Optimal replacement of continuously degrading systems in partially observed environments

Motivated by wind energy applications, we consider the problem of optimally replacing a stochastically degrading component that resides and operates in a partially observable environment. The component's rate of degradation is modulated by the stochastic environment process, and the component fails when it is accumulated degradation first reaches a fixed threshold. Assuming periodic inspection of the component, the objective is to minimize the long‐run average cost per unit time of performing preventive and reactive replacements for two distinct cases. The first case examines instantaneous replacements and fixed costs, while the second considers time‐consuming replacements and revenue losses accrued during periods of unavailability. Formulated and solved are mixed state space, partially observable Markov decision process models, both of which reveal the optimality of environment‐dependent threshold policies with respect to the component's cumulative degradation level. Additionally, it is shown that for each degradation value, a threshold policy with respect to the environment belief state is optimal if the environment alternates between two states. The threshold policies are illustrated by way of numerical examples using both synthetic and real wind turbine data. © 2015 Wiley Periodicals, Inc. Naval Research Logistics 62: 395–415, 2015     

Effects of cueing in cooperative search

We consider the effects of cueing in a cooperative search mission that involves several autonomous agents. Two scenarios are discussed: one in which the search is conducted by a number of identical search‐and‐engage vehicles and one where these vehicles are assisted by a search‐only (reconnaissance) asset. The cooperation between the autonomous agents is facilitated via cueing, i.e., the information transmitted to the agents by a searcher that has just detected a target. The effect of cueing on the target detection probability is derived from first principles using a Markov chain analysis. In particular, it is demonstrated that the benefit of cueing on the system's effectiveness is bounded. © 2006 Wiley Periodicals, Inc. Naval Research Logistics, 2006     

The impact of manufacturer rebates on supply chain profits

Manufacturer rebates are commonly used as price discount tools for attracting end customers. In this study, we consider a two‐stage supply chain with a manufacturer and a retailer, where a single seasonal product faces uncertain and price‐sensitive demand. We characterize the impact of a manufacturer rebate on the expected profits of both the manufacturer and the retailer. We show that unless all of the customers claim the rebate, the rebate always benefits the manufacturer. Our results thus imply that “mail‐in rebates,” where some customers end up not claiming the rebate, particularly when the size of the rebate is relatively small, always benefit the manufacturer. On the other hand, an “instant rebate,” such as the one offered in the automotive industry where every customer redeems the rebate on the spot when he/she purchases a car, does not necessarily benefit the manufacturer. © 2007 Wiley Periodicals, Inc. Naval Research Logistics, 2007     

Quadratic programming with quadratic constraints

A program with a quadratic objective function and quadratic constraints is considered. Two duals to such programs are provided, and an algorithm is presented based upon approximations to the duals. The algorithm consists of a sequence of linear programs and programs involving the optimization of a quadratic function either unconstrained or constrained to the nonnegative orthant. An example involving production planning is presented.     

Information in two-stage programming under uncertainty

The random variables in two-stage programming under uncertainty are generally treated in a passive manner in that no information regarding the random variables or the process generating the random variables may be obtained. This paper develops the economics of information for the case in which the probability distributions are discrete. A multinomial process is assumed to generate the random variables, and the parameter vector of that process is assumed to be unknown. A Dirichlet prior distribution on the parameter vector is used, and the computation of the value of information thus involves a Dirichlet-multinomial distribution on the random variables. An example involving producing to meet uncertain demands is presented.