Principles to Automate Inventive Problem Solving Based on Dialectical Negation, Assisted by Evolutionary Algorithms and TRIZ
Durán Novoa, Roberto A.
MetadataShow full item record
An inventive problem (IP) can be defined as a human perception of a situation that has to be changed, but with at least one obstacle which impedes achievement of the desired goal. In practice, they are solved generally using random trial and error, despite the fact that in literature there are several structured approaches to stimulate creativity and deal with them. During the development of products IP solving is particularly important throughout the concept generation, being often conducted intuitively by field experts. This dependence of intuition and expertise is a bottleneck in the design as a whole: intuition for being unpredictable, and expertise for being rare and thus expensive. This dissertation proposes a series of steps, based on dialectics, to decrease IP solving user dependence. First, the reasons behind this dependence are investigated, describing the critical tasks to be performed and developing the necessary characteristics of the tools to be utilized, simplifying several existing ones under a coherent framework that can withstand different levels of expertise. In order to explore and develop the proposed model, it is studied the complementation between people's innate creativity and Computer Sciences, aiming to ideally solve IP automatically. Due to its empirical results, Evolutionary computing and TRIZ techniques are utilized for the development of the study cases, whose results shows the viability of the dialectical hypothesis in the concept generation. Finally new and complementary research directions, likely to deliver concrete results, are proposed