Experimenter functionality

Experimenter functionality

pieter_cecat
Not applicable
84 Views
1 Reply
Message 1 of 2

Experimenter functionality

pieter_cecat
Not applicable

[ FlexSim 16.0.8 ]

Hi

I am wondering what is the best way to test the sensitivity of my simulation model with respect to total cost. I am trying to use the experimenter, but don't know if I am doing it correctly as it keeps on returning the same value for each scenario. My settings are as follows:

Three performance measures:

-> the total cost (= a cell in a global table)

-> the service level (= a cell in a global table)

-> the fill rate (= a cell in a global table)

As variables:

-> a cost table (= an entire global table)

As scenarios:

-> several alternative cost tables for the cost variables (generally 3 alternative tables containing costs). Is it possible to change a global table entirely without having to declare every cell individually?

As replications:

-> How many is generally sufficient to get reliable results for each scenario within reasonable time?

As export:

-> Save values of the three performance measures for every scenario to global table/graph

Please find attached the model. mean.fsm

0 Likes
Accepted solutions (1)
85 Views
1 Reply
Reply (1)
Message 2 of 2

SCHamoen
Advisor
Advisor
Accepted solution

@Pieter Cecat

1. In your experimenter you have the variable from the table OverTimeCost. But when I search the model this parameter doesn't seem to be used. So changing it in the experimenter has no impact so that is why your results are always the same.

2. If you want to change a whole table that could be done by using an array with tablenames and using a global variable or label as index. So where you use the cost you can write gettablenum( TableArray[getlabel(current,"index")] and as variable you select the index label and give it values 1,2,3...

Other option could be that you create a user event that goes off at time = 0 and then copy the correct table into the other table.

3. How many replications should be sufficient? Normal approach would be to make the runs and look at the spread of the result. If the spread is too large you can either make more replications or have a longer runtime.

4. concerning results. If you have run your experimenter at the bottom you can look at your results with the View Results button. There you can export everything.

5947-view-results.png

0 Likes