On Pattern Recognition in Rule-Based Topology Modification

Classical topology optimization aims at achieving a problem suited material distribution in a structure by identification of lightly loaded areas and local element-wise reduction of stiffness. The resulting topologic layout often contains small substructures which are complicated to manufacture, hence requiring an additional manual smoothing during the structural interpretation phase. One major drawback of this approach is that the results still have to be interpreted by an engineer and consequently be translated into a feasible structure. In order to gain a first conceptual yet topologically sound design proposal for composite structures, this paper presents an alternate method for an explicit, pattern based topology modification approach combined with numerical simulation of tape-laying technology. It is assumed that certain patterns exist in stress fields that are extractable by pattern recognition algorithms known from image processing. In the case that prototypical structural reinforcements for such stress patterns can be defined, an automatic topology modification algorithm with the goal of increasing the stiffness is feasible. The classification of these stress patterns is achieved by using dimensionless features matching the stress patterns with their appropriate reinforcements. When integrated into a rule-based conceptual design environment, this explicit topology modification offers the potential to generate simple and easily manufacturable topological reinforcement proposals in an automated structural design loop.