Design and Simulation of a Feedback Control System for a Steel Plate Storage

摘要

A discrete event heuristic feedback control system fo r a steel plate storage at Odense Steel Shipyard Ltd is developed and implemente d in a computer-based simulation model. The plant is subject to stochastic dist urbances. The control system is able to handle this stochastic behaviour bec ause of the feedback design. The present simulation results indicate that a bene fit in the range of 30%～40% is reachable by modifying the plant. Plant description The steel plate storage is located at Odense Steel Shipyard Ltd, Denmark. The st eel plate storage contains steel plates used to build ship sections. The size of the storage is 600x35 meters. It consists of 32x8 stacks of plates. Plates are delivered and removed dynamically. The plate storage contains 5.000 steel plates in average. Over time, new plates delivered to the storage replace plates removed from the storage. The maximum si ze of a plate is 17x4 meters with a weight of up to 22 tons. The plates are uniq ue and dedicated to a specific purpose but in average the 5 000 plates consist o f 2 500 types. Plates are transported around the plate storage by two portal cranes. The cranes drive on a common rail and hence they cannot overtake each other. The cranes us e magnetic force to lift the plates. The crane can deliver the lifted plate to a conveyor belt or to another stack depending on whether the plate has been order ed for the currentday or later. The plant is subject to different kinds of disturbances. The processes performed at the plant are stochastic. Furthermore, the due dates of the plates are frequ ently changed and hence the effect of previously sorting might be lost. The control task is to dispatch movements of plates to the two cranes. The overa ll goal is to minimize the costs at the plate storage and hence the two cranes a re used for executing the movements as fast as possible without colliding and re spect other physical constraints. Problem solution Different approaches have been investigated to solve the problem. A discrete eve nt heuristic feedback controller is used as an online algorithm. The controller is combined with a local search heuristic for improving the initial plan deliver ed by the controller. The improved plan is used as a reference for another feedb ack controller. To indicate the size of the problem, consider an example where the online algori thm is used for executing 600 movements for one day. Through the day the algorit hm finds the locally optimal movement between 20 executable movements. Then the number of different solutions is 20 600! Notice that the possibility of cho osing another destination for the plates to be moved is not taken into account! Results Without time disturbances the controller using an optimised plan as a reference supply the best performance. Nevertheless, when time disturbances are introduces the discrete event heuristic feedback controller performs as well. The comparis on depends on the size of the disturbances. When the disturbances reach a certai n level the discrete event heuristic feedback controller supply the best perform ance and robustness. If the performance is compared with the working system of today, it seems realis tic that a 30%～40% reduction in time is reachable.