Adapting material requirements planning systems to effectively handle uncertain yield rates



Journal Title

Journal ISSN

Volume Title



Material Requirements Planning (MRP) Systems implicitly make a number of assumptions about the environment in which they operate. One significant result of these assumptions is the elimination of uncertainty. Much research has been done where this uncertainty originates from outside the system such as demand or supply variances. These assumptions have been handled through a number of buffering techniques. However, very little work has been done to handle uncertainty which originates internally in the production system. For example, variances in yield rates. This largely uninvestigated aspect of internal uncertainty arises when the output quantity from a production process is uncertain. The typical assumption made is that the output has some deterministic relationship to the input quantity so that appropriate fixed yield rates are incorporated in the bills of material. However, actual yield rates may vary for a variety of reasons: operator error, tool wear, material flaws, etc. Clearly an MRP system must compensate for this internal uncertainty if it is to be effective as a planning and control system. Previous research in this area has largely focused on the single period, single process optimal reject allowance problem. Research in the more complex manufacturing environment such as faced by MRP systems has not provided any detailed discussion of how the problem is structured for analysis, nor has a full range of compensation techniques been investigated. The principal focus of this dissertation is to fill this significant void in the literature. Three compensation techniques studied are: safety stocks, inflated lot sizes, and rescheduling. These techniques could be implemented using either an overall approach which concentrates all compensation activity at the end item level or a level by level approach which spreads the activity throughout the product structure. The effectiveness of these techniques and approaches has been measured through the achieved service level, number and duration of stockouts, and inventory levels. The policies have been investigated under a number of experimental conditions which included: demand variation (high or low), part commonality (high or low), lead time duration (short or long), and order policy (lot-for-lot or period order). The findings have been that demand variability does influence the choice of a compensation policy. The lot inflation policy was the preferred one when the degree of demand variability was greater. We also found that higher levels of component part commonality were more costly but that this did not favor any one policy. The lot sizing policy used was found to be significant since those policies which consolidate several period's demands provide an inherent buffer stock and obviate the need for any further compensation. Our investigation of the influence of lead time duration found that these were not significant unless the lot sizing policy used was insufficient to cover the lead time demand. [...]



Materials management--Mathematical models, Production control--Mathematical models