A new adaptive algorithm for determination of stocks in spare parts inventory systems

This paper treats the problem of the stock levels of spare parts required for maintenance of end products in the presence of uncertain demand for spares. Considerations are restricted to spare parts characterized by low demand and high unit price. Demand for spares is assumed to be insufficiently known, i.e. the uncontrolled input to the spare parts inventory system is stochastic in nature. It is supposed that only the type of a priori distribution of input demand is known, whereas distribution parameters are unknown. Reliability analysis may yield only some initial estimates of input distribution parameters. The following task is stated: develop an applicable learning mechanism allowing a priori assumption about uncertain demand to be corrected over time by realizations of demand, and use the improved knowledge about demand to determine the subsequent stocks of spare parts. Since reorders of spares in the system under consideration are corrected over time on the bases of the present improved knowledge about the uncertain demand, they have been termed ‘adaptive reorders’. These reorders should be superior in quality to ones obtained by classical optimization procedures with fixed a priori estimates of demand characteristics.