Computer aided design is vital in modern engineering design. It is used across essentially every discipline to accelerate a project through its detailed design phase. This report explores the application of automated computer aided design to generate novel designs in simple mechanical systems, with the goal of producing functional innovation, proposes and tests a methodology to achieve such a goal given sufficient information.
It is suggested that an application of a genetic algorithm to the parameterisation of a design problem along with the abstraction/realisation method might produce positive results, as similar methods have been shown to be successful individually in prior work. The genetic algorithm requires a chromosome and a fitness function to operate, the first being derived from the parameterisation of a design concept while the second can be more elusive.
The search for a fitness suitable fitness function for applications with changing functional components is one of the key goals of this thesis. The abstraction/realisation process is used as a means to determine suitable replacements for functional artefacts in mechanical systems by comparing their inputs, outputs and characteristic equations.
Implemented in a python computer model, the methodology was tested on the idea of a bicycle frame. It was found that given an initial design similar to very early bicycle frames, the model was able to output a family of solutions which are of similar form to those that are found in modern bicycles.