A methodology for estimating expected usage of repair parts with application to parts with no usage history

Abstract

In this paper a model is presented which focuses on the difficult problem of predicting demands for items with extremely low usage rates. These form the bulk of repair parts in military systems. The basic notion underlying the model is the pooling of usage data for common design items with movement for the purpose of estimating usage rates for similar items which have shown no movement.A unique feature of the model is that it also makes possible the estimation of usage rates for items newly introduced into a system for which no previous usage history is available. INTRODUCTIONThe problem of predicting demands for individual repair parts in military inventory systems has received much attention over the last two decades. This problem is a complicated one because of the sporadic nature of demands for military repair parts. For most repair parts, no demands are registered over long periods of time and when items are demanded, they are generally demanded only once or twice. This fact has now been documented by many usage studies t and is once again documented in this study. To illustrate the nature of the demand problem under consideration, usage data for 61 submarine patrols are shown in Table 1. As may be seen from the first entry in the table, no usage 535 536S. E. HABER AND R. SITGREAVES was recorded for the vast majority of items, i.e., the vast majority of items was not demanded in any one of the 61 patrols. Of those that were demanded, one-half were demanded in exactly one patrol.Thus for most repair parts with usage, the problem of estimating usage rates is a difficult one. This difficulty is compounded by an order of magnitude for the bulk of the items whose usage history shows zero units used.In this paper we will be concerned with the estimation of expected usage of repair parts for the purpose of computing shipboard allowance lists. A shipboard allowance list is defined as the range and depth of repair parts to be stocked aboard ship to meet uncertain demands. The range of repair parts refers to the number of different items to be stocked. The depth refers to the number of units stocked of an item.Given that repair part usage is sporadic, several demand prediction strategies are available. The most widely practiced approach is that of employing usage estimates provided by technicians, i.e., supply personnel responsible for provisioning of repair parts. In practice, these have been found to be conservative and lead to relatively expensive stockage lists. Such conservatism, however, is preferred to a much more extreme approach which might assign a zero usage estimate to a repair part until positive usage is experienced. The difficulty with this latter approach is that many repair parts are only one-time movers. Failure to have an adequate quantity of stock aboard ship or in the supply system prior to the first demand can thus lead to a large range of shortages and an unsatisfactory level of readiness.Another approach that has been utilized to estimate usage of slow moving items is exponential smoot...

A unified model for demand prediction in the context of provisining and replenishment

“…From the figures in Table 1, the important finding to be noted is that the adoption of the approach developed in [5] to the estimation of component unreliability factors preserves the relationship between component unreliability and repair part usage found in [6].…”

“…This is done to acquaint the reader with the structure of these models and to facilitate the transition to the unified model. In the first model [5], classes of repair parts are established on the basis of a specified criterion, for example, nomenclature. For a given part I in the class C , the total quantity of units demanded over T time periods, say, yl, is assumed to be a random variable with a Poisson distribution whose mean is BIT.…”

“…The major innovation of the integrated model is that it applies the methodology of [5] to the problem of estimating C,. Attention is now directed to this new model which makes possible the estimation of positive values of C , in (9) for components that have not malfunctioned, or have not been subject to wear because they have not yet been introduced into the logistics system.…”

“…The second, Model IA, is a modified version of Model I; discussion of Model IA is temporarily deferred. The third, called the Pooled Repair Part Model, is described in [5]; the particular version applied here is Model IIA (see [5]) in which usage data is pooled by individual repair part class. As previously indicated, unlike the unified model, in the Pooled Repair Part Model no attempt is made to utilize component data to modify individual repair part usage rates.…”

“…From these examples, it is seen that when demand is sporadic and low, the requirement for demand prediction techniques with desirable properties becomes doubly urgent. An approach incorporating the desirable features of two earlier demand prediction models [5] and [61 is the subject of this paper.…”