| In this dissertation, we investigate in depth the probabilistic behavior of warranty cost of repairable complex systems from manufacturers' point of view by considering a comprehensive set of warranty cost factors, such as characteristics of warranty policies, system structure, product failure mechanism, warranty service cost, impact of warranty service, value of time, warranty service time, and warranty claim related factors. Three types of existing warranties including free repair warranty (FRPW), free replacement warranty (FRW) and pro-rata warranty (PRW) are studied. In addition, we propose two novel warranties that could be attractive to both buyers and sellers: renewable full service warranty (RFSW) and repair-limit risk-free warranty (RLRFW).; A traditional way to model warranty cost for warranted multi-component systems is the black-box approach that does not utilize the information of system structure. By studying the proposed RFSW policies with explicit consideration of four types of systems: series, parallel, series-parallel and parallel-series, we demonstrate the importance system structure information.; Some researchers have studied the discounted warranty cost (DWC), a measure incorporating the value of time, for black-box systems under various warranties. We extend their results by deriving the moments of DWC for minimally repaired series systems with an FRPW or a PRW while considering both the continuous and discrete time discounting. The optimal warranty reserve level and the optimal warranty duration are obtained for lot sales.; The majority of researchers studying warranties of repairable products assume perfect repair, an assumption that can hardly hold in practice. This leads us to investigate the problem of warranty cost determination for RLRFW policies considering minimal repair or imperfect repair. We derive some useful results for truncated quasi-renewal processes and truncated non-homogeneous Poisson processes, based on which the exact expressions of the expectation and variance of warranty cost are obtained.; Warranty managers are often puzzled by the difficult decision of choosing the best warranty with appropriate policy parameters among many possible policies for a certain type of products. Motivated by this question, we develop and analyze a general non-linear optimization model considering random demand, which maximizes the expected utility of profit under some realistic constraints. |
