Abstract
dc:descriptionDue to the stochasticity not only of real demand, but also supply processes, inventory management deals with uncertain environments. Supply uncertainty is a diachronic and perhaps unavoidable phenomenon. It can be distinguished into two main categories: lead-time and yield uncertainty. The former concerns situations in which the exact delivery time is a random variable, whereas the latter is met when the batch received may differ from the order placed. Yield uncertainty can be further distinguished in four types: (1) random yield, in which the amount delivered does not equal to the quantity ordered; (2) random capacity, in which the supplier provides either the quantity ordered (if available) or the maximum available stock; (3) random quality, in which a batch received may contain imperfect items; (4) supply disruption, in which a placed order may fail to materialize. Some of the causes leading to supply uncertainty include complex production runs, machine breakdowns, inadequate raw materials, damage in transit, extreme adverse weather events, strikes, pandemics, or even just an unreliable supplier. Consequently, the firm is exposed to inefficient inventory management and so unplanned shortages, hence reduction of customer service, firm reputation and profit. In this thesis, consideration is given to systems facing either quality uncertainty or supply disruptions. More specifically, in the context of quality uncertainty, we study a continuous review inventory system in which each batch received is subject to a time-consuming screening process in order to detect any imperfect quality item. Concerning the supply disruptions, we study two periodic review inventory systems; one examines the impact of the disruption profile (combination of time until disruption occurrence and recovery), and one evaluates the benefits of advance delivery information sharing. In the existing literature, several mitigation and contingency strategies have been investigated to deal with quality uncertainty and supply disruptions, with the most commonly used being the safety stock. Although the safety stock reduces the unplanned shortages and improves the overall system performance, it is not always efficient. A higher one results in holding cost increase, whereas a lower one may not protect the system against unplanned shortages. In order to determine the optimal safety stock, the decision maker should take advantage of any inherent knowledge or available information about the outcome of the supply process. Regarding the quality uncertainty, the main hitherto action against unplanned shortages is to use a safety stock equal to the total demand during the screening process. The proposed safety stock is independent of the shortage cost, a result that is intuitively unappealing. Hence, in this dissertation, adopting the reorder level - order quantity (r,Q) policy, we appropriately formulate the shortage cost and determine the optimal safety stock that minimize the long-run average cost per unit time. Furthermore, we propose a new safety stock that effectively prevents any unplanned shortage and results in significantly lower costs compared to the one suggested in the literature. Yet, there is still a cost increase compared to systems with quality reliability, which becomes more intense as the proportion of imperfect quality items increases. As for supply disruptions, although the disruption profile has been considered in continuous review inventory systems, it is ignored in the periodic review. Instead, only the limiting disruption probabilities are taken into account. However, two systems may have the same limiting disruption probabilities but completely different disruption profiles (ranging from frequent-short to rare-long). Hence, in the existing literature, such systems are treated in the same way. By obtaining full insight of the supply process and operating under the order-up-to level - reorder interval (S,T) policy, we build a mathematical formulation that effectively distinguishes the systems with the same limiting disruption probabilities but different profiles. We determine the policy variables that minimize the long-run average cost per unit time and, also, propose an approximated solution for the reorder interval. Theoretical and numerical results demonstrate that systems facing rare but long disruptions are associated with inventory cycles of higher variance, although the expected length is kept constant, need higher safety stocks, and are more costly than systems facing frequent but short. This result indicates that a different strategy (such as the use of a backup supplier) may be more efficient against rare but long disruptions. As for the performance of the approximated solution, it performs quite well, opposite to the solution that ignores the disruption profile. Besides, we study a periodic review inventory system under endogenous disruptions (described by Bernoulli as a stationary process) in which the supplier provides the buyer advance delivery information. Two extreme regimes are considered: full delivery information (FDI) where the timing of the next delivery is always known, and partial delivery information (PDI) where the timing of the next delivery is only known if it coincides with the next planned delivery. For both regimes, a natural adaptation of the base stock policy is used, having multiple order-up-to levels, each corresponding to a particular level of information available when ordering. For both models, we determine the optimal policy variables that minimize the long-run average cost per unit time. The solution is obtained in closed-form, directly analogous to the classical EOQ. Numerical results show that advance information directly reduces operating costs, with the magnitude of the reduction being positively associated with the amount of information. However, since even under full information the costs remain high, the decision makers should always seek reliable suppliers aiming at ensuring disruption-free operation.
Degree
thesis:*- Grantor dc:publisher
- University of Ioannina
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Karakatsoulis, Georgios
- Καρακατσούλης, Γεώργιος
Subjects
dc:subject × 10Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/54033