The effect of engineering changes and demand uncertainty on MRP lot sizing: a case study

Abstract

Much has been said on the various aspects of lot sizing for single-level and multi-level (assembly) systems in MRP. Numerous heuristics have been developed and tested on (rather small) problems with finite horizon and deterministic, time-varying demand.

However, in practice, the hypothesis of a finite horizon and deterministic demand contains several inherent deficiencies. Indeed, the finite horizon assumption ignores that decisions are usually made on a rolling schedule basis. Moreover, lot-sizing decisions are based on uncertain information about future demand. Timing and quantity of future requirements are, at least partially, based on forecasts and hence imply forecast errors and the need for rescheduling. Finally, engineering changes can also lead to very high scrap costs and hence should be incorporated into the lot-sizing decisions.

In this paper the influence of these factors on lot-sizing decisions is examined for a real life problem in a company that produces electronic components to he used in high technology telecommunication systems.

Demand for these components has the following characteristics:

highly time varying due to the occurrence of several large orders for the entire system, as well as numerous smaller ones. The production of all components (several hundreds) has to be smoothed over time in order to account for the available capacity.

frequent rescheduling of orders, which results in very nervous demand for components (changes in timing as well as production quantities).

all subassemblies (printed circuit boards) have a life cycle of at most two to three years, and most of the components are then replaced.

The cost-effectiveness of several single-level lot-sizing techniques under these conditions was investigated using a simulation package, and the influence of each of the above mentioned factors was evaluated.