Title:
Geometric feature extraction in support of the single digital thread approach to detailed design
Geometric feature extraction in support of the single digital thread approach to detailed design
Author(s)
Gharbi, Aroua
Advisor(s)
Mavris, Dimitri N.
Chau, Duen Horng
Chau, Duen Horng
Editor(s)
Collections
Supplementary to
Permanent Link
Abstract
Aircraft design is a multi-disciplinary and complicated process that takes a long time and requires a large number of trade-offs between customer requirements, various types
of constraints and market competition. Particularly detailed design is the phase that takes most of the time due to the high number of iterations between the component design and
the structural analysis that need to be run before reaching an optimal design.
In this thesis, an innovative approach for detailed design is suggested. It promotes a collaborative framework in which knowledge from the small scale level of components is shared and transferred to the subsystems and systems level leading to more robust and real time decisions that speed up the design time. This approach is called the Single Digital Thread Approach to Detailed Design or shortly STAnDD. The implementation of this approach is laid over a bottom-up plan, starting from the component level up to the aircraft level. In the component level and from a detailed design perspective, three major operations
need to be executed in order to deploy the Single Digital Thread approach. The first one is the automatic geometric extraction of component features from a solid with no design
history, the second phase is building an optimizer around the design and analysis iterations and the third one is the automatic update of the solid. This thesis suggests a methodology
to implement the first phase.
Extracting geometric features automatically from a solid with no history(also called dumb solid) is not an easy process especially in aircraft industry where most of the components have very complex shapes. Innovative techniques from Machine Learning were used allowing a consistent and robust extraction of the data.
Sponsor
Date Issued
2016-12-08
Extent
Resource Type
Text
Resource Subtype
Thesis