Guided generation of strut-and-tie models for reinforced concrete structures with parametric graph grammatical evolution

Strut-and-tie models are typically a manual design approach that follows a truss analogy for designing reinforced concrete structures with discontinuities. Their advantage of simplicity, flexibility and their sound mechanical basis are negated by the requirement of engineering judgement and time-consuming iterative development. Automation attempts with topology or discrete layout optimisation often struggle to incorporate soft requirements such as practicality or manufacturability. This limits their applicability in real-world use cases. To address this shortcoming, the authors have previously proposed a graph grammar-based approach using rewrite rules to iteratively transform an initial graph into a truss by incorporating domain biases. Yet, the method relies on user guidance and faces scalability issues. This work proposes parametric graph grammatical evolution, an extension of grammatical evolution combining formal grammars with evolutionary algorithms, to guide the generation of strut-and-tie models. A novel selection method suggests diverse solutions based on the optimisation objectives and the graph topology. Validation demonstrates that the method can generate simple, valid solutions with an explainable generation process. For more complex geometries, it benefits from more intricate initial trusses, which can be developed by the user, emphasising its suitability for human–computer interaction. • Guided and semi-automated explainable generation of strut-and-tie models. • Extension of grammatical evolution to parametric graph grammars. • Selection method of diverse strut-and-tie models for effective design exploration. • Validation of the proposed method with three case studies.