Welcome to MS383 Simulation

1
Welcome to MS383 (Simulation)
2
Course Introduction

• Welcome to the course!
• Thanks for taking the course with me.
• This is an OR minor course focused on modeling
  systems using simulation techniques.
• We will meet thrice a week in the E slot here in
  HSB 334.
• The course overview MS Word file.

3
Course Objectives

• The course hopes to answer
• How to build models of real systems using
  simulation techniques (as against analytical
  techniques)?
• How to statistically analyze these models?
• How to gain managerial insights about the real
  systems from the simulation models?
• To achieve these objectives, we will spend
  considerable time on statistical tools and
  techniques relevant to simulation.
• However, no pre-requisite knowledge of statistics
  is required for the course.

4
Course Introduction

• Why is this not CS383?
• True, we are going to mimic the real systems by a
  computer code (or software).
• The focus is going to be more on analyses rather
  than the code itself.
• We are more interested in finding the performance
  parameters of the system using simulation. So
  that we can take decisions based on the models.
• Simulation software packages and tailor-made
  codes are going to be just tools to achieve our
  objective which is to draw conclusions about
  the real system.

5
Simulation

• To imitate The operations of various kinds of
  real-life facilities or processes can be
  simulated using a computer code.
• The facility of process of interest is the
  system. System is a collection of entities that
  act and interact together toward the
  accomplishment of some logical end.
• To imitate the real-life processes, we may have
  to make some assumptions.
• The logical or mathematical relationships amongst
  these assumptions is a model.

6
Simulation

• If the inter-relationships are simple, we can
  represent them in the form of exact mathematical
  expressions. These are the analytical models. The
  results we get from these are exact called
  analytical solution.
• When the inter-relationships are complex, we
  analyze them numerically using computers. These
  are simulation models. Here, we gather data in
  order to estimate the desired true
  characteristics of the model.

7
Systems

• State of a system collection of variables that
  describe the system completely (with respect to
  the study) at a particular time instance.
• Discrete systems is the one for which the state
  variables change instantaneously at separated
  points of time.
• e.g. Bank with of customers as state variable.
• Continuous systems is one for which the state
  variables change continuously with respect to
  time.
• e.g. Fluid flow in pipe-lines.

8
Systems
Real System
Experiment with actual system
Experiment with a model
Physical model
Logical model
Analytical solution
Simulation
9
Systems

• Static vs. dynamic models In static models,
  there is no time axis.
• Deterministic vs. stochastic models No
  probabilistic components in the deterministic
  models.
• Continuous time vs. Discrete time models Same
  as earlier definitions of continuous and
  discrete time systems.

10
Pros and Cons of Simulation

• Advantages
• Mathematical models (which can give analytical
  solutions) fail to describe many complex,
  real-life systems accurately.
• Using simulation, performance of the existing
  system can be estimated for a range of operating
  conditions.
• Alternatives to existing system could be tested
  with relative ease.
• Control in simulation experiment is better than
  that in experiment with the system itself.
• Time axis can be played around with relative ease.

11
Pros and Cons of Simulation

• Disadvantages
• Most often, the results of simulation are
  estimates. Several runs of the model could be
  required to study the system in details.
• Often are time consuming and expensive.
• Large volume of numerical data and animation
  (with the use of latest software) create at
  times, false impression of accuracy and hence,
  confidence.

12
Pitfalls of Simulation

• Failure to have clearly defined objective at the
  beginning of the study.
• Inappropriate level of modeling details.
• Failure to communicate with the management on
  regular basis.
• Treating it as if it is a complex exercise in
  computer programming.
• Failure to have people with OR and Statistics
  background in modeling teams.
• Using commercial simulation software without
  knowing its limitations.
• Misuse of animation.

13
Pitfalls of Simulation

• Failure to account for correct source of
  randomness.
• Using arbitrary distributions as input.
• Analyzing output data from one simulation run
  without any consideration of reality.
• Making single replication of the model. And
  comparing alternate systems based on the same.
• Using wrong measures of performance.

14
Steps in Simulation Study

• Formulate the problem and plan the study Clear
  statement of purpose of the study, its overall
  objectives, and specific issues to be addressed
  is a must. The alternatives system designs to be
  studied should be delineated, and criteria for
  evaluating the efficacy of these alternatives
  should be given.
• Collect data and define a model Information and
  data should be collected and used to specify
  operating procedures and probability
  distributions for the random variables used in
  the model. Data on performance of the system
  should be collected for validation.
• Check for validity Though validation should be
  carried out through the study period, it is
  required at scoping stage.

15
Steps in Simulation Study

• Write a computer program and verify Decision
  about either using a general purpose computer
  language (C, C etc) or a simulation package
  (ARENA etc) should be made. General purpose
  language might be preferred by the modeler but to
  include complex features to the model simulation
  package might be better.
• Make a pilot run Pilot runs of the verified
  model are made for validation purpose in Step 6.

16
Steps in Simulation Study

• Check for validity Pilot runs can be used to
  test the sensitivity of the models output to
  small changes in input parameters. This will
  emphasize the need for better estimates of input
  parameters. The data from system (if it exists)
  and model could be compared for closeness.
• Design the experiments If testing more than one
  alternatives, decide about what system design to
  simulate. For each design to be tested, decide
  about the initial conditions for the run(s),
  warm-up periods of each run, the length of the
  simulation run(s), and number of independent
  replications.

17
Steps in Simulation Study

• Make production runs Production runs are made to
  provide performance data on the system design of
  interest.
• Analyze output data Statistical techniques are
  used to analyze the output data from production
  runs. Goal is to construct confidence interval
  for a measure of performance on one particular
  system or to decide which simulated system is the
  best choice.
• Document, present, implement results Compilation
  of the entire study, assumptions, model
  description, results, and conclusions.