Early-Stage Ship Design Operational Considerations as a Thin Abstraction Enabled by a Grid-Supported Markov Decision Process Directional Decision Ensemble Framework

摘要

Design always works with a reduction of the problem's complexity. Independent of the design stage, design always involves a reduction in fidelity from the final operational product. This fact is even more prevalent in the design of large marine products. The reduction of the designed vessel's complexity is also known as an abstraction, and designers utilize abstractions during all phases of design. The term abstraction means that designers connect "the world of events that actually occurred or can occur" and "the imagined world of hypothetical descriptions". Currently, researchers have focused on creating thick abstractions through specific frameworks, which can richly describe a certain scenario of the event with as much detail as possible. However, little has been done to enable thin abstractions, which only reserve key factors to ensure condensed but not scenario-specific descriptions of the event. Because of this gap, it becomes challenging to understand the operational performances of a conceptual design with adequate multidisciplinary trade-offs. If suitable key factors exist, designers would then be able to model ship operations at a reduced-order level, consistent with what a conceptual design supports but rich in the implications of how multiple disciplines are synthetically balanced. In the evaluation of ship operations, thick abstractions are the predominant approach being taken. The research presented in this thesis focuses on the creation of a novel thin abstraction of ship operations so that the appropriate key factors of describing sea transport performances in concept design can be obtained. The Grid-Supported Markov Decision Process (GS-MDP) framework has been developed to analyze ship operations as a thin abstraction. The framework blends a newly developed gridding approach, Markov Decision Process (MDP), and frequency-domain seakeeping codes. The GS-MDP framework uniquely identifies directional decisions as the key factor required to execute operational evaluation as a thin abstraction. A directional decision is the determination of whether a direction at a location deserves to be maintained or adjusted with respect to reaching the destination. By setting up MDP based on a novel ocean grid, a vessel can be simulated to make directional decisions for all directions at all locations over the entire ocean under any circumstance. Linking frequency-domain seakeeping codes to MDP ensures the incorporation of physics-based ship motions to the sea transport simulations. Furthermore, aggregating directional decisions solutions across a large simulation space creates thin abstraction operation ensembles. The operation ensemble can provide valuable knowledge for designers to understand a conceptual design. Beyond the novel framework, new decision metrics have been developed that enable design decisions utilizing the thin abstraction. Based on the utilization and statistical analysis of an operation ensemble, these metrics enable the designer to understand the potentials of operational efficiency or operational difficulty. The ability to quantify efficiency or difficulty allows designers to explain the underlying causation associated with the operational potentials. Two case studies are presented in this thesis. The first case study discusses the usefulness of the GS-MDP framework in identifying main contributors and underlying contexts with respect to certain operational outcomes. The second case study expands the application of this framework and maps it onto both transit events and on-site operational events, which illustrates the value of a thin abstraction.