Defining inter-arrival time for simulation

JT

Jim Thompson on March 18, 2011

I'm a beginner in the use of the simulation tool in ARIS with BPMN models.

I've learned enough to create a simple process model with one task to learn how to run the simulation tool.

From what I can tell, I specify the inter-arrival time of work using the Frequency group attributes associated with the BPMN Start event. These attributes seem to allow only specifying constant inter-arrival time.

For those familiar with queueing theory, my simple process model is a D/M/1 queueing model.

However, I'd like to specify other types of inter-arrival distributions. For example, I'd like to define exponential inter-arrival time to create a M/M/1 queue to test out the ARIS simulation tool. Since the M/M/1 has closed form formulas, I can validate the results I get from the simulation with this simple queueing model. This way I know if I've parameterized simulation model correctly.

For processing time, I can specify a variety of time distributions, e.g., deterministic, exponential, normal, ... I'd like to do the same for inter-arrival times.

Is there a way to specify a non-constant inter-arrival time distribution?

Thank you.

Tags: #simulation; #bpmn

Sign in or register to reply.

Notify Moderator

3 Replies

Ivo Velitchkov on March 19, 2011

Jim,

I don't think there is a straight-forward way to specify exponential inter-arrival time in the start event. But why don't you try the following: create a model with one Task. Select exponential distribution and choose Start Event instantiation and Processing time according to your simulation requirements.

Then make and occurence of the End Event as the Start Event of the real business process that you want to simulate.

I haven't tryed it myself but I believe it should work.

If you decide to try it, let me know, please. I'm really curious about it.

Like

Sign in
or register to reply.

Notify Moderator
Ralf Angeli on March 22, 2011

The frequency attributes are quite limited. As you've already noticed you can only set a fixed interval with those. You can get more control and flexibility with schedules and calendar events. A schedule is a model type which can be assigned to a start event. The calendar events which can be added to such a model serve as instantiators. And while they don't support distributions directly at the moment, you might be able to get an approximate solution for your problem by adding multiple calendar events which gradually instantiate more processes, either over periods of time or at single points in time. For example, 2 instances at 9am, 4 at 10am, 16 at 11am, 256 at 12am, and so on.

Information about schedules can also be found in the ARIS help topics below the group "Optimize and control processes / Simulate processes / Valuable information / Modeling / What can the Schedule and Composite schedule model types be used for?".

Like

Sign in
or register to reply.

Notify Moderator
Geoff Hook on March 23, 2011

Jim,

I work for Lanner we are the company that supplies the simulation software which is integrated within ARIS. We are UK based but operate globally, simulation is pretty much all that we do. Great to hear that you are using the ARIS Business Simulator...if you would like to talk about particulra issues the business faces and whether simulation might be relevant I would be very happy to schedule a call. Hopefully the advice from Ralf has helped with your specific question....I reckon Ivo's method would work, but may proviode challenges with statistics. essentially you can create tokens using an inter-arrival time and specify a quantity to arrive at that time (both values can be sampled) or use the schedule approach. The schedule allows you to specify a volume that arrives in the schedule period. There are some posts on simulation in ARIS Community which may be of help...

best regards

Geoff

Like

Sign in
or register to reply.

Notify Moderator

There are currently no upcoming events.

Sign in

Reset Password