Dashboard & Experiment Questions

Good morning - I have a few questions here:

1. On the dashboard if I have a line graph, how can I remove the time 1:00:00 - 1:08:00 from the graph. I am only simulating during working hours but the chart shows times that are not relevant. And to a certain extent will take that time into average - what am I missing here?
2. When using the Experiment feature, is there custom code or a place to change a the Process Time under an activity in the track? So if I wanted to show different processing times in different scenarios as its set to a variable?
3. Similar to #1, when I am using the Experiment feature how can I make sure that the data shown in the dashboard is only during working hours.

Thanks in advance for the feedback!

Robert

Originally Posted by Robert Rodgers

Good morning - I have a few questions here:

1. On the dashboard if I have a line graph, how can I remove the time 1:00:00 - 1:08:00 from the graph. I am only simulating during working hours but the chart shows times that are not relevant. And to a certain extent will take that time into average - what am I missing here?

I have a presentation tomorrow and I'm hoping to remove the excess time on 2 of my line graphs. The simulation only runs during business hours however it graphs all the earlier time of that day.


Thanks,


Robert

first workaround: use a table to collect your data and present this at a visual tool.

If you insist to use the dashboard, have a look at commands to manipulate the bundle data type.

An Example at.

Jörg

Another option is to start your simulation at 0:00 with the first events at 8:00. Add a user event at 0.001 with this code inside:

step();

With that the simulation time jumps from 0:00 to 8:00

Originally Posted by Robert Rodgers

Good morning - I have a few questions here:

(2.)When using the Experiment feature, is there custom code or a place to change a the Process Time under an activity in the track? So if I wanted to show different processing times in different scenarios as its set to a variable?

Yes, there is a construct called label in Flexsim, which act as a variable. Add a number Label to your processor. Set the value to your first process Time. Get this value and return it in the Process Time Function at your Processor with the command getlabelnum(current,"name of your label").
Change the label value in your experimenter scenarios.

Jörg

When running experiments, I think the most appropriate dashboard chart type is the box plot because it gives you distributed statistics over a user-specified period of time during the run and across multiple replications, but for a single scenario. The example I just posted here is a good example taken from an original post by AJ regarding the use of dashboards in an experiment.

Here's a screen capture showing box plot statistics for the length of stay of PCI 1 type patients for the scenario named "6-2-2" just during the daytime hours of 08:00 to 18:00.



Did you know you can change the time scale of the line graph with the scroll bar at the bottom of the line graph widget? There are three adjustments that can be made with the scroll bar using your mouse. Drag the left end of the bar to adjust the earliest time displayed in the graph window, drag the right end to adjust the latest time displayed in the graph window, and drag somewhere in the middle of the bar to slide the graph window left or right.

Here is an example of a line graph displaying staff utilization in a model where the staff don't really do anything until about 6am.



Here's a screen shot after I dragged the left end of the scroll bar to the right.



As for changing an activity's process time during an experiment, there are a couple options. One option Jörg described above is to use a label to hold the numeric value for the process time, and then simply reference the label value within the process time field itself with getlablenum() command. Take not that in Jörg's explanation, he had the might have been the process time of an Item Processing object in mind rather than the process time of a patient's activity, so you would need to adapt the references accordingly (i.e. patient rather than current). Also note that his example assumes you are changing a constant numeric value for the process time.

If you want to experiment with stochastic process times using different distributions for different scenarios, then you need an approach that will let you modify and execute distributions on the fly. You could still a label or a global table to store the process time, but it would need to be a label/table that stores text strings rather than numbers, AND you would need to execute the string to get a number out of it. There happens to be a picklist option called "Based on global table lookup" that lets you specify whether it should return a number from the table, or execute the string it looks up out of the table as an expression.

Having said that, I would suggest you simply use the experiment variable type called "Track Activity Parameter" as shown here:



Then simply select the particular track activity and parameter (i.e. ProcessTime) you want to change as shown here:



Then type in the distributions you want to use for your scenarios as I did in the "Scenario 1" and "Scanario 2" columns of experiment in my example as shown in both screen captures above.

Good luck,

Cliff