Continuous (s,S) inventory policy with non-stationary stochastic demand -

Abstract

We consider a single-item inventory model with non-stationary stochastic demand. Non-stationary stochastic demand is applicable to a large number of real world supply chain systems. Dynamically changing (st , St ) policies are shown to be optimal in the existing literature, Song and Zipkin (1993). In this thesis, we present relatively a new approach to model the non-stationary stochastic demand and inventory position processes. Our analytical model considers both a general phase-type (Pht ) distribution and a special two-level mixture of Erlangs of common order (2-MECO) Pht distribution to serve as an approximation of the demand process. The approximate Pht distribution allows us to compute the expectation and variance of the demand, inventory position, net inventory and number of orders in function of time. We then propose an optimization heuristic to compute the dynamic time dependent reorder and order up-to levels (st , St ) that minimizes the total expected cost. Finally, we test our findings using numerical examples.