Allocating Variable Resources Over A Finite Time Horizon to Combine Aircraft Sizing and Airline Planning

This paper considers the problem of simultaneous aircraft-design and airline-aircraftallocation. In this problem formulation, an airliner allocates "variable" resources and an aircraft manufacturer develops a new aircraft based on a set of "variable" design speciflcations. This is a multidisciplinary problem in which an optimization framework flnds an optimum decision for both allocation and design domains. The primary objective of the work presented on this paper is to explore and expose issues when the simultaneous problem considers variable-resource allocation over a flnite-time horizon. The quoted term "variable" indicates that for the purpose of ∞eet planning, an airliner allocates a number of "nonexisting" aircrafts in addition to the existing ∞eet in order to satisfy passenger market demands which evolve over time. The "nonexisting" aircraft constitutes a new aircraft concurrently being sized and developed by an aircraft manufacturer. A set of speciflcations of the new aircraft is expected to optimize the objective of ∞eet planning over a flnite horizon. As a solution approach to the two-domain multidisciplinary problem: aircraft-design and resource-allocation, the authors use an integrated optimization framework consisting of dynamic programming (DP), nonlinear programming (NLP) and linear integer programming (LIP). For numerical explorations, this work considers a small scale allocation problem similar to the authors’ previous work on static problem. 1 However, this work extends the problem by introducing dynamics to the original static problem. When dynamics of passenger demands, monetary ∞uctuations, maintenance activities change over time, a solution to static problem no longer holds for simultaneous problem over a flnite-horizon in most cases. As an interesting example, in the work on static problem, 1 one of the conclusions from the numerical explorations is to retire one aircraft. However, a solution presented on this paper shows the contrary even though the averaged problem parameters of the dynamic problem are similar to those of the static one. Exploration on this simultaneous problem reveals that in general a large number of computation is necessary to obtain a solution. Only for some cases, where inter-time dependency is weak or predictable, original problem over a flnite-horizon can be decomposed and solved for each individual time.