Three inventory models for non-traditional supply chains
This work considers three different non-traditional supply chain structures with similar demand and replenishment parameters, and similar solution techniques. In the first article, we develop an inventory model that addresses inventory rationing based on customer priority. We use the framework of a multi-echelon inventory system to describe the physics of a critical level policy. To extend from previous research, we allow multiple demand classes while minimizing a cost objective. We assume a continuous-review, base stock replenishment policy and allow for full backordering. Simulation is used to estimate total expected cost, applying variance reduction to reduce sampling error. First differences are estimated using a Perturbation Analysis unique to inventory rationing literature, heuristics are used to minimize costs. In the second article, we consider a stockless hospital supply chain with inaccurate inventory records. The model presented here is conditional on the level of accuracy in a particular hospital department, or point-of-use (POU). Similar to previous research on inventory inaccuracy, we consider both actual net inventory and recorded inventory in deriving the performance measures. The resultant model is a periodic-review, cost minimization inventory model with full backordering that is centered at the POU. Similar to the previous article, we assume a base stock ordering policy, but in addition to choosing the optimal order-up-to level, we seek the optimal frequency of inventory counts to reconcile inaccurate records. We present both a service level model and a shortage cost model under this framework. In the final article, we consider a hybrid hospital supply chain with both regular and emergency ordering when inventory records are inaccurate. The resultant model is an extension from the previous article where there are opportunities for both regular replenishments and emergency replenishments. We seek an optimal solution to an approximate cost model, and then we compare the results to a simulation-optimization approach.