresentationPhysical Layout

• BCNN 4th Ed., Example 2.6, Figure 2.15, page 50

10
Graphical RepresentationNetwork Model

• BCNN 4th Ed., Example 2.8, Figure 2.18, page 54

11
Graphical RepresentationBlack Box

• BCNN 4th Ed., Figure 10.5, page 367

12
Graphical RepresentationComponent Relationship

• BCNN 4th Ed., Example 14.4, Figure 14.10,
  page536 Website configuration

13
Simulation Study Representation(after Banks et
al, Figure 1.3, Page 15)
Set Objectives and Project Plan
Problem Formulation
(Re)Conceptualize Model Collect Data
Yes
No
Translate Model
Can Model be Verified?
No
Can Model be Validated?
Yes
DOE - Design Experiments
Runs and Replications
Analysis
No
Results Clear and Able to be Described?
Document, Report and Recommend
Yes
14
Simulation Study

• Identify problem(s), improvement(s), and/or plan
  new capabilities
• Specify the system select boundaries, identify
  inputs, entities, attributes, events, activities,
  processes, and state variables - specify
  output(s) and their desired values
• Build a conceptual and operational model of the
  system build a representation of inputs,
  entities,

15
Simulation Study (cont.)

• Verify and Validate (as best you can) the
  operational model against existing system only
  partial model verification/validation may be
  possible for new systems
• Perform screening experiment(s) to identify IVs
  with significant effect on desired output(s)
  proceed with only these IVs
• Select ranges of IVs which reduce variability to
  acceptable levels, if necessary (Critical
  Step!!!)
• Experiment with model to identify values of
  inputs which optimize output or achieve goal
• Build system or prototype to test results of
  study

16
System Description, Problem, Objectives, Project
Plan

• Verbal description/linguistic analysis
• Identify problems and/or (re)design objectives
• Identifying relevant
• Entities
• Attributes
• Events
• Activities/processes, and
• state variables
• to address problem(s) and/or objectives
• Develop project plan may follow STEPS FOR
  EXPERIMENTAL DESIGN in Schmidt and Launsby on
  pages I-26 and I-27

17
Simulation Model Components

• Entities named physical/conceptual objects
  (improper nouns used for UML classes, proper
  nouns for UML objects)
• Attributes named characteristic or property
  (adjectives)
• Methods named activities or operations the
  entity can perform (predicates verb
  direct/indirect object(s))
• States named set of conditions, standings,
  circumstances, and positions describing an entity
  at a particular time (adjectives, verbal nouns
  gerunds)
• Processes named groups of activities
• Events named noteworthy occurrences, often at
  the beginning or completion of one or more
  activities and/or processes

18
Identify Variables

• Output (dependent) variables whose values will be
  the problem solution/design improvement
• Operational definitions
• Range of values
• Appropriate output analysis
• Transient vs. steady state
• Statistical tools (confidence intervals, t-tests,
  ANOVA, regression/model building)

19
Identify Variables (cont.)

• Factors among whose combination of values will
  provide the problem solution of optimum design
• These will be varied by the investigator
  according to some experimental design (DOE)
• Operational definitions, range of values, level
  values, potential interactions (for eventual
  assignment to DOE columns)
• Factor model relates factors to output variables
  developed in modeling experiments

20
Identify Variables (cont.)

• State variables whose change of values determine
  the events
• Other variables necessary for a complete model
• Identify stochastic variables and collect data to
  specify their distributions
• If close to known mathematical distributions then
  identify their parameters
• Else use as empirical distributions
• Collect data for constants these may have to be
  fitted from the data

21
(Re)Conceptualize Simulation Model and Collect
Data

• Simulation model relates all variables to output
  variables
• Representation tools
• natural or domain specific language/jargon
• mathematical notation
• code (e.g., Java, GPSS) and pseudo-code
  (primitive action, choice, iteration)
• flow charts
• UML
• PERT/CPM diagrams
• pictorial images
• storyboards/movies
• Build Simulation Model and the Simulation itself

22
Verify and Validate

• Verify that calculations in implementation are
  correct
• Validate the results against output known to be
  an accurate reflection of reality
• May only be possible for parts of the model or
  highly restricted situations
• If not make reasonableness checks

23
Design and Conduct Experimental Runs

• Do experiments
• Screen experimental runs (2-level?) to find the
  significant few factors
• Model
• further or new set of experimental runs (3 or 5
  levels) to develop factor model equations
• fit equations by regression
• Optimize
• solve equations for optimum or
• make experimental runs to drill down to best
  combinations of factors
• Check local optimum (simulate all neighbors)

24
Solutions/Design Identification and Report

• From simulation runs identify the solution to the
  problem and/or the optimum design
• Write Report
• Abstract (may only be needed for research or
  archive reports)
• Executive Summary non-technical problem
  statement, solution/design, justification (not
  usually in research reports)
• Technical Report complete details so that entire
  project could be repeated by others including
  equations, code, distributions, run results
• Technical Appendix

25
Simulation ReportSee SimulationStudyReportOutlin
e.doc for details of each section

• Abstract
• Executive Summary
• Full Technical Report
• Situation, Problems, Opportunities, Goals, and
  Objectives
• Background
• System Specification
• Performance Measures
• Input Factors
• System Representation/Model
• Project Activities
• Input Specification and Model Implementation
• Verification and Validation
• Experiments and Results of the Simulation Runs
• Analysis and Results
• Conclusion and Recommendations
• Technical Appendix

26
Assignments

• Choose one application from Banks 1.1 or your
  selection for a DESS project write sections
  3.a)-c) of the report (specify the entities
  make a symbolic representation using flow charts,
  UML, or ). This can be a group exercise.
• Individual exercises, Banks 1.6
• Prepare a brief written report (include copy of
  papers if possible) and
• Prepare an even briefer set of slides for
  presentation to the class (unless the subject of
  your paper is particularly interesting you may
  not be asked to actually make the presentation
  in any case the presentation will be informal)

27
Model Classification

• Does system evolve over time?
• Static one time period or steady state
• Dynamic changes occur over time period of
  interest
• How often do we have to specify changes?
• Discrete Event changes only occur at instances
  separated in time
• Continuous Event changes occur constantly
• How predictable is the system?
• Deterministic we assume we can model the system
  as if we know all that needs to be known about
  the system
• Stochastic (Stochs) we know certain aspects of
  the system only as a probability distribution
• Totally Unpredictable cannot model

28
How Various Models are Studied