Suggestions on Experimenter and Report

The exprerimenter and report in the new version is greatly improved. I really like it. It saves me a lot of time for post processing the simulation results.

One bug exists when you View Results in Performance Measures. If your number is happen to be integer, the number will be displayed with hundreds of tailing 0s. For example, if it is 1, then you will see 1.0000000000000000000000... in both Performance Measurement window and the Experiment Report. See attached picture.

Here I have several suggestions:

1.
It would be good if we can choose only some of defined PFMs will go into report. This is good if you defined a lot of PFMs but some times you only want to report part of them.

2.
It would be good if we can choose only output graphics, and/or raw data.

3.
At the beginning of the experiment report, it would be good to output the following information:
Model file name;
Number of Scenarios;
Number of Replication Per Scenario;
Warmup Time;
End Time;
Scenario Settings (the table as you see under the Experimenter panel);

4.
For group type of PFMs, it would be good to choose what statistics are included in the report. Now it seems only Group Average, Combined Graph and Individual Graphs are outputed. Sometimes we may also want Group Total, Group Maximum ect. And if we have dozens of individual objects in a group, we may not want to include all of them in the report. The group statistics would be good enough.

5.
It would be good to put Save Experimenter Tree and Load Experimenter Tree in the experiment. Thus we can view previous experiment results if the saved tree is loaded. This can be done in Tree Edit Window. But it might be easier if it is put in the GUI. And this is a nice example to use Flexsim tree files. (Before Anthony told me, I was not aware of the "Load Into Node..." and "Save Node As..." items under the Right-Click menu of tree windows.

6.
It would be good to have a way to organize the experimenter outputs by scenarios and replications. The way I am using is to create a directory same as the model name, and under the that directory, there are two levels of sub-directory. The first level is scenarios, the second level is replications. So if I have a model calls flexsimmodel, after the experimenter of 2 scenarios and 2 replications each, I will have a directory structure like this:
-flexsmmodel
-+1
---+1
---+2
-+2
---+1
---+2

There may be other ways to organize simulation results from replications. But it would be nice at least to provide the capability of creating directories without using C++ code.

These are just my suggestions. People may agree or disagree. Flexsim is far from perfect. But I think all of us would like to see a better and better Flexsim simulation software. With our help and the great efforts of Flexsim team, it will.

It seems not many people using experimenter since I did not get any feedback for the post.

One update on the 1.0000000000000000000000... shown in the experiment report. It is not because of the integer as I originally thought. It is due to the following code in the node("VIEW:/standardviews/PerformanceMeasure>creategraph"):

int precision = 0;
double range = scenariomax - scenariomin;
while(range < 2*pow(10,1-precision))
    precision++;

If the range is 0, potentially we will have infinite loop. But due to rounding error, the while loop will exit when you have a very big number of precision. I found the following code can be used to fix the problem.

int precision = 0;
double range = scenariomax - scenariomin;
if(range>0){
    while(range < 2*pow(10,1-precision))
    precision++;
}else{
    precision = -trunc(log10(fabs(scenariomax)))+1;
    if(precision<2) precision = 2;
}

However, since the standardviews will be changed back to default. Each time when you open the Flexsim, you have to do the change again.

Tried to put in the development list, not sure if it is added.

Hey Alan,
Yeah, this was a change that was made to fix the issue where the range was very small. Before, it would only get to a precision of 2 decimals, which was problematic if your range is very small. This was fixed in 4.32, but it causes this problem if the range is 0. I've checked the development list, and your item has been added. I think the reason that you're not getting much response from the community is because it probably hasn't happened to anyone else. The only time it would happen is if the range is 0, which for a performance measure is very unlikely, because it indicates there is no variability in the different runs of your model.

Yeah. It could be very unlikely to get the 0 range. But if you have defined a list of PFMs and in a particular experiment, you want to check effects on only some of the PFMs, you may get the 0 range for some other PFMs.

I think a good design of experiment and report is critical for a serious simulation software.

By the way, I just noticed the "Posts below this point are new since your last visit." statement in the post. It is pretty good. I like it.

Thanks.
Alan

Hi,

how is it possible to save and load a Experimenter Tree?

Thanks

Congshi

Hi,I'm a student from Temasek Poly and currently using flexsim for my project. I'm facing a problem on how to generate the experiment and report. I want to know the total cycle time from source to sink,(the normal report only provide me the individual process min,max and avg stay time), how can i do this? and if i set experiment 30 replications, how can i get the report? Thank you very much for your reply and help. I'm looking forward to hear from you!

Hi,

we write a On Entry trigger code in the sink to get the cycletime and write this value to global table.
settablenum("globaltablename",row,col,time()-getcreationtime(item));

Jörg