Systematic bio-inspired design: from concept to validated functionality

Millions of years of evolution have allowed organisms to produce innovative solutions for the same problems designers encounter every day. During bio-inspired design, the interesting working principles that make the biological solution work are extracted and transferred to the design problem. Previous research has shown that this results in more novel designs. Furthermore, since biologists have already been studying organisms for centuries, their publications contain a large catalogue of proven solutions. The recent rise in open-access publishing and more permissive text-mining policies allow for this interesting source of information to be more widely available.

Despite these promising boundary conditions, bio-inspired design has not yet been widely adopted as a design strategy and only a limited number of bio-inspired products are commercially available. During the design process, designers often lack biological background knowledge, turning the search for relevant biological strategies into a frustrating and time-consuming task. Next, the process of analysing, selecting and understanding these strategies is riddled with pitfalls.

This dissertation aims to develop and validate a bio-inspired design support tool that can be directly integrated into the systematic engineering design process. The first contribution is an automatic search method that starting from the functional requirements allows the designer to retrieve the most relevant biological documents. This is done by applying machine translation, and automating and scaling the generation of organism aspects. The second contribution is a comparison between the proposed search method, AskNature and a biologist for retrieving bio-inspiration. Most notably, the search methods managed to retrieve more working principles, but with the caveat that processing biological literature directly comes with a large cognitive burden. The third contribution is the pilot of the first end-to-end validation experiment using bio-inspiration using outcome-based metrics. The main finding is that when designers must search, analyse and select bio-inspiration during a concept generation session, the novelty increase is more limited than reported in previous research. Furthermore, due to the high time cost associated with processing biological literature directly, two different approaches to bio-inspiration were identified: one where the goal is to gain inspiration quickly and which uses pre-processed strategies and another where the goal is to find novel biological strategies, at the cost of having to search through biological publications. The fourth contribution is a general method and dataset for clustering biological strategies by their working principle. Due to the general nature of the proposed method and benchmarking tool, the pipeline can be easily adapted when more performant language models are released. The fifth and final contribution is the proposal of three different metrics to re-rank the generated clusters based on the novelty, robustness or current understanding of a biological strategy.