Title: A container allocation model at container terminals with double wide gantry cranes
1A container allocation model at container
terminals with double wide gantry cranes
- 11-04-2007
- Presentation by
- Karthik Mohan
- Advisor Dr. Anne Goodchild
- Civil and Environmental Engineering
- University of Washington, Seattle
-
2Problem statement
- Minimize processing time containers at storage
yard container terminal double-wide yard gantry
cranes
3Motivation and significance
- Container terminals efficiency operating costs,
congestion, throughput - Operational efficiencies New container terminal
system LVRT (Low Viaduct Horizontal Rail
Transportation System) Double wide yard crane,
horizontal rail interface, Automated Guided
vehicles (AGVs) - Application to non-automated, automated systems,
DSS
4Generic container terminal
5 LVRT system
LVRTS / Automated Container terminal system
Water side equip. Gantry crane
Interface Two way Rail system, Gantry Crane
on viaduct, AGVs
Land side equipment Double-wide Gantry crane
on rail, AGVs
6 LVRT system contd.
- Quay crane
- Container ship
-
- Ref Liangcai Dong, Anne Goodchild, and Hanzhi
Ding. A Trolley Routing and Scheduling Strategy
for Low-Viaduct Rail Transportation System with
Double-Wide Yard Crane. Transportation Research
Board Annual Meeting 2007
7Assumptions in the study
- Problem scope
- Half-bay import/export
- Two trolleys and three cranes
- Capacity of bay 60 containers
- Influence area
- Containers continuously loaded and unloaded from
ship Imports picked hoist - Export containers previous day
- Assume no rehandling of containers required
8Problem formulation
Export/ Import assignment
Transfer and Drop-off/pick-up
Non-overlapping Influence areas
Bay capacity
3) Constraints
Minimize Processing time
LVRT problem
1) Objective
2) Decision variables
Import transfer
Export transfer
Import drop-off
Export Pick-up
Section bay 1
Section bay 2
9Decision variables
Section bay 1
Section bay 2
Import Drop-off bay
Import transfer bay
Export transfer bay
Export Pick-up bay
10Model components
- Optimization model G.A.
- Objective function Simulation model
11Simulation model
- Used to compute the objective function
- Discrete event deterministic simulation
- Simulation based on cycles
- In each cycle, one import and one export
container processed - Entities in simulation Cranes and trolleys
- 5 states for crane and 3 states for the trolleys
- Next event Chosen based on current time (time at
last updation) comparisons of trolleys at state
changes
12Simulation example
Event Pick trolley 1 State change 0 gt 1
Trolley 1 Current state 0 Current time 400
Trolley 2 Current state 2 Current time 405
13Simulation example
Event Pick trolley 2 State change 2 gt 0
Trolley 1 Current state 1 Current time 412
Trolley 2 Current state 2 Current time 405
14Simulation example
Event Pick trolley 1 State change 1 gt 2
Trolley 1 Current state 1 Current time 412
Trolley 2 Current state 0 Current time 489
15Solution Methodology for the model
Start
Initial solution set
New solution set
Selection
Fitness evaluation through sim. model for each
solution in pop.
Is gen. lt total Gen.
Mutation
Crossover
Stop
16G.A. parameters
- Mutation probability 0.0001
- Crossover probability 0.9
- Pop size 300 500
- Generations 600 800
- Implementation NSGA II
17Sensitivity to no. generations
18Sensitivity to pop. size
19Input to the model
- Crane speed 0.91 ft/s (1.2 m/s)
- Trolley speed 0.366 ft/s (3 m/s)
- Number of bays 15
- Capacity of each bay 265 60
- Number of cranes 3
- Number of trolleys 2
- Number of containers 420 imports 420 exports
- Time taken to unload/load a container 60
seconds - Time taken at the hoist 102 seconds
20Processing time for different trolley speeds
21Processing time for different crane speeds
22Sensitivity of section bays to trolley speed
23Section bay variation with trolleyspeeds
Trolley Speed 1.2
Trolley Speed 2
Trolley Speed 4
24Conclusions
- Takes 13 hrs to process 840 containers
- Processing time is more sensitive to trolley
speeds than crane speeds (24 improvement - 1.2
m/s to 4 m/s for trolley as compared to 5.6
improvement 1.2m/s to 4m/s for crane) - Storage section bays move away from the hoist as
the trolley speed increases
25 The road ahead
LVRT model Future work
Adding one more crane/ trolley
Dynamic variation of Storage sections
Extension to other models
26(No Transcript)
27Additional slides
28(No Transcript)
29Simulation variables
- Trolley
- current time - ct_t(i),
- current state - state_t(i),
- current cycle
- Crane
- current time - ct_c(j)
- current state - state_c(j)
- current distance from the hoist
30Simulation model flowchart
Pick trolley i
Start
-
- no
no -
- yes yes
yes -
- yes no
- no
-
- yes
Is State_t (i) 2
Is State_t (i) 1
Is State_t (i) 0
Update cycle i
Update Ct_c (j) State_c (j) 1,2,0
Update Ct_c (j) State_c (j) 3,4,0
Check for termination
Update Ct_t (i) Set State_t (i) 0
Update Ct_t (i) Set State_t (i) 2
Update Ct_t (i) Set state_t (i) 1
Terminate
Is Ct_t (i) gt Ct _t(1-i)
i 1-i
31Simulation model
Pick trolley i
Current state , current time
New current state, New current time
Is New current time i gt Current time(1-i)
Set i 1-i
32 Genetic Algorithms (G.A.)
33Outline
- Problem Statement
- Motivation for the study
- Problem formulation
- Model components
- Results
- Future work