Using tabu search to determine the number of kanbans and lotsizes in a generic kanban system

A generic kanban system designed for non-repetitive manufacturing environments is described. The purpose of this paper is to determine the number of kanbans and lotsizes to maximize system performance. System objectives include minimizing cycle time, operation costs and capital losses. A scalar multi-attribute utility function is constructed and a tabu search algorithm is proposed to search for the optimal utility value. Simulation is used to generate objective function values for each system setup. Four different variations of tabu search are employed. It is shown that a random sampling of the neighborhood provides good results with the shortest computation time. The tabu search algorithm proposed performs much better than a local search. The results are then compared to those from a modified simulated annealing algorithm. Due to the planar nature of the objective function, it is shown that tabu search can provide excellent results, yet a simulated annealing approach provides the same results with better computation time.