Arena is innovating simulation of high speed systems.  We’re solving bigger problems, faster, and more accurately to help our customers remain competitive in their fast paced business.

Past Approach

The modeling of high speed systems such as canning and bottling lines has always been a challenge using Discrete Event Simulation. With such systems there is usually a large number of entities in the system. This large number of entities along with a significant number of attributes can cause a Discrete Event model to run very slowly.

In the past we have usually approached this problem in one of two ways:

• Batching the entities so that we have fewer entities in the system concurrently.
• Using a continuous approach within the framework of a discrete language where the system is modeled as rates and levels.

Both approaches do work but in both cases one may encounter certain issues.

Maintaining Fidelity

When one is batching entities, the question becomes what should the batch size be in order to maintain a certain level of fidelity for the model. And this batch size may not be easily achievable because a given batch size for one process might not be appropriate for another process. This can be addressed by changing batch sizes, but this can lead to more complexity in the model. And one would like to keep the model as “simple” as possible so that verification and validation can be easily performed.

Maintaining Accuracy

For the continuous approach, since we are no longer advancing the time from discrete event to discrete event, one has to decide upon how often the levels are updated. This is sometimes called the step or delta for the reevaluation of the model status. This can lead to misleading results if this step is too large or to very long run times if the step is too small. So to get it just right can be a challenge.

Innovating with Arena – Any level of complexity and Any level of detail

With the advent of Arena 15, we can now model the system to the level of each entity. The model can concurrently have millions of entities without jeopardizing the run time of the model. So we need not worry about batch sizes or steps in our modeling efforts. We can more clearly maintain the fidelity of the model leading us to better statistical outputs to analyze the given system. The power comes from the strategic decision to make Arena 15 a native 64 bit application, opening our users to a whole new world of modeling capabilities.

Not only can the model have large numbers of concurrent entities but the model can also have a large number of attributes for each entity. Again this leads to a better model of the real world system having to make fewer assumptions to get the model to work. All in all this leads to a simpler approach for the modeling effort allowing the analyst to have more time to use the model in the analysis portion of the project.

Improving runtime from 45 minutes to 28 seconds

Just to punctuate the ability of Arena 15, a continuous approach model that took 45 minutes to complete a 12 hour run was converted to a discrete even model not using batching. When converted to the discrete event approach and not batching, the model ran 12 hours of simulated time in 0.46 minutes. Now, we are not guaranteeing that we can have that kind of decrease in run time for all models, since all models behave differently, we can certainly say that with Arena 15 the run times for certain models can definitely be improved.

More importantly the ability to use Arena’s Discrete Event environment under 15 gives one the ability to model any level of complexity and any level of detail. This allows the modeler to capture the fidelity to fulfill the objective of the model as described in the project’s functional specification.