Randomness

[JaimeSotomayor94]

My company is developing a big model that simulates the arrival of passengers to an airport. It is important to us to control the randomness of the model for two reasons:

-          We want to be able to replicate the outputs of an experiment if the changes had not been significant.

-          When running different scenarios where only part of the configuration has changed, the part of the model that has not changed should remain the same.

It is possible to control the random generated number of the distributions that Simio offers, but we hadn’t been able to fix the randomness in the next cases:

-          When using the function RandomRow. Any change in the model that modifies the running time when the function is called changes the row selected.

-          When selecting a path from its weight. As well as with the RandomRow function, we have seen that only moving one meter the node that is connected to more than one path, the randomness of the path selected varies.

Furthermore, because of the dimensions of the simulation, some processes take time to end. That means that in some cases when we add more data to the input tables, data that is independent to a previous part of the simulation, everything changes because the process takes a little more time to end and that changes the result of the random decisions.

We would like to know if it is possible to control the randomness on this events, or what other alternative can we use.

There is a mini model attached that shows the described behavior. It consist on a source that generates 100 entities, and two possible paths with the same weight.

If the node that connects to both path is in the center of the facility, 45 entities go up and 55 down, like in the next picture.

If the node is moved to another location, the number of entities that go up changes to 48.

Finally, adding a process (with a loop that doesn’t do anything) before selecting the path, the number of entities varies again to 47.

Randomness_example.spfx