Thesis
Multi-fidelity methods for multi-objective optimal control applied to aerospace systems
- Creator
- Rights statement
- Awarding institution
- University of Strathclyde
- Date of award
- 2026
- Thesis identifier
- T17632
- Person Identifier (Local)
- 201953140
- Qualification Level
- Qualification Name
- Department, School or Faculty
- Abstract
- Continuous, super-linear increase in the overall complexity of modern Multidisciplinary Design Optimisation problems inexorably tempers the benefits of advancing computational power. The pursuit of computational efficiency therefore remains of crucial importance, including within all modelling and analysis algorithms employed throughout. This work presents two methodologies for reducing the computational burden associated with solving complex Multidisciplinary Design Optimisation problems, with a particular focus on Multi-Objective Optimal Control Problems. The first involves adaptively combining separate levels of modelling fidelity into one single approximation. This approach is particularly suited to cases where one or more subsystems can be represented using alternative numerical models/datasets of competing accuracy and/or computational cost. The second involves systematically upgrading an existing level of modelling fidelity defined by the mathematical formulation of the problem itself. This approach is particularly suited to problems where solution fidelity is more dependant on the number of variables (both static and dynamic) used to describe the system/subsystem(s) of interest. Each methodology is incorporated into separate numerical frameworks intended for application in scenarios where in-depth and exclusively high-fidelity models/analyses are not viable. Several isolated numerical tests are examined in each case to systematically validate the fidelity management steps employed within each framework. Finally, each framework is demonstrated in application to a complex engineering design scenario. It is found that the proposed approaches offer viable and efficient methods with which to adaptively upgrade modelling fidelity for complex Multidisciplinary Design Optimisation problems. These methods can be formulated in a general manner, without relying on extensive a priori information, particularly suiting their application to early-stage preliminary design problems. With this class of approach, the design space of complex systems can be more thoroughly explored in terms of dimensionality, modelling fidelity and multi-objective performance criteria, at earlier stages in the design process. This helps to quickly identify promising search directions for future development whilst reducing the risk of significant later stage design alterations due to misleading or poorly-modelled physical behaviour/inter-disciplinary interactions.
- Advisor / supervisor
- Maddock, Christie
- Resource Type
- DOI
- Funder
Relations
Items
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
PDF of thesis T17632 | 2026-04-15 | Public | Download |