Basic Use of the Experimenter

I have a Processor with a mean cycletime of 100 sec. I want to see the effect of dropping the mean to:

1) 90
2) 80
3) 70
4) 60
5) 50


Attached are screenshots of the experimenter setup. I only want one replication per scenario. The results of the experiment show that it never gets past the default 100 sec…so it never indexes to the next scenarios.

Gavin,

Looks to me as the experiment is in the 5th scenario (cycle of 50). But with just 1 replication !!
As a rule of thumb you want to experiment with at least 5 replications per scenario.

Regards,
DJ

Everything looks right except that your Performance Measure tab says your object is Processor1 and your Experimenter tab says it's Processor2. Should these be the same object?

Gavin,

Does the attached model do what you describe?

Yes, I see but the results tell a different story (See attached).

Thanks guys, Alex was right on this one.

Gavin,

As in the model I sent you, make sure that your processor is using that label (Cycle) for its cycle time. And also as Alex mentioned, make sure that in your performance measure tab, that it is referencing Processor2 and not Processor1.

Using the experimenter, is there a way to have different scenarios for my processor which uses statistical distribution for my cycle time? I'd like to use different statistical distributions for my scenarios. Thanks for any help

Jamie,

If you are not going to be changing the distributions very often then you could have the experimenter change a label on the Processer. Then you could use a switch statement to choose the distribution. It would look like the following:

switch(getlabelnum(current, "experimenter"))
{
case 0: return uniform(5, 10, 1);
case 1: return normal(10, 2, 1);
default: return 10;
}

I hope that this helps. There are other way to do this that would allow you easier access to the distributions so that you could change them but they would be harder to code and slower to execute.

Good Luck,
Brandon