Cancel
Showing results for 
Search instead for 
Did you mean: 

experiment manager

Pioneer
Pioneer
Im running an experiment trying to analyse the energy use of a machine to try and reduce energy consumption.
However when i run the manager i dont receive any results. I receive the error 'The analysis of variance is impossible.
At least two experiments with random output required'.
Ive attached images of my inputs and outputs as well as the error message im receiving.
Is it possible to analyse energy in this way? many thanks for your help
7 REPLIES

Re: experiment manager

Gears Esteemed Contributor Gears Esteemed Contributor
Gears Esteemed Contributor
do you have any stochastic behavior in your model (e.g. failures, random distributions)? If not, every experiment will deliver the same result and you don't need the experiment manager.

How do you want to reduce the energy consumption?
Steffen Bangsow
freelance simulation specialist  
web: www.bangsow.eu
mail: steffen@bangsow.net

Re: experiment manager

Creator
Creator
thanks for your response. Im looking to analyse the total energy consumption of my shopfloor using the total energy consumption from each machine (i have around 13 machines using energy in total). Id like to see how changing the schedules of each machine will effect the overall energy at 30 min intervals (this 'overall energy' is coming from an excel datasheet not the machines as this total energy also includes building energy data).
So ideally, the end goal would be to determine where i have spikes in energy consumption and reduce this by changing the schedule.
I dont have any stochastic behavior currently.
many thanks

Re: experiment manager

Gears Esteemed Contributor Gears Esteemed Contributor
Gears Esteemed Contributor
then I think you need only to record the total energy consumption every 30 minutes. The easiest way is to write a method to sum up the consumption and call this method every 30 minutes (using a generator). The values you can write into a table. If you have different scenarios, write the values in different columns (one column one scenario), so you can compare them easily.
Steffen Bangsow
freelance simulation specialist  
web: www.bangsow.eu
mail: steffen@bangsow.net

Re: experiment manager

Creator
Creator
is this best done with the experiment manager or GAwizard/ GA optimisation? or neither? Im currently not finding anyway to include the shifts/schedule into the experiment manger or GA tools.
Im a little unsure on the differences between the experiment manager and GA tools.
thank you.

Re: experiment manager

Gears Esteemed Contributor Gears Esteemed Contributor
Gears Esteemed Contributor
you need to prepare the experiments, so that the experimentManager can handle them. E.g. if you want to change the shiftCalendar, you need to prepare for all variants one shift calendar, set up a table, where you insert all shiftcalendar objects, one globale variable to switch between the variants. Then you need a init-method that sets the shiftcalendar objects according to the selected variant (global variable). The global variable finally, you can set as input value for the experimentManager. The GA you can use, if there are too many experiments to run (e.g. more than 10.000).
In your case you can run the experiments manually. Change the shift calendar manually, start the experiment, record (copy) the consumption data into a result table.
Steffen Bangsow
freelance simulation specialist  
web: www.bangsow.eu
mail: steffen@bangsow.net

Re: experiment manager

Creator
Creator

thanks for your reply. what do you mean by a global variable to switch between the variants? is this a parameter that id like to change in the object eg cycle time etc? im guessing here that the variants would be different machines i am analysising.

many thanks

Re: experiment manager

Gears Esteemed Contributor Gears Esteemed Contributor
Gears Esteemed Contributor
the experimentManager will change this parameter. You can read the value of the parameter e.g. in an init-method an make all setting, that are necessary for the actual experiment (e.g. activate/deactivate machines, change the number of worker etc.)
Steffen Bangsow
freelance simulation specialist  
web: www.bangsow.eu
mail: steffen@bangsow.net