Assignment #4 - GPSS

Modeling and Simulation (308-522A)

David Bélanger
dbelan2(à)cs.mcgill.ca

Problem

Assumptions

I am assuming the following:

Model

ATOM3 was used to draw the model and generate the GPSS code.

A screen grab of the model is shown below. It may be easier to read the numbers in the postscript version: model.ps.

Description

Each time unit corresponds to 1 minute. This gives enough precision to express the half hours quantities in integers.

The arrival process of ships is modeled by the generator blocks. A new ship arrives at every hour plus minus one half (60 plus minus 30 minutes).

Storage 1 of size 6 restricts the number of ships in the harbour to the number of berth (available space). Queue 1 represents the ships waiting outside the harbour for a free berth.

After a ship was able to get inside, it starts to be unloaded if one crane is available (represented as Storage 2 of size 5, five cranes). Otherwise, it waits for a free crane (Queue 2). Note that there can be at most one ship in Queue 2 as there can be no more than six ready to unload. Unloading takes 7 1/2 plus minus 3 hours (450 plus minus 180 in minutes). This is represented by an ADVANCE block.

After a ship is unloaded, the crane is free to service the next ship waiting. This explains the LEAVE block for storage 2. Now, 10% of the ships may need to be refueled. The decision is done by a TRANSFER block. Refueling takes 1 plus minus 1/2 hours (60 plus minus 30 minutes), modeled by an ADVANCE block. When a ship leaves the harbour (refueled or not), it leaves it places to the next one (a LEAVE for Storage 1).

Finally, we want to simulate for 100 ships leaving the harbour. A TERMINATE block that increments the counter by 1 every time a ship leaves is added.

Generated Code

The code generated by ATOM3 is found here.

Output

Runs were done using the Web GPSS Simulator.

Run 1

Complete output is here.

Some comments:

Simulation begins.

RELATIVE CLOCK: 9458.6161   ABSOLUTE CLOCK: 9458.6161

The clock time makes sense. Unloading takes on average 450 minutes per crane. If the work is equally divided, then each crane unloads 20 ships (450 * 20 = 9000). Plus some time before cranes can be get busy. Plus some delay due to refueling.

BLOCK CURRENT     TOTAL  BLOCK CURRENT     TOTAL  
L0                  158  FUEL                 12 
L1         53       158  L10                  12 
L2                  105  DONE                100 
L3                  105  L11                 100 
L5          1       105 
L4                  104 
L6                  104 
L7          4       104 
L8                  100 
L9                  100 

L1 makes sense. L1 is the numbers of ships queued, waiting for a berth. 53 makes sense as ships arrive much faster than they can be unloaded. Assuming the last five ships are being unloaded and 750 * 5 = 3750 minutes = 62.5 hours to unload the ships, around 62 (usually less as ships may be already half unloaded) may arrived.

L5 is the queue for the crane. It must be at most 1 as 5 cranes are available and at most 6 ships may be ready (in a berth) at any time.

L7 is the advance block for unloading. There can be at most 5 ships unloading at the same time. Four is ok.

There could be some ships in the FUEL block when the simulation ends. In this run there are none (see run 3).

The total of ships (transition) that went through the blocks also makes sense. Note the 12 in FUEL is about 10% (10% is 10 ships). Also, there are no more than 100 that went through the TERMINATE block as this was the terminating condition.

          --AVG-UTIL-DURING--                                                                                          

 STORAGE  TOTAL  AVAIL  UNAVL     ENTRIES    AVERAGE   CURRENT  PERCENT    CAPACITY     AVERAGE     CURRENT     MAXIMUM
           TIME   TIME   TIME               TIME/UNIT   STATUS   AVAIL                 CONTENTS    CONTENTS    CONTENTS
       1  0.979                       105     529.320    AVAIL   100.0            6       5.876           5           6
       2  0.982                       104     446.610    AVAIL   100.0            5       4.911           4           5



   QUEUE     MAXIMUM      AVERAGE       TOTAL        ZERO     PERCENT      AVERAGE      $AVERAGE     QTABLE     CURRENT
            CONTENTS     CONTENTS      ENTRIES     ENTRIES     ZEROS      TIME/UNIT    TIME/UNIT     NUMBER    CONTENTS
       1          53       26.673         158           6        3.8      1596.783      1659.814                     53
       2           1        0.881         105          12       11.4        79.345        89.583                      1



  RANDOM    ANTITHETIC     INITIAL     CURRENT      SAMPLE   CHI-SQUARE
  STREAM      VARIATES    POSITION    POSITION       COUNT   UNIFORMITY
       1           OFF      100000      100375         375      0.96

The average contents of storage 1 and and storage 2 is almost equal to their respective capacities. The arrival rates being small compared to the processing rate, the berth and cranes are almost always kept busy (except at the beginning). Both have reach their maximum content.

The average contents for the crane queue is 0.881 out of a maximum of 1. For the berth it is 26.673 and it reaches a maximum of 53. The $average time/unit is 1659.814 minutes waiting for a berth. This is 27.65 hours! Once a ship has a berth, it waits 89.583 minutes (1.5 hours) on $average to get a crane.

Run 2

The simulation was runned 3 times consecutively without reseting the random number generator. Similar results were obtain in all three meaning that the results are consistent through the runs. Nothing really interesting.

Complete output is here.

Run 3

Complete output is here.

As previously explained, it is possible to have one or more ships in the FUEL block when the simulation ends. The simulation was runned several times and it was obtained. Note that there is no ship waiting for a crane. This can be explained as follows. Just before the end, 5 ships were unloading and one ship was done unloading and was getting refueled. No ships were waiting in the crane queue as no berth were available. Then 1 ship finished unloading and left. The simulation terminated. No ship entered in the berth freed.

BLOCK CURRENT     TOTAL  BLOCK CURRENT     TOTAL  
L0                  150  FUEL        1        13 
L1         45       150  L10                  12 
L2                  105  DONE                100 
L3                  105  L11                 100 
L5                  105 
L4                  105 
L6                  105 
L7          4       105 
L8                  101 
L9                  101 

Conclusion

The harbour would require more berths and more cranes to be efficient. The ships are accumulating waiting to enter the harbour. Continuing the simulation for a longer time would probably yield a longer queue.
$Revision: 1.2 $