A neural network procedure for kanban allocation in JIT production control systems

In this paper, we develop a Generalized Systematic Procedure (GSP) for determining the optimum kanban allocation in Just-In-Time (JIT) controlled production lines. This procedure is based on a meta-model that incorporates (1) a factorial design approach to select the appropriate kanban combinations, (2) a simulation model to simulate the JIT production line, and (3) a trained neural network model to evaluate the line performance over the entire domain of possible kanban combinations. The GSP is then applied to a case problem and the results are presented.