Two-stage automatic structural design of steel frames based on parametric modeling and multi-objective optimization

Traditional structural design involves drawing recognition, repeated modeling, parameter tuning, and numerous mechanical analyses by skilled designers, which is time-consuming and inefficient. To address those problems, a two-stage automatic structural design of the steel frame based on expert experiences is proposed. In the first stage, from the computer-aided design plain drawing, semantic features (walls and openings) and geometrical information of architectural elements are extracted by a layer classification method. The segmentation of rooms is conducted by an enclosed region detection method and the connectivity graph is generated using the connected component analysis method. Based on expert experiences considering both structure and architectural function requirements, the structural member configuration and the floor load distribution are automatically established to obtain the parametric structural model. In the second stage, a modified particle swarm optimization (MPSO) based on expert experiences is proposed for single-objective structural optimization according to the design codes. Then based on MPSO, a hierarchical multi-objective optimization method is adopted to obtain more available solutions with different economic benefit and redundant safety. The results show that the proposed two-stage structural design framework is fully automatic and highly efficient. It integrates parametric modeling and structural optimization, and also enables effective transfer of different data items including architectural plan and structural model. It provides a guideline to automatic structural design of steel frames.