Exception error with dispatcher causes staff to freeze - call center model

Question

question

Christopher Morgan asked Apr 28, '17 Cliff King answered May 4, '17

Exception error with dispatcher causes staff to freeze - call center model

Hello,

I've been developing a fairly basic call center model and have run across an odd exception error. I was hoping i could get some feedback on what might be the root cause. Interestingly, the error only seems to fire for "Staff_1" and only when scheduling is activated. After the error "staff_1" becomes inactive.

I've attached my model and a txt of the error. Any feedback would be greatly appreciated

error-log.txt

callcenter2.fsm

Software Version:

FlexSim HC 5.1.0

exceptions exception message

error-log.txt (1.7 KiB)

callcenter2.fsm (268.4 KiB)

· 3

5 |100000

Attachments: Up to 12 attachments (including images) can be used with a maximum of 23.8 MiB each and 47.7 MiB total.

Sam Stubbs ♦ commented · May 02, 2017 at 01:14 PM

I still haven't figured the error issue out. But I did have a thought: Do you have access to a standard FlexSim license (as opposed to just HC)? Because this particular model of a call center, seems like it would be better suited to the standard software.

0 ·

Christopher Morgan Sam Stubbs ♦ commented · May 02, 2017 at 02:17 PM

Unfortunately i don't have any other licenses for any other packages. This is a bit of a one off for us so i was trying to figure out a creative way to make it work.

Is there anything you can think of that i should try to start debugging the problem?

0 ·

Sam Stubbs ♦ Christopher Morgan commented · May 02, 2017 at 04:58 PM

@Matthew Gillespie

0 ·

Answer 1 · 2017-05-03T17:37:19Z

Cliff King answered May 3, '17

This post dealt with a similar problem caused with patients leaving the model early per the "Patient Leaves Early" option for a waiting room. Problems can occur if a patient gets up out of the waiting room and exits the model when they have already called for and/or allocated a staff member for the next activity. In the other post we helped the user by adding some checks in the Patient Leaves Early code to not let the patient leave early if they've already allocated staff, and then later on in the same post, we realized it was also necessary to terminate the escort activity with a little delay. I've implemented these changes with the exception of using a shorter delay time (0.01 rather than 0.1) for the terminateactivity command to try and match the timing of your model better. The model doesn't throw any exceptions now, but it has other problems that I need your help in resolving because I don't understand some of the timing issues in your model. Play around with this model a little bit and see if you can figure out what's missing to get the patients to stop backing up so bad in the queue. If we can't figure out a solution quickly, I'd be happy to help you implement another modeling solution to for dropped calls that doesn't involve using the "Patient Leaves Early" functionality of a waiting room.

callcenter2-ck.fsm

callcenter2-ck.fsm (268.7 KiB)

5 |100000

Attachments: Up to 12 attachments (including images) can be used with a maximum of 23.8 MiB each and 47.7 MiB total.

Answer 2 · 2017-05-04T19:32:37Z

Cliff King answered May 4, '17

Here's a small example of how to model a certain percentage of patients leaving the model early because they are tired of waiting. The patient's tolerance for waiting (Lou calls this the impatience curve) is defined by a weibul distribution. Rather than use the "Patient Leaves Early" option for a waiting room, this model triggers a delayed start of an LWBS activity in the patient's track, and then later terminates the event if the patient leaves the wait queue. Here's a list of the key modeling points of the model:

Activity 30_Go to Waiting Room uses the "Start an Activity" pick list option in its Activity Finished Trigger to schedule the start of activity 31_LWBS after waiting weibul(90.4, 2.08, 73.3) minutes.
Activity 31_LWBS has the following Activity Start Condition so that only a percentage of patients who exceed their maximum tolerance for waiting will actually leave the model: bernoulli(30,true,false). This start condition can be removed or set to bernoulli(100,true,false) if you want every patient who exceeds their impatience curve sample time to leave the model early.
Activity 40_Bed Placement activity has activity 30 as its predecessor, but has an Activity Start Condition defined so that it will not actually start until both an exam room and an escort available.
Activity 40_Bed Placement has an Activity Start Trigger to destroy the pending LWBS starting event once activity 40 actually begins.
Activity 60_Departure has an Activity Finished Trigger to notify any patients who may be waiting in the waiting room, that a bed and/or nurse are now available. This is done by using the "Start an activity.

lwbs-based-on-threshold-waittime-and-percentage.fsm

lwbs-based-on-threshold-waittime-and-percentage.fsm (113.2 KiB)

5 |100000

Attachments: Up to 12 attachments (including images) can be used with a maximum of 23.8 MiB each and 47.7 MiB total.

Answer 3 · 2017-05-03T20:19:54Z

Cliff King answered May 3, '17 Christopher Morgan commented May 4, '17

You might be interested in taking a look at this Hospital Call Center model and paper that @Lou Keller put together awhile back. It's a simplified model demonstrating a technique for building a call center model using just Item flow objects. It might give you some ideas. I'm going to look for a good example for modeling LWBS (Leave Without Being Seen) patients as well, because it will give you some good ideas as well.

hospital-call-center.zip

model-screen-capture.png (791.4 KiB)

hospital-call-center.zip (3.5 MiB)

· 1

5 |100000

Attachments: Up to 12 attachments (including images) can be used with a maximum of 23.8 MiB each and 47.7 MiB total.

Christopher Morgan commented · May 04, 2017 at 03:54 PM

Lou's model was actually where i got the idea to try modeling this process. Its a great example!

I had to shift away from utilizing items since we wanted to look at the arrival rates per hour of day since there is a large variation in demand by time of day. Since Scheduled arrivals by hour of day wasn't available for arrival options, I chose to model the calls as "patients" and compensated for travel time by experimenting with character speed (i think i edited this out of the model i uploaded while troubleshooting the errors).

0 ·

question