Page 634 - ExtendSim User Guide
P. 634

Model Execution
Other Units
☞ Time unit conversions are only applicable if the model uses specific global time units and does not use Calendar dates.
Other Units
Flow units
Flow units (gallons of liquid, cartons of cereal, rolls of paper, tons of flour, and so forth) can be defined in the Rate library blocks. For more information, see “Units and unit groups” on
page 329.
The Transport and Convey Item blocks (Item library) utilize a unit of distance (feet or meters) to calculate the delay required to move an item from one point to another. Similarly, the Con- vey Flow block (Rate library) uses a user-defined unit of length to calculate a delay for moving units of flow.
Length and number of runs
An important consideration when building a model is to determine how long and how often the simulation should be run. Your entries for the End time and Runs in the Simulation Setup dia- log will depend on four factors:
• Whether the system being modeled is terminating (has a natural end point) or non-terminat- ing (has no obvious end time)
• Theperiodofinterest(whatportionoftimeyouaremodeling)
• Your modeling objectives (estimating performance, exploring alternatives, or other)
• How the samples for statistical analysis are obtained (from running multiple short simula- tions or by analyzing portions of one very long simulation run)
A brief discussion of terminating and non-terminating systems follows. A comprehensive dis- cussion on this matter is beyond the scope of this manual.
Terminating systems
Some systems obviously lend themselves to a natural determination of end time. For instance, most service operations have a point at which activities end. In these terminating systems there is an obvious time when no more useful information will be obtained by running the simulation longer. For example, when modeling one day at a walk-in customer service center that is only open 8 hours each day, you could safely set the simulation end time for 8 hours. Since custom- ers would not wait overnight in line for service, the service queue would be empty or cleaned out at the end of the day and further simulation time would be unproductive.
Because terminating systems do not typically reach a continuing steady state, your purpose in modeling them is usually to look for changes and identify trends, rather than obtain system- wide statistical averages. For instance, in the customer service center mentioned above, it is more important to determine the peaks and valleys of activity than to calculate overall aver- ages. Basing your decisions on average utilization in this case could obscure transient prob- lems caused by multiple periods of understaffing.
Since the initial conditions in terminating systems will have an impact on results, it is import- ant that they be both realistic and representative of the actual system. Terminating systems are simulated using multiple runs for short periods of time using different random seeds for each
How To

   632   633   634   635   636