Conceptual structural system layouts via design response grammars and evolutionary algorithms

摘要

Two new methods to generate structural system layouts for conceptual building spatial designs are presented. The first method, the design response grammar, uses design rules-configurable by parameters-to develop a structural system layout step by step as a function of a building spatial design's geometry and preliminary assessments of the structural system under development. The second method, design via optimizer assignment, uses an evolutionary algorithm to assign structural components to a building spatial design's geometry. In this work, the methods are demonstrated for two objectives: minimal strain energy (a commonly used objective for structural topology optimization) and minimal structural volume. In a first case study three building spatial designs have been subjected to the methods: Design via optimizer assignment yields a uniformly distributed Pareto front approximation, which incorporates the best performing layouts among both methods. On the other hand, results of the design response grammar show that layouts that correspond to specific positions on the Pareto front (e.g. layouts that perform well for strain energy), share the same parameter configurations among the three different building spatial designs. By generalizing, specific points on the Pareto front approximation have been expressed in terms of parameter configurations. A second case study addresses the use of a generic material and generic dimensions in the assessment of structural system layouts. The study applies a technique similar to topology optimization to optimize the material density distribution of each individual structural component, which can be regarded as a part of determining materials and dimensions in more advanced stages of the design of a system layout. This optimization approach is applied to the layouts that are part of the Pareto front approximations as found by the evolutionary algorithm in the first case study, the study shows that-after optimization-the fronts remain the same qualitatively, suggesting that the methods produce results that are also useful in more advanced design stages. A final case study tests the generalization that is established in the first case study by using the found configurations for the design response grammar, and it is shown that the generated layouts indeed are positioned near the desired positions on the Pareto front approximation found by the evolutionary algorithm. Although the evolutionary algorithm can find better performing solutions among a better distributed Pareto front approximation, the design response grammar uses only a fraction of the computational cost. As such it is concluded that the design response grammar is a promising support tool for the exploration and structural assessment of conceptual building spatial designs. Future research should focus on more types of structural elements; more objectives; new constraints to ensure feasible solutions, especially stress constraints; and the application of state-of-the-art techniques like machine learning to find more generalizations.