Managing variable-dimension structural optimization problems using generative algorithms

This article introduces a novel design abstraction concept for efficient truss topology and geometry optimization. The core advancement introduced here is to represent truss topology and geometry using rules of generative algorithms, and to operate on generative algorithm rules using a genetic algorithm rather than directly on the design description. This indirect design representation supports efficient exploration of variable and high-dimension design topologies. Generative design strategies are also independent of any kind of ground structure, thus avoiding the inherent limitations of ground structure approaches that may hinder innovative design solutions by defining a priori what topologies may be considered. We present new generative algorithm strategies that automatically satisfy structural stability constraints, and that can produce truss topologies with a diversity of patterns within an individual truss design. Truss topology and geometry is optimized in an outer-loop by a genetic algorithm that operates on generative algorithm rules, and size optimization is performed in an inner-loop for each candidate topology using sequential linear programming. The proposed methodology supports concurrent optimization of truss topology, geometry, and size. The generative algorithm abstraction layer also supports the design of variable-dimension structures, which can be generated from the same fixed-dimension rule set. Finally, we demonstrate the effectiveness of the new methodology by examining archetypal two- and three-dimensional truss design optimization problems.