AI-Based knowledge extraction for automatic design proposals using design-related patterns

• DOI: 10.1016/j.procir.2021.05.093
• Publication Date: 2021-06-02T21:43:41Z
• Authors: Carmen Krahe; Maksym Kalaidov; Markus Doellken; Thomas Gwosch; Andreas Kuhnle; Gisela Lanza; Sven Matthiesen
• Keywords: Component (thermodynamics); Generative Design; Tree (set theory); Digitization; Design process

Engineering competence and the digitization of all processes along product development process are highly decisive for today's success industrial companies. The design is very individual strongly based on engineers' experience. Part this knowledge result approach fixated in existing variations generations, but difficult to extract formalize. Conclusions about design-related patterns between products different generations or variants can be drawn from model tree representing engineer's thinking each CAD model. However, has hardly been used so far. aim paper examine whether there exist any common models certain component classes by exemplary use case area mechanical engineering. To identify out complex data sets, Machine Learning (ML), especially Deep Learning, proven an immense capability. Finally, learned patterns, meaningful next steps proposed form assistance system. results show that various components. It illustrated class those train system Recurrent Neural Networks (RNNs). corresponding were extracted application partner. By transferring these new variants, one hand itself thus time market shortened. On other hand, previous contained preserved. For further research ML algorithms a contribution faster extrapolation.