期刊
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH
卷 58, 期 13, 页码 4050-4070出版社
TAYLOR & FRANCIS LTD
DOI: 10.1080/00207543.2019.1641238
关键词
slab pre-marshalling; self-training; hybrid algorithm; dominance rules
资金
- Fund for Innovative Research Groups of the National Natural Science Foundation of China [71621061]
- Major International Joint Research Project of the National Natural Science Foundation of China [71520107004]
- Major Program of National Natural Science Foundation of China [71790614]
- 111 Project [B16009]
We study the slab pre-marshalling problem to re-position slabs in a way that the slabs are stored in the least number of stacks and each stack contains only the slabs of the same group, which can be utilised interchangeably. In this way, when a slab of any group is required, the topmost slab can always be picked up without shuffling. During pre-marshalling, however, at most two slabs can be moved by one operation. In this paper, we present a network model with three valid inequalities to solve this problem. With a small amount of labelled data from the model approach, a self-training technique is applied to train a function for predicting the best next move. Then, a new hybrid algorithm is developed to solve the practical problems by combining the self-training technique, heuristics, and the branch-and-bound algorithm with five dominance rules. The experimental results demonstrate the effectiveness of this network model and valid inequalities, and the performances of different components of this algorithm. The new algorithm produces high-quality solutions within seconds.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据