Question about Queue, send to port, Random Available port and its randomness

Hey,

I am trying to compare two different warehouse storage assignment policies: class-based storage (ABC storage) and random storage in full-pallet order-picking.

ABC storage means that warehouse is divided into, for example three areas so that the itemtypes are located in those specific areas based on the demand rates of itemtypes. Thus, highest demand itemtypes are located inside area close to output point (A sector) and smallest demand itemtypes are located inside area farthest from the output point (C sector). Random storage means that any itemtype is located anywhere inside the warehouse.

My problem is related to random storage scenario. I want my simulation to start from some moment in time in warehouse. Hence, I want boxes to be in storage locations when order-picking starts. In source I have 10 arrivals: itemtype1 and its quantity, itemtype2 and its quantity, and so on, total of 10 different itemtypes. The arrival time for all arrivals is set to 0.00.

The results were really confusing for total travel distance of transporter (order-picker, travel with only one box at a time) in random storage scenario. The travel distance was smaller for random storage than for ABC storage what did not make any sense to me because I had demand curve of 20/76, that is 20% of products made 76% of total demand. The travel distance was 14km in random storage scenario. When I changed the order of itemtypes in the arrival list in source a little so that itemtype1 was last in the arrivals list and itemtype10 was first in the arrival list, the result changed little so that now it was 13,3km for random storage scenario. After this I tried one more thing, more accurately I imported from Excel a table which had 2062 arrivals so that one arrival represented only one itemtype and quantity of one. Now the results changed dramatically and the results made more sense to me because ABC storage had much smaller travel distance than random storage scenario, travel distance for transporter was 28,7km in random storage scenario.

I am trying to achieve a goal which is that Docking area queue would sent all the itemtypes and its boxes totally randomly to any storage location in racks, bays and levels. Transporter would then pick the boxes based on real FIFO, that is transporter picks the oldest box of specific itemtype first. I have queue, output send to port option, Random Available port but it seems that it won’t solely do the trick.

My question is, how does the queue random available port work? I mean if loads of itemtypes are sent to one rack or port, how do they spread around that rack or how do it solve the issue? Should I use 2062 arrivals in source arrival schedule (the quantity of itemtypes to be stored in racks) instead of 10 arrivals? It seems it has huge effect on results.

I attached model of random storage with 10 arrivals in source and random storage with 2062 arrivals in source so that one can see the differences in results which I explained earlier in terms of travel distance of transporter and the source arrival list.

Thanks beforehand!

Hi Johannes,

You are facing a problem of the pseudo random generator. If you load a model and start it, the results are going to be the same at each simulation run. You can prevent this, if you combine the start values with the system time. You find a switch in the environment tab of the Global Preferences.
The next issue depends on the use of only a single random stream in all statistical distributions. By default you use the random stream number zero. The randomness decreases with each used statistical distribution. You get a systematic fault. It is a good practice to use a different random stream at each statistical distribution to minimize the systematic effect. Back to your model there might be indeed a difference creating more flowitems at a single schedule arrival in comparison to create a flowitem at a schedule arrival event. The flexsim engine maybe process at each creation event all steps at the simulation time null and then goes to the next creation event at time null.
Maybe someone of the development can explain in detail how the engine process the creation of one or more flowitems in the source and the effects causing the process of the items at the same simulation time.

Jörg