Experimenter: Performance Measures

I think I have the experimenter figured out for the most part. However, I’m having a problem setting it up to capture the correct data. I want to know how much time varies from scenario to scenario to move, say.... 45 flow items to a sink. So when I view results, I should see sim time varying across different scenarios. Do I make sense <crickets chirping>?

Thanks

Sure you make sense.
You're probably running into a problem when there are no more events to do, it stops running the model. To make the model run forever, execute code like this in an OnReset trigger somewhere (or Tools>Model Triggers>On Run Start):

createevent(model(),2147483647,1,"run model forever",model());

What I would do is set the simulation end time in the experimenter to be much larger than I think the model will ever take to run (2147483647 time units), then record the node Sink>stats/content/stats_contensince. If there is other stuff happening after all the items enter the sink, use an OnEntry trigger in the sink to somehow disable them, maybe by closing all the ports in the whole model.

All,

The reason Alex is using 2147483647 as the number is that this is the highest signed value that an integer can have. So trying to set the value of an ingeter higher (say by adding 1) would result in getting a negative value as you wrapped around to the lowest value.

Flexsim uses a double value for the event times so the largest value possible is 1.79769e+308. This should be more than is ever required.

Brandon

Yes, I know the upper bounds of the sim time for the model...no problem. Sim time will drop from the upper bounds after each senario is run. What do I put under the performance measure tab and what other steps do I need to do (say in the sink) to make this happen? Oh yes, nothing happens in the model once 45 items enter the sink (there are no other flow items).

For this specialized of a data capture, you'll have to write your own picklist option for the Performance Measure, something like this:

/**Last content change of an individual object:*/
/** \nObject Name: */
string objname = /**/"Sink1"/**strlistpath:MAIN:/project/model*/;
treenode current = node(concat("/", objname), model());
//pt(getnodename(current));pr();
//pt(getnodename(node(">stats/content/stats_contentsince",current)));pr();
return getnodenum(node(">stats/content/stats_contentsince",current));

There is an OnEntry picklist option called "Stop the Model Run or Replication" that was intended for these types of situations.

I am trying to get the mean value of a column of data in a global table using experimenter.
edit: I think I solved my problem. Thanks anyway.

Hello,

it's me again with a question about the experimenter.
Is it possible to define simulation run-time and warm up time as experiment variables?
I can select them in the tree view and assign values but it doesn't work properly, I can't make the simulation run for varying durations in each scenario.
Has anyone tried something similar to that before?

Thanks.

Hi Yasemin,

Here is a simple model to show how to do this. Check the code in the Start of Replication in Experimenter window.

Thanks for the code. Unfortunately, I got one error with this code. I want to use this to increase the warm up time successively while remaining the simulation time a the same. Therefore Iadd the same amount of time to both warm up time and simulation end time by the start of every scenario (not replication). The Output of every scenario stays about the same (slight differences comes due increased warm up time), but the processing state reduces with the longer warm up time. I guess that the output will correctly computed regarding the constant simualtion time, but the processing state is corresponds to the stoptime. And the longer the stoptime is, the lower will the processing time be.

Does anyone know the reason for this?

Thanks