Knowledge-based Feature Recognition for CAD Model Generation of Bifurcated Sheet Metal Parts

This dissertation develops a concept for the knowledge-based feature recognition of bifurcated sheet metal structures within a graph. The objective is to automate the feature-based 3D CAD generation within the algorithm-based product development process. This concept addresses the research gaps in CAD modeling of bifurcated sheet metal parts, pattern recognition, and knowledge-based approaches.
The scientific core of this dissertation is the development of knowledge-based methods and algorithms for the recognition of features within a graph representing the optimized solution of a bifurcated sheet metal part. These methods and algorithms exploit the capabilities of existing pattern recognition and knowledge-based approaches. To support these algorithms, a knowledge-based feature library has been defined. This feature library cluster different knowledge representations necessary to describe features for bifurcated sheet metal parts. Additional methods and algorithms are introduced to derive the part’s geometry from the given graph, calculate the parameter values of the recognized features, and abstract the feature-based model history. The latter stores the sequence and dependency of features.
The validation results confirm the capabilities of the developed algorithm to interpret a given graph in terms of features and parameter values and to automatically generate the 3D CAD model. The concept developed in this dissertation contributes to the increase of the amount of knowledge that integrated into the 3D CAD models. Using this knowledge, knowledge-based algorithms facilitate the achievement of the final part's solution. As a result, a reduction of time to market and an increase in product quality can be achieved.