Dynamical multi-parameter sizing of DDMRP buffers in finite capacity flow-shops

Abstract

The DDMRP (Demand Driven Material Requirements Planning) methodology uses buffer stocks to (i) maintain a high level of service, (ii) stop the spread of uncertainty and (iii) adapt to market changes. According to theory, the size of these buffer stocks should be defined regularly. This sizing involves several parameters and policies to update them, but very little information is available on this subject. We aim to help practitioners choose sizing policies, while maximizing the performance of a given workshop. We have developed an experimental design to compare many combinations of flow-shops and bottleneck constraints, taken from industrial use-cases, and using discrete event simulation. The results show that (i) different degrees of dynamism are needed depending on the performance metric chosen by practitioners, (ii) completely dynamic control does not systematically lead to better performance, and (iii) contrary to what the existing literature on DDMRP suggests, varying buffer sizes may be less effective than fixed ones for an important part of use-cases.