hi all,
I want to generate customer weekly demands which are normal distribution(for example, mean=100 and variance=10).
However, in the Source Atoms, only the interal-arrival time of producing each demand can be set.
Is Inverse Gaussian distribution useful in this case or
is there other way to generate demand with mean value and variance?
i hope my question is clear
thank you in advance
Jimmy
Demand and interal-arrival time
Re: Demand and interal-arrival time
PS: I tried to set Inter-arrival time like this :
hr(72/normal(125,1)), where 72 hours is the total active time of a week and 125 is the expected mean demand.
each week, it do creat demand around 125.
But when I increase the variance, the mean demand doesn't remain as 125.
hr(72/normal(125,1)), where 72 hours is the total active time of a week and 125 is the expected mean demand.
each week, it do creat demand around 125.
But when I increase the variance, the mean demand doesn't remain as 125.
Re: Demand and interal-arrival time
Hi Jimmy,
I suggest to make use of the arrivallist atom, you'll find that atom in the category TIME of the library. You could use this to generate each week 1 arrival (repeat time 72 hours) where you set the quantity of that arrival to a number normally distributed.
In this way the total number of customers will arrive at the beginning of the week. I expect that you'll want to have the customers arrive randomly over the week. You can do this by adding a multiservice (category PROCESSES) behind the arrivallist. Set the capacity to a big number and the process time to uniform(0, hr(72)). This will spread the customers uniform over the week, the time between the customers will be negative exponentially distributed.
Regards, Marlies
I suggest to make use of the arrivallist atom, you'll find that atom in the category TIME of the library. You could use this to generate each week 1 arrival (repeat time 72 hours) where you set the quantity of that arrival to a number normally distributed.
In this way the total number of customers will arrive at the beginning of the week. I expect that you'll want to have the customers arrive randomly over the week. You can do this by adding a multiservice (category PROCESSES) behind the arrivallist. Set the capacity to a big number and the process time to uniform(0, hr(72)). This will spread the customers uniform over the week, the time between the customers will be negative exponentially distributed.
Regards, Marlies
Re: Demand and interal-arrival time
dear marlies,
Thanks for your reply. This is excatly what i need.
the uniform(0, hr(72)) in multiservice randomly spread the weekly demand into days.
Is that possibible to specify the different weights to spread for each days?
For example:
monday(0-12) takes 12% of the weekly demand, tuesday(12-24) takes 11% etc..
Best regards,
Jimmy
Thanks for your reply. This is excatly what i need.
the uniform(0, hr(72)) in multiservice randomly spread the weekly demand into days.
Is that possibible to specify the different weights to spread for each days?
For example:
monday(0-12) takes 12% of the weekly demand, tuesday(12-24) takes 11% etc..
Best regards,
Jimmy
Re: Demand and interal-arrival time
Hi Jimmy,
That's possible, you could of course make an arrival in the arrivallist for each day instead of per week.
What you could do as well is: add an empirical distribution atom to your model (category TOOLS), create 6 classes and define the chance per class (= day in the week). Note that the percentages need to be cumulative.
In the multiservice (or in the arrivallist by assigning a label to your products) determine the day of the week by calling the empiricaldistribution. The servicetime of the multiservice will be something like this:
MyEmpiricalDistribution * Uniform(0, hr(12))
Regards, Marlies
That's possible, you could of course make an arrival in the arrivallist for each day instead of per week.
What you could do as well is: add an empirical distribution atom to your model (category TOOLS), create 6 classes and define the chance per class (= day in the week). Note that the percentages need to be cumulative.
In the multiservice (or in the arrivallist by assigning a label to your products) determine the day of the week by calling the empiricaldistribution. The servicetime of the multiservice will be something like this:
MyEmpiricalDistribution * Uniform(0, hr(12))
Regards, Marlies