Abstract
Design of composite components is inherently multi-disciplinary problem; the current linear design method is not very well suited for holistic design optimization and frequent iteration of design. A novel methodology for comprehensive design of a composite component is suggested in this thesis. The methodology utilizes Python scripting to create an interconnected system of simulations (SySi), including CAD, basic aerodynamic analysis, meshing, structural FE, kinematic braiding simulation, and resin infusion simulation. An SQL database is used to manage and trace iteration data. The models are self-generated with given iterated variables, to allow for a multi-simulation optimisation and fast component rework.A demonstrator part was developed to demonstrate the suggested methodology. A small UAV braided spar was selected. It is sufficiently simple for the demonstrator development, yet it retains some of the complexities of typical composite component.
All the modules were successfully automated, providing variety of standalone benefits. A Latin hypercube sampling method proved highly successful in initial sampling of the design space. It also supports troubleshooting, as it generates all allowed combinations of parameters. It was also used as an input for surrogate models. Non-surrogate optimisations proved too computationally expensive. With surrogate model this problem was alleviated to some extent. Surrogate model was successfully created with minimal errors present in the validation. However, significant number of datapoints was required to create high quality surrogate. The subsequent optimisation was not particularly useful for the demonstrator part, as the optimised values could be obtained from plotting the sampling data.
The alternative, highly automated design process was successfully demonstrated. Some of the envisioned benefits are clear from the demonstrator, while others require further study.
The demonstrator system with its scripts is available on https://github.com/Ellutze/sysi .
| Date of Award | 21 Jun 2022 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | James Kratz (Supervisor), Thomas C S Rendall (Supervisor) & Amit Visrolia (Supervisor) |