Technical Report: Optimizing the Slab Yard Planning and Crane Scheduling Problem using a Two-Stage Approach

In this paper, we present The Slab Yard Planning and Crane Scheduling Problem. The problem has its origin in steel production facilities with a large throughput. A slab yard is used as a buffer for slabs that are needed in the upcoming production. Slabs are transported by cranes and the problem considered here, is concerned with the generation of schedules for these.

The problem is decomposed and modeled in two parts, namely a planning problem and a scheduling problem. In the planning problem a set of crane operations is created to take the yard from its current state to a desired goal state. The aim of the planning problem is twofold. A number of compulsory operations are generated, in order to comply with short term planning requirements.

These operations are mostly moves of arriving and leaving slabs in the yard. A number of non-compulsory operations with a long term purpose are also created. A state of the yard may be more or less suited for future operations. It is desirable to keep the yard in a state, where it lends itself well to the future requests.

Partial knowledge of future requests may exist and hence the yard can be prepared for those. In the scheduling problem, an exact schedule for the cranes is generated, where each operation is assigned to a crane and is given a specific time of initiation. For both models, a thorough description of the modeling details is given along with a specification of objective criteria.

Variants of the models are presented as well. Preliminary tests are run on a generic setup with artificially generated data. The test results are very promising. The production delays are reduced significantly in the new solutions compared to the corresponding delays observed in a simulation of manual planning.

The work presented in this paper is focused on a generic setup. In future research, the model and the related methods should be adapted to a practical setting, to prove the value of the proposed model in real-world circumstances.