Repeat Random Stream with model parameters

Question

question

Sebastien asked Mar 16, '21 Phil BoBo commented May 11, '21

Repeat Random Stream with model parameters

Hi,

We have a model where we can change parameters of a system within Global Tables, that changes for instance the arrival rate of the material flow.

If we have our first simulation run with x=20 and then use another value like x=30 and run the simulation again, we obtain different results with our models which is normal.

But once we change it back to 20, should we have the same results or is it normal that we obtain different random values ?

I don't know exactly how the random stream works in that case. When does it change ?

Software Version:

FlexSim 20.2.3

flexsim 20.2.3 random streams

· 1

Ben Wilson ♦♦ commented · Mar 19, 2021 at 03:35 PM

Hi @Sebastian, was one of Ralf Gruber's answer helpful? If so, please click the red "Accept" button at the bottom of one of the answers. Or if you still have questions, add a comment and we'll continue the conversation.

If we haven't heard back from you within 3 business days we'll auto-accept an answer, but you can always unaccept and comment back to reopen your question.

0 ·

Answer 1 · 2021-03-16T10:36:03Z

Ralf Gruber answered Mar 16, '21 Phil BoBo commented May 11, '21

Sebastian,

if Random Streams are set to repeat, the same initial model should get you the same results with the same parameters. Are you sure, your model is always reset to the same initial status? E.g. do you put Task Executers on a designated spot at reset, or do they start where ever they have been left in the previous run?

Check your model for that

Good luck

· 9

Joerg Vogel commented · Mar 16, 2021 at 10:45 AM

Repeat Rand Streams is an option in Main Menu “Statistics“.

0 ·

Sebastien commented · Mar 16, 2021 at 12:48 PM

@Ralf Gruber thank you for your answer. As far as we know all the necessary values and Task Executers are reset and the random stream option in statistics is activated. We went through the manual and checked all the points adressed there.

Yet, we have'nt initialised all the variables to a default value but I don't think we would have some repetition then. We will look futher into it. Maybe we missed something otherwise we will try to initialize all the variables to a default value.

0 ·

Sebastien Sebastien commented · Mar 16, 2021 at 01:42 PM

@Ralf Gruber. How is it possible that our model is repeatable on our computers but not on a computer from a partner, knowing that he did not change anything to the model ?

0 ·

Ralf Gruber ♦ Sebastien commented · Mar 16, 2021 at 03:08 PM

For as long, as they do have a set value, you should be fine.

0 ·

Sebastien commented · May 11, 2021 at 10:55 AM

@Ralf Gruber,

Sorry for digging up the subject again. We dug deeper in the issue. The Event Log shows a difference once the first patient comes into the system and should walk to the first waiting line. The only difference is with the parval(2) although the events happen at the same time this parval(2) is different (parval(2): 2516010663936,000000 != parval(2): 2512319938560,000000). This is the line in the log:

What is this variable and how is it that it is different between two model runs ?

410,502655

BeginTask: NodeFunction

/Patient1

/Tools/People/Flows/Activities/Walk/Travel>variables/callbackNode

parnode(1): /Tools/People/Flows/Activities/Walk>stats/instances/Patient1

parval(2): 2516010663936,000000

parval(3): 0,000000

parval(4): 0,000000

0 ·

Phil BoBo ♦♦ Sebastien commented · May 11, 2021 at 12:40 PM

That's just a pointer difference. That's expected.

When using the event log to find why a model isn't repeatable, you need to find the first event where the time is different. Whatever generated the event with a different time is the source of the repeatability issue. Having different memory addresses in the params is expected.

1 ·

Sebastien Phil BoBo ♦♦ commented · May 11, 2021 at 01:09 PM

Oh I see. Thank you ! I'll look deeper into the issue.

0 ·

Show more comments

Answer 2 · 2021-03-16T15:24:14Z

Ralf Gruber answered Mar 16, '21

Are you doing any calculations OnDraw? These would be dependent on screen refresh rate, which could explain different results on a different computer. You should never do that. Or are you maybe reading positions of objects without calling updatelocations, e.g. items on conveyors?

These are all guesses, the model would probably help ;)

question