This research presents a framework system dynamics (simulation) model that evaluates the effect of maintenance policies on system performance and life-cycle cost. The model highlights factors such as learning, aging and the technological upgrades that occur during the life-cycle of a system. The metrics used to measure the effectiveness of maintenance policies are the system life-cycle cost and cumulative breakdowns. In this research, a varying maintenance policy has been modeled using system dynamics methodology to determine the future performance of the system that is dependent upon its past performance when breakdowns occur randomly. The main objective of this modeling approach is to balance the cost of preventive maintenance actions with the opportunity losses due to system breakdowns. The approach used in this research primarily involves forecasting future breakdowns using an average of accumulated opportunity losses. This research effort was mainly aimed at developing a (framework) model to determine effective maintenance policy for a system and evaluating the effect on the life-cycle cost for various scenarios. This model could further form the basis of a decision support system for maintenance modeling. / Master of Science

Maintenance models are critical for evaluation of the alternative maintenance policies for modern engineering systems. A poorly selected policy can result in excessive life-cycle costs as well as unnecessary risks for catastrophic failures of the system. Economic dependence refers to the difference between the cost of combining the maintenance of a number of components and the cost of performing the same maintenance actions individually. Maintenance that takes advantage of this difference is often called opportunistic. Large number of components and economic inter-dependence are two pervasive characteristics of modern engineering systems that make the modeling of their maintenance processes particularly challenging. Simulation is able to handle both of these characteristics computationally, but the complexity, especially from the model verification perspective, becomes overwhelming as the number of components increases. This research introduces a new procedure for maintenance models of multi-unit repairable systems with economic dependence among its components and under opportunistic maintenance policies. The procedure is based on the stochastic Petri net with aging tokens modeling framework and it makes use of a component-level model approach to overcome the state explosion of the model combined with a novel order-reduction scheme that effectively combines the impact of other components into a single distribution. The justification for the used scheme is provided, the accuracy is assessed, and applications for the systems of realistic complexity are considered.

In many equipment-intensive organizations in the manufacturing, service and particularly the defense sectors, service parts inventories constitute a significant source of tactical and operational costs and consume a significant portion of capital investment. For instance, the Defense Logistics Agency manages about 4 million consumable service parts and provides about 93% of all consumable service parts used by the military services. These items required about US$1.9 billion over the fiscal years 1999-2002. During the same time, the US General Accountability Office discovered that, in the United States Navy, there were about 3.7 billion ship and submarine parts that were not needed. The Federal Aviation Administration says that 26 million aircraft parts are changed each year. In 2002, the holding cost of service parts for the aviation industry was estimated to be US$50 billion. The US Army Institute of Land Warfare reports that, at the beginning of the 2003 fiscal year, prior to Operation Iraqi Freedom the aviation service parts alone was in excess of US$1 billion. This situation makes the management of these items a very critical tactical and strategic issue that is worthy of further study. The key challenge is to maintain high equipment availability with low service cost (e.g., holding, warehousing, transportation, technicians, overhead, etc.). For instance, despite reporting US$10.5 billion in appropriations spent on purchasing service parts in 2000, the United States Air Force (USAF) continues to report shortages of service parts. The USAF estimates that, if the investment on service parts decreases to about US$5.3 billion, weapons systems availability would range from 73 to 100 percent. Thus, better management of service parts inventories should create opportunities for cost savings caused by the efficient management of these inventories. Unfortunately, service parts belong to a class of inventory that continually makes them difficult to manage. Moreover, it can be said that the general function of service parts inventories is to support maintenance actions; therefore, service parts inventory policies are highly related to the resident maintenance policies. However, the interrelationship between service parts inventory management and maintenance policies is often overlooked, both in practice and in the academic literature, when it comes to optimizing maintenance and service parts inventory policies. Hence, there exists a great divide between maintenance and service parts inventory theory and practice. This research investigation specifically considers the aspect of joint maintenance and service part inventory optimization. We decompose the joint maintenance and service part inventory optimization problem into the supplier s problem and the customer s problem. Long-run expected cost functions for each problem that include the most common maintenance cost parameters and service parts inventory cost parameters are presented. Computational experiments are conducted for a single-supplier two-echelon service parts supply chain configuration varying the number of customers in the network. Lateral transshipments (LTs) of service parts between customers are not allowed. For this configuration, we optimize the cost functions using a traditional, or decoupled, approach, where each supply chain entity optimizes its cost individually, and a joint approach, where the cost objectives of both the supplier and customers are optimized simultaneously. We show that the multiple objective optimization approach outperforms the traditional decoupled optimization approach by generating lower system-wide supply chain network costs. The model formulations are extended by relaxing the assumption of no LTs between customers in the supply chain network. Similar to those for the no LTs configuration, the results for the LTs configuration show that the multiobjective optimization outperforms the decoupled optimization in terms of system-wide cost. Hence, it is economically beneficial to jointly consider all parties within the supply network. Further, we compare the model configurations LTs versus no LTs, and we show that using LTs improves the overall savings of the system. It is observed that the improvement is mostly derived from reduced shortage costs since the equipment downtime is reduced due to the proximity of the supply. The models and results of this research have significant practical implications as they can be used to assist decision-makers to determine when and where to pre-position parts inventories to maximize equipment availability. Furthermore, these models can assist in the preparation of the terms of long-term service agreements and maintenance contracts between original equipment manufacturers and their customers (i.e., equipment owners and/or operators), including determining the equitable allocation of all system-wide cost savings under the agreement.