Analysis of assembly systems with service level constraints

In this paper we propose a decomposition method for general assembly systems, that decomposes the assembly network into purely divergent multi-echelon systems. The key idea behind the method is to pre-allocate common components to (sets of) end-products. The decomposition method is inspired by Rosling's proof of the equivalence of pure assembly systems and serial multi-echelon systems. For the divergent multi-echelon systems it is possible to calculate near-optimal order-up-to-levels satisfying a fill rate constraint. These order-up-to-levels are then used in the original assembly system. We restrict detailed analysis to assembly systems that decompose into series inventory systems. Using discrete event simulation we compare our pre-allocation policies with several commonly used allocation policies. These allocation policies distribute the common components at the latest possible moment among orders of the subassemblies and end-products that use them. The simulation experiments show that the pre-allocation policies perform quite well compared to the allocation policies, both with respect to costs and service.