Basic HTML Version
Page 635 - ExtendSim User Guide
P. 635

``````Model Execution 609
Length and number of runs
run. As discussed in “Confidence intervals” on page 645, the more frequently a simulation is run, the more confidence you can have in the results.
Non-terminating systems
A non-terminating system does not have a natural or obvious end time. Models of non-termi- nating systems are often called steady-state systems since, if they are run long enough, the results tend to a steady state. In these situations, simulation runs could go on indefinitely with- out materially affecting the outcome. Most manufacturing flow systems and some service situ- ations (for example, 24-hour convenience stores, emergency rooms, and telephone service centers) are non-terminating systems.
Systems that have off-shift periods, for instance a manufacturing plant that operates only two shifts a day, may still be considered non-terminating. If the operation does not clear out at the end of the second shift, but instead the first shift starts up where the second shift ended, the system is considered non-terminating and the off shift period is just ignored for modeling pur- poses.
The important considerations when modeling non-terminating systems involve eliminating the initial bias caused by the warm-up period, deciding how to obtain samples for statistical analy- sis, and determining the length of the run.
• Thewarm-upperiodistheperiodfromstart-uptowhenprocessesoperateattheirnormalor steady-state level. In simulation models, start-up conditions may be unrealistic or nonrepre- sentative of the actual system and may bias simulation results. To overcome this, you can either wait until after the warm-up period to gather statistics, reset statistics as discussed in “Clear Statistics” on page 644, or run the simulation for a long period of time to “swamp” the biasing effect of the initial conditions.
• To obtain multiple samples for statistical analysis, you could perform repeated runs after eliminating or compensating for the warm-up bias or you could do one extremely long run and calculate statistics on results occurring during various windows of time. As with termi- nating systems, the greater the number of samples, the higher the confidence in the results.
• The run length of a non-terminating system depends on various factors, including how you obtain samples, your period of interest, and your modeling objectives, as discussed below.
Determining the length and number of runs
When modeling terminating systems, the length of the simulation run is usually determined by the natural end point of the process being modeled. For instance, the 8 hours that a bank would be open is modeled for 8 simulation hours. For statistical analysis purposes, however, you may want to build a model that looks at a specific time period of a terminating system. For example, you could model the bank’s busiest time period (say from 11 AM to 2 PM) to run multiple times to get a better statistical picture of how the bank operates during that time period.
For non-terminating systems, the length of the run depends on how you decide to obtain your samples (as discussed above) and on your period of interest. Theoretically, a model of a non- terminating system could be run indefinitely. In reality, it is usually simulated until the output reaches an adequate representation of steady-state. For example, you would run a model of a manufacturing operation for a long enough period of time that you feel confident that every type of event happens at least several times. In other situations you might want to limit the run
How To

``````
 633   634   635   636   637