Page 84 - ExtendSim User Guide
P. 84

Simulation Concepts
The modeling process
• Where is the required data? It is useful to start collecting data early in the model building process because it can often take a while to obtain all of the necessary information. You will also need to know if the input data is an absolute value or if the data is from a statistical dis- tribution. Additional data requirements may surface once the model building process has begun. Your model may, for example, lead you to explore alternatives that had not been con- sidered before.
• How shall the model be conceptualized? Before even running ExtendSim, think about what the various components of the system represent. Roughly determine the time delays, resource constraints, flows through the system, and any logical actions that occur in the model. This will help you determine how to build the model.
• Whatalternativeswillbeinvestigated?Althoughthemodelmayleadyouintonew,unex- pected directions, try to think ahead so that the model can be easily changed from one alter- native to the next.
☞ It is common to use a constant or a uniform (integer or real) distribution in the early stages of model building so that modeling problems and variations can be more easily detected. After the model is verified, you can easily change the distributions to correspond to real-world pro- cesses.
Refining models
It is important to remember that models may not give you a single “correct” answer. Instead, they make you more aware of gaps in your thought process. These problems may involve over- simplification in the model, false assumptions on your part when creating the model, or miss- ing connections between parts of a model. Refining your model step by step helps eliminate these and other pitfalls.
Every model can be made more complex by adding assumptions and interconnections. The model-building process commonly begins with the creation of a simple model. After analyzing the simple model, complexity is added, followed by further analysis, the addition of more com- plexity, and so on. The complexity takes one of two forms:
• Takingoneblock(aprocess)andturningitintomanyblocks(amorecomplexprocess)
• Adding a connection between two previously unrelated blocks, usually through a mathemat- ical operation (finding an interconnection between two processes)
At each step, look at your results and make sure they make sense relative to the data. If you can, verify the results in the real world. If one result is way off, check the output from each step to determine where the process went awry.
Model verification
The process of debugging a model to ensure that every portion operates as expected is called model verification. In the tutorial, you performed part of this verification process by building the model in stages and with minimal detail, then running it at each stage to observe the results. A common verification technique could be termed reductio-ad-
absurdum (reducing to the absurd), which means reducing a complex model to an aggressively simple case so that you can easily predict what the outcome will be. Some examples of “reduc- ing to the absurd” are:
• Removeallvariabilityfromthemodel,makingitdeterministic

   82   83   84   85   86