| Summary: | 碩士 === 國立東華大學 === 運籌管理研究所 === 99 === Ocean transportation becomes more important by containerization. Nowadays vessels are built much bigger as throughput of ports increases. Planners need to reduce the berthing time of vessels and improve the operational efficiency of container terminals. Container yard is one of the areas in a container terminal highly affected by the increase in container throughput (Vis and Carlo, 2010). Yard crane is the key equipment in container yard. Good schedules of yard cranes are indispensable to improve the productivity of container terminals.
Our thesis considers two yard cranes retrieving a given collection of export containers in a block. The two yard cranes have to be apart for the safety distance and the retrieval of containers follows the order of the group identifications of containers. The objective is to minimize the completion time for the given group of containers. We modify Liou’s (2008) integer programming model leading to a model with shorter computation time. Because we use different concepts to control movement ranges of the two yard cranes, the present model can handle bigger scale problems with more containers or stacks. In these bigger problems, our model sometimes takes more than 1 hour or even a few days to solve. So we develop a rule-based heuristic which can solve faster and more accurately.
According to the concept of Ng (2005), we separate two zones of equal size for the two yard cranes. This is called Basic Assignment in the thesis. We observe optimal assignments of yard cranes and find factors affect the Basic Assignment. Three factors are identified by us for our two-step heuristic. The heuristic has two steps. The first step decides the assignment for most containers by rules that we build. The second step schedules the two yard cranes and assigns yard cranes for the remaining containers by CPLEX-OPL. These remaining containers are important because they perhaps affect finding a good feasible solution.
Based on the results of numerical runs, the computation time of our model is really faster than Liou’s (2008) model. For the gap defined as the difference between a solution of heuristic to the corresponding optimum solution the average gap of our heuristic is also smaller than Liou’s (2008) heuristic. Because of the heuristic, we can get acceptable feasible solutions quickly. In bigger scale problems more containers or more stacks, the feasible solution of the heuristic is acceptable too. The best gap is about 1.20% and the worst gap is 7.41%. Finally, we check the scale of problems that our heuristic can solve.
|
|---|
