Model Verification and Validation
Published: January 23, 2017
Categories: Consultant’s Corner
There are several key phases to any simulation project to ensure success:
- Plan the model through a functional specification.
- You can see the risks of modeling building without a functional specification in this article from Newsletter #1.
- Build the model in Arena.
- Verify and validate the model to ensure we can trust it to solve our problem.
- Run validated model through various alternatives.
- Analyze results to determine the best path forward.
In order to ensure that your model is correct, it is critical that it be verified and validated against actual plant performance. Without this step, there is no guarantee that your results will have any value whatsoever.
Verification and validation of your model is a skill that takes time to learn and often develops over each and every project. Here’s a few important things to consider in the verification and validation phase of your project.
Ensuring that the model behaves in a way it was intended, is most likely the easier of the two tasks. If the model is constructed in segments, then each segment should be verified separately as it is completed. It is a good idea to verify the model as each of the segments are put together. The final verification must be performed with the completed model. Adding some animation constructs can aide in the verification process. You can display key elements on the screen to observe important interactions occurring in the model.
For example you might display, the current simulation time (TNOW), the current number in queue (NQ) or counter value (NC). The resource status (idle, busy, failed, inactive), resource capacity (MR) or number busy (NR) of a resource. A few measures that may be red flags are larger than expected queue lengths, resources that are unexpectedly under-utilized or over-utilized, counter values that are smaller or larger than expected. There are multiple Smart examples installed with Arena you may reference for additional information for displaying variable values in your model. The animation of these variables can be helpful but should not be the only analysis tool.
During the verification stage, you should be sure to check the function of the model under extreme conditions. One condition, would be to create a single entity and follow it through your model logic. You could also consider using deterministic times that will allow you to more easily predict outcomes for simple simulation runs. Often modelers test the developing model at various stages under the same set of conditions. Occasionally, you should make extended runs to assure that the randomness in the model does not create circumstances that were not previously considered. You are trying to create a variety of different situations to verify your model.
Ensuring that the model behaves the same as the real system, can be difficult. If the system currently exists, then some kind of comparison can be made to ensure that the model represents the real world. If the system does not exist, but similar ones do, then the simulation results can be compared to the similar system, and at least a partial validation can be performed. If there is no real system to compare with the simulation, then true validation cannot be performed. If this is the case, you will want to be sure to dedicate time to the verification process with interaction with the proposed system experts. For additional validation information, please see Transforming Strategic Models into Operational Models from Issue #2 of the Arena Newsletter.
Verification and validation is an important part of the simulation process. As you complete each project you will become more adept at the process. Simulation is a valuable tool in any analysts set of skills and will improve your ability to sell change within any organization.
Sign up for the Newsletter
Arena User Group
Arena has a very active user community on LinkedIn. Ask questions, learn about best practices and connect with other simulation professionals.