question

The raninit() command is the correct way to set the seed value of a random number stream.

Using the Toolbox window, add Modeling Logic > Model Trigger > On Reset to your model.

In the OnModelReset trigger, execute the randinit() command to reset the seed value for whatever streams you want to set. For example:

randinit(8, 1234);

Then, when you sample a distribution from that stream, it will be sampled using the specified seed. For example, if you then subsequently call the following code, it will return 0.1160 because you initialized stream 8 to seed 1234:

return uniform(0, 1, 8);

Subsequent distribution calls using stream 8 will continue to sample from that stream that had been initialized with seed 1234.

Attached is a simple example using Process Flow. randinit_example.fsm

If you reset the model and then press Step twice, the token will be created at time 4.53. If you then comment out the randinit() call in OnModelReset, apply it, reset the model, and then press Step twice, the token will be created at a different time based on the default seed value of stream 8. If you then uncomment the randinit() call, apply, reset the model, and then press Step twice again, you'll see that the first event is again at 4.53 because that's what the distribution returns when sampling from the stream that was initialized to seed 1234.

The getstream() command returns a stream number that is unique for the object that you passed to it.

When running a simulation, it is a good practice to use different random number streams for different objects and activities. This is discussed at length in various Discrete Event Simulation textbooks. The getstream() command is an easy way to accomplish that result. FlexSim will initialize those streams automatically (if Repeat Random Streams is checked).

If you want direct control over the streams and seeds you use, then you should pass the particular stream and seed values you want to the various distribution functions in your model, as shown in the example above that uses stream 8 initialized with seed 1234.

If you run the above example, you will get the exact same results every time, including when using the experimenter. To get different results in the experimenter, you'll need to initialize your streams in the Start of Replication trigger to different seed values depending on the scenario and replication number.

If you build your entire model this way with a different stream number for each activity and object, calling randinit() in On Model Reset and Start of Replication to initialize each of those random number streams to unique seed values for each scenario/replication combination, then you will have replicated what FlexSim does automatically for you when simply using the Repeat Random Number Streams checkbox and the getstream() command.