Quality controls to set a quality rate of the model

Question

question

Marcos MNA asked Jan 08 2021 at 3:29 PM Jeanette F commented Jan 12 2021 at 12:15 AM

Quality controls to set a quality rate of the model

Hello, I need to set a quality rate of my model and I have two quality controls in my model (one at the end of the Processor 2 and one at the end of the Processor 4). I have tried to set a model quality rate of 95% setting on these processors "Output-Send to Port-By Percentage" but I have noticed that it always gives me the same results... I mean, in all the different simulations I did, product number 5 (for example) is sent as "bad piece" and I want to have this randomly.

I attach my model below.

Is it possible to do with the free license? Thank you.

Prueba.fsm

Software Version:

FlexSim 20.1.0

flexsim 20.1.0 random experiments quality control

prueba.fsm (92.2 KiB)

· 1

Answer 1 · 2021-01-08T16:02:45Z

Marcello Rosadini answered Jan 08 2021 at 4:02 PM Marcos MNA commented Jan 10 2021 at 4:04 PM

Hi Marcos,

most likely, you need to disable Repeat Random Streams to generate different replications on each model run.

1610121644012.png (6.0 KiB)

· 6

Joerg Vogel commented · Jan 08 2021 at 4:23 PM

And in an Express License this feature is disabled. You get only results equal to the setting Repeat Random Streams. You can only run the model some time. Stop running, reset only the stats and then run a simulation as long as the experiment lasts. If you have different self set warmup times you get a pseudo randomness in your results, because you use a different amount of random values of the same streams.

0 ·

Marcos MNA Joerg Vogel commented · Jan 08 2021 at 5:22 PM

Could you explain me this more in detail please? I mean the last part of your answer, I guess it could be useful in my case! Thank you very much!

0 ·

Joerg Vogel Marcos MNA commented · Jan 08 2021 at 11:36 PM

You have only one set of random streams. If you reset your model the same number stream is injected in the same order as in every run. It is equal to the option Repeat Random Streams. If you let run your model longer the stream returns still the same set of values. But if you stop your model after 245 seconds run time and you reset the stats values you have implemented a manual warmup time. On the next run you reset only the stats after a time of 1224 seconds. This is different warmup time. Then you reset the statics and the result of your model differs, because you use a slightly different set of random values. It isn’t a statistically correct procedure to implement randomness, but it is one possible method to get different results. If you want to optimize this and your experiment should last for example 1200 seconds. You stop your model after 1200 seconds, reset only the stats and then you let run the model the next 1200 seconds. Each run has got a unique set of random values. You can combine this with manually activated warmups. In a discussion of randomness this procedure will be disqualified and has a poor level of randomness, but it is still a kind of randomness.

1 ·

Show more comments

Jason Lightfoot ♦♦ commented · Jan 08 2021 at 4:43 PM

If you have a license to run experiments then, running one run at a time it should repeat the random stream.

When you run an experiment the randomness will be applied even with the 'repeat streams' enabled - which you would need for debugging and verifying replications.

If you want to run a replication interactively one at a time you should change that in the experimenter.

0 ·

1610124203524.png (4.6 KiB)

question