Evanthia Kallou
(Advisor: Prof. Dimitri Mavris)

will propose a doctoral thesis entitled,

A Set-Based Methodology for Aircraft Design Space Exploration and Enrichment with Higher Fidelity Data

On

Monday, July 26 at 2 p.m.
https://bluejeans.com/207710340/7755

Abstract
Due to the lack of validated models for novel aircraft concepts, design decisions are made using uncertain information. Selecting a point design prematurely results in redesign cycles later in the design process when higher fidelity data is available. These redesigns are thus very computationally expensive and time-consuming. However, keeping design options open longer results in an impractical dimensionality increase due to the necessary analysis cases to be run. From the above, the need emerges to reduce the computational cost and make robust design decisions, as the design process progresses.

The challenges of performing Design Space Exploration (DSE) differ as the analysis of more detailed subsystems becomes available. During the conceptual design phase, the aircraft (system) analysis presents a lower computational cost for each analysis run and thus, more design cases can be run. The dimensionality of the input design space is also more manageable since initial estimates of system-level metrics are available at this phase.

As the design progresses to preliminary, the decomposition of the aircraft to its subsystems requires more detailed metrics. This, in turn, increases the dimensionality of the input design space for each subsystem and the computational budget for each analysis run. At this point, dimensionality may become prohibitive to perform optimization, let alone DSE.

Additionally, the communication of a single point design between the aircraft and its subsystems faces organizational barriers and causes costly reiterations. Throughout these iterations, the uncertainty increases if the disciplinary analysis model of appropriate fidelity is not selected. When higher fidelity models are introduced, time dependencies and field responses come into play, rather than scalars. The increase in complexity of the response, in combination with the dimensionality increase, and the point design communication, may render DSE impractical during preliminary design.

This research proposes a novel multi-level Set-Based Design Space Exploration method that uses classification and supervised dimensionality reduction methods to define and communicate design sets between the disciplines in a decomposition level and between the system and the subsystem levels. This approach also advocates for the consideration of the physics assumptions that need to be captured for the discipline at each decomposition level.

Committee

• Prof. Dimitri N. Mavris – School of Aerospace Engineering (advisor)
• Prof. Graeme J. Kennedy– School of Aerospace Engineering
• Prof. Daniel P. Schrage – School of Aerospace Engineering
• Dr. Burak Bagdatli – School of Aerospace Engineering