Cancel
Showing results for
Did you mean:

# Re: Determination warm up period, run length, number of replications

Gears Esteemed Contributor

the different products I would create randomly (calculate the rate for each product), quantity by changing the interval of the source, and of cource non blocking

 Steffen Bangsowfreelance simulation specialist  web: www.bangsow.eumail: steffen@bangsow.net

# Re: Determination warm up period, run length, number of replications

Solution Partner Phenom

You now have a created a cyclic simulation model (weekly cycle). Therefore you introduce some bias that you need to take into account when you set the run-length (weekends, with no production).

For very long runs the effect will diminish, but when the run-length is only a few weeks long, the weekend losses do have a statistically significant effect. To avoid this, make sure the length is a multiple of 7 days.

Gert Nomden
Digital Factory Consultant
cardsplmsolutions.com

# Re: Determination warm up period, run length, number of replications

Builder

Dear @Steffen_B, @Nomden,

Do you agree with the following calculation in terms of the quantity of products?

Let`s assume that for a certain product there are 120 orders per year. The system is operational for 5 days a week for 5.9 hours a day (+-1500 hours a year). This are (5/7)*365 = 261 days per year. Hence, 120/261 = 0.46 products will arrive per day from this product type. I repeat this calculation for the complete product scope of 76 products and sum up the total amount of products per day. Assume this is 8.86. Then, 5.9 (working hours per day) / 8.86 = 0.67 hours = 40 minutes. Hence, I will set the negative exponential interval with a mean of 40 minutes and products will be selected based on the portion calculated above (0.46 for the example product). Lastly, the "blocking" must be cleared (deactivated).

Furthermore, @nomdem I will set the run length on a multiple of 7 days.

Kind regards.

# Re: Determination warm up period, run length, number of replications

Gears Esteemed Contributor

theoretically: yes

from a practically point of view, it is not secure enough

if you only have 120 orders per year and you will distribute the parts randomly, then you need I guess 100-500 years simulation time to get a statistical distribution around 120

a better way is a statistical method "pull without replacement" (in German ziehen ohne zurücklegen);

here you create an anonymous MU and you dice a product (e.g. with z_dEmp --> see help). After producing one part, you degrease the quantity of this part in the production plan and calculate a new empirical distribution for the next call until the production plan is empty --> stop simulation

 Steffen Bangsowfreelance simulation specialist  web: www.bangsow.eumail: steffen@bangsow.net

# Re: Determination warm up period, run length, number of replications

Solution Partner Phenom

The calculation looks plausible to me, if you want just a randomized mix.

I did notice that you filled in some parameters for the set-up behavior, but no set-up times. This may be on purpose, I just could not help to notice.

Gert Nomden
Digital Factory Consultant
cardsplmsolutions.com

# Re: Determination warm up period, run length, number of replications

Builder

Dear @Steffen_B,

Thanks for your reply! In principle, it works since I obtain somewhat the throughput that I expected after running the model for 1 year for instance.

However, if this is nog secure enough, then I really would to implement your suggestion into my model. Could you check the current arrival distribution in my model with the table file? I am not common with "pull without replacement", so could you please provide an example or some recommendations on how I should adjust my source?

Kind regards.

# Re: Determination warm up period, run length, number of replications

Solution Partner Phenom

How sure are you about the product ordering distribution?

Are they ordered totally independent from each other? Then you need many observations (excessively long runs do not make sense to me), as Steffen pointed out. But this could be the fact of life in your case.

Or is there some form of correlation? Then Steffen's method may have to be used. I would run some sensitivity-analysis on the product distribution anyway.

A hint on the frequencies: you can use absolute numbers, Product1: 120, etc. This can make life a bit easier...

Gert Nomden
Digital Factory Consultant
cardsplmsolutions.com

# Re: Determination warm up period, run length, number of replications

Gears Esteemed Contributor

you can check it by yourself

--> drain -> type statistcs -> detailed statistics table

 Steffen Bangsowfreelance simulation specialist  web: www.bangsow.eumail: steffen@bangsow.net

# Re: Determination warm up period, run length, number of replications

Builder

@Dear @Nomden ,

There is no correlation, so they are independent!

Concerning the set-up without times, this is indeed on purpose.

Could you please elaborate more on what you mean with your last sentence concerning absolute numbers? Can you provide an example?

Kind regard.

# Re: Determination warm up period, run length, number of replications

Solution Partner Phenom

Of course, I guess it is simpler htan you expected...

Gert Nomden
Digital Factory Consultant
cardsplmsolutions.com