From 2007-2011 I undertook a Ph.D. at the University of Cambridge in the Engineering Design Centre. My supervisor was Professor John Clarkson, and I worked closely with Dr David Wynn (then a member of the EDC). The title of my thesis was “Developing a computational approach to support product architecture design”.
During my Ph.D. I studied the design process for complex products, through interviews with a broad range of engineers at a firm producing internal combustion engines. Based on the insights from the case study I developed a computational method to help engineers develop new concepts or architectures for such artefacts, at an abstract level. The method, which I termed EPA, entails developing a schema for a particular problem consisting of two main aspects:
- An ontology describing the potential constituent parts of an architecture in terms of the types of components that may make it up and the types of relations or connections that may link them.
- A set of constraints that determine whether a certain arrangement of components and connections constitutes a valid architecture.
Based on a schema, it is possible to synthesise solutions to the problem computationally – either starting “from scratch” or based on a pre-existing solution. I was able to show preliminary results that suggested the method was usable by engineers with appropriate training, and useful in practice by generating potentially improved solutions and by helping to formalise otherwise-tacit knowledge. The method is similar in principle to morphological analysis but has greater expressive potential to handle more complex scenarios.
In the course of developing and evaluating the EPA method, I implemented it within the EDC’s Cambridge Advanced Modeller tool for linkage-based modelling as an EPA “toolbox”. CAM provided a robust and usable GUI, aiding construction of schemas represented as graphical models, as well as additional supporting frameworks for evaluating generated architectures. I developed two algorithms for synthesising solutions, one based on state-space search and one by converting an EPA schema to a model in the Alloy language which is then solved by reduction to a binary satisfiability problem. I applied the method to a number of problems myself, and supported others in doing so for the purposes of evaluating the method and identifying shortcomings.
As well as my own research, during my Ph.D. I also contributed extensively to aspects of the CAM framework, and developed a separate tool called “Set Visualiser” (not currently available online) to generate interactive Euler diagrams for visualising categorical data.
My full publications list may be found on ORCID: http://orcid.org/0000-0002-1114-7994