Methods and Algorithms for Systematic Innovation (Methoden en algoritmen voor systematische innovatie)

The vital importance of innovation on enterprises, and by extension, economies is often expressed as innovate or vegetate and innovate or evaporate . However, these expressions provide no insight on how to innovate, or on how to be inventive. The creative spark necessary for this kind of inventive thinking is typically thought of as situated in the realm of the human mind. Computers programs and algorithms can however support creativity by providing stimuli in different forms, e.g. random words, function databases, or previously documented solutions. This dissertation aims at stretching the boundaries of the level of support that algorithms bring to creative product development, i.e. the development of novel and useful products. The approach taken in this dissertation is to propose stimuli based on patent databases, which contain a vast number of previously documented solutions. Theunderlying assumption is that similar creative solutions might already have been implemented in a different application domain. The recollection of similar solutions over all application domains is not an easy task for humans. In contrast, computers excel at this kind of processing given that the data is in a suitable format, leading to the formulation of the first research goal: Research Goal 1: The development of history-based ideation methods, and automatic and scalable supporting algorithms applicable on patent databases, which are aimed at increasing the novelty and usefulness of the creative outcome of engineers and designers.Often ideation methods adopted in industry are not thoroughly scrutinized. In this respect, the increase of novelty and usefulness aimed for in Research Goal 1 is also too vague. An in-depth quantitative evaluation is needed to understand the effect of the proposed ideation method, as made explicit in the second research goal: Research Goal 2: The development of an experimental test method basedon ideation effectiveness metrics, which can be used to provide quantitativeevidence of the effect of the developed ideation method and algorithms.The structure of the dissertation is subdivided in two parts. The first part addresses Research Goal 1, and describes PAnDA (Products Aspects in Design-by-Analogy), being the method and algorithms developed in order to support engineers and designers by proposing cross-domain candidate products for Design-by-Analogy based on the textual extraction of a patent corpus. The second part addresses Research Goal 2. This part describes the metrics used in the quantitative evaluation, being the quantity, quality, novelty and variety, and furthermore proposes several refinements to the latter two. The novelty metric is refined to allow for the evaluation of systematic ideation methods. For both the variety and novelty metrics, level-based analyses are proposed in order to understand the effects of an ideation method. The variety metric is furthermore refined to account for the distribution of concepts on each abstraction level. Based on these refined metrics, the second part furthermore describes the experimental setup, and analysis of the results, together forming a framework which enabled the in-depth evaluation of the ideation method proposed in the first part.The contributions from the first part, describing the PAnDA tool, to the state of the art are: Automated: PAnDA is an ideation method based on large external sources of data needing no manual intervention to add products. Scalable: PAnDA scales well, as adding patents can be accomplished by folding-in techniques. Domain relatedness: PAnDA differentiates and informs the user whether the proposed products are inter-domain or cross-domain. High novelty, same technical feasibility: PAnDA increases the novelty of the generated concepts by up to 56%, while not reducing the technical feasibility.For the second part, which describes the metrics, experimental setup and analyses of the results, the contributions to the state of the art are: Novelty: The refinement to the novelty metric allows it to be applied to systematic and priming ideation methods. Variety: The refined variety metric accounts for the distribution of concepts over ideas on each abstraction level. In-depth analyses: This dissertation provides an in-depth understanding of the quantitative results of an ideation method. As such, it can serve as areference framework for the testing of other ideation methods.status: publishe