Please use this identifier to cite or link to this item:
|Title:||Dynamic inventory policies for aerospace service parts supply chain||Authors:||Aghil Rezaei Somarin||Keywords:||DRNTU::Business::Operations management::Inventory control
DRNTU::Business::Operations management::Supply chain management
DRNTU::Engineering::Industrial engineering::Operations research
|Issue Date:||2014||Source:||Aghil Rezaei Somarin. (2014). Dynamic inventory policies for aerospace service parts supply chain. Doctoral thesis, Nanyang Technological University, Singapore.||Abstract:||Service parts, which replace defective parts in aerospace repair facilities, are expensive. On the other hand, on-time fulfillment of demand for these parts is crucial as there are hefty financial penalties for aircraft schedule delays. Therefore, it is crucial to design an inventory and distribution system which minimizes the total cost of inventory investment. Recent changes in commercial aviation, e.g. substantial increase in fuel and labor costs and continuous growth of low fare carriers, have forced airlines to improve the efficiency of their MRO operations. As a result, it is not un-common for MRO operations to be outsourced to overseas service providers and for service providers to have more than one airline customer. All this has resulted in a complex and decentralized service parts logistics system. Research so far has focused mainly on static decision making for service parts networks with a few warehouses. This research deals with generating dynamic inventory policies for larger service parts supply chains modeled as Markov Decision Processes (MDP). Capitalizing on the real-time information of parts in the resupply network, three types of decision are investigated: allocation of service parts to the bases (stock allocation policy), reallocation of service parts among the bases (stock reallocation policy) and emergency resupply of service parts from alternative source options (emergency resupply policy). Each policy is generated based on the optimal relative value function of the respective dynamic program modeled as MDP. It is shown that stock allocation policy can be characterized by a set of switching curves. Stock reallocation policy defines regions of imbalance. When the inventory levels enter these regions, reallocation of a service part is initiated. Emergency resupply policy can be defined by a set of limiting boundaries; when these boundaries are reached, it is optimal to fulfill the demand from alternative sources. To overcome the common problem with MDP, "curse of dimensionality" in particular, heuristic techniques are proposed to approximate the optimal relative value function. The optimal relative value function for a single-base model is derived by solving the respective difference equation systems and is used as the basis for developing these techniques. The relative value function of a single-base model is proposed to be used for stock allocation policy generation. Aggregate queues are developed and utilized in approximating the optimal relative value function for stock reallocation. The optimal relative value function of an inventory system without safety stocks is derived and a difference operator is introduced. Using the proposed difference operator, the optimal relative value function for emergency resupply policy is approximated. For each of the developed inventory policies, computational experiments are conducted to evaluate the optimality gaps and cost reductions. Computational experiments of each policy consist of a general problem set and series of sensitivity analysis. Based on the numerical results, proposed policies perform very close to the optimal policies. The models developed in this research could serve as a foundation to develop decision support systems to improve the efficiency and reduce costs in service parts supply chain management.||URI:||https://hdl.handle.net/10356/61820||DOI:||10.32657/10356/61820||Fulltext Permission:||open||Fulltext Availability:||With Fulltext|
|Appears in Collections:||MAE Theses|
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.