Page 85 - ExtendSim User Guide
P. 85

Simulation Concepts 59
Additional modeling terminology
• Run the deterministic model twice to make sure you get the same results
• Output detailed reports or traces to see if the results meet your expectations
• Run a schedule of only one product line as opposed to several
• Reduce the number of workers to 1 or 0 to see what happens
• Uncouple parts of the model that interact to see how they run on their own
• Run very few or very many items through the model to determine if the model responds properly.
Other methods for verifying models include making sure that you can account for all the items in a model, animating the model or portions of the model, or using diagnostic blocks from ExtendSim’s libraries. For more information, see “Debugging Tools” on page 711.
☞ You can also verify that your models output the same results given a new release of ExtendSim. For more information, see “Automated test environment” on page 727.
Model validation
Once the model is verified you need to validate it to determine that it accurately represents the real system. Notice that this does not mean that the model should conform to the real system in every respect. Instead, a valid model is a reasonably accurate representation based on the model’s intended purpose. When validating, it is important to make sure that you know what to compare to and that you verify that measures are calculated in the same manner.
For validation, your model should accurately represent the data that was gathered and the assumptions that were made regarding how the system operates. In addition, the underlying structure of the model should correspond to the actual system and the output statistics should appear reasonable. While you would normally compare critical performance measures, it is also sometimes helpful to compare nonessential results that may be symptomatic and therefore show the character of the system.
One of the best validation measures is “Does the model make sense?” Other methods involve obtaining approval of the results by those familiar with the actual process and comparing sim- ulation results with historical data. For example, when validating model performance com- pared to historical data, try to simulate the past. If you have sufficient historical data, break the actual system performance into various windows of time, where all of the input conditions cor- respond to the input conditions for multiple runs of your model.
For more information, see “Debugging Tools” on page 711.
Additional modeling terminology
In addition to the following general information, each of the modules in this User Guide has a section with tips specific to that module. For additional general information about using ExtendSim, see also the How To module that starts on page 564.
Model parameters, variables, inputs, and outputs
A parameter is any numerical characteristic of a model or system. Parameters describe some- thing about the model and are known or can be estimated.
• An input parameter is a value that is required as part of the model specification.

   83   84   85   86   87