Agent-based modelling methods for fabrication-oriented design of adaptive modular structures made from precast concrete

• Applicant: Professor Achim Menges
• Subject Area: Structural Engineering, Building Informatics and Construction Operation ; Architecture, Building and Construction History, Construction Research, Sustainable Building Technology
• Term: from 2019 to 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 424070949

Highly efficient, load-adapted structures are characterized by a high degree of component differentiation. According to the principle "individuality on a large scale - similarity on a small scale", this variability is to be realized by adaptive basic modules under the conditions of serial production. In order to activate the potential of adaptive modular construction methods with flow-based manufacturing methods for the building industry with regard to speed, quality and simultaneous flexibility, computational design methods are therefore necessary, which enable both adaptability of the component geometry as well as production-oriented planning and design. The focus of the proposed project is therefore the question of suitable design methods for the design of modularised concrete structures, taking into account multi-criteria aspects of segmentation, joining, production and assembly. The goal is the development of consistent digital models that meet the requirements of the design process and form the basis for digital fabrication. However, the focus is not only on consistency, but also on the integration into the design through systematic feedback of requirement parameters from design phases that are typically only considered during construction development phases. This is not a sporadic feedback in the design process, but high-frequency iterations through automated, rule-based digital "loops". In addition to the aforementioned criteria, special attention is also paid to the interfaces with structural design and LCA and LCC.The proposed approach is agent-based modelling (ABM), whose ability to map complex and reciprocal relationships beyond database-driven or parametric planning methods has been demonstrated in numerous preliminary work of the applicant in the field of integrative, computational design. Based on this methodical framework, the ABM method is to be extended with regard to the requirements from the planning and production of precast concrete constructions. The development of ABM methods for the fabrication-oriented design of adaptive modular construction methods from precast concrete parts therefore includes the analysis and systematisation of system and process parameters, e.g. with regard to modularisation, functional integration and formwork planning, as well as the transfer of the relevant system and process parameters into a rule-based agent model. The synthesis of the requirements in the agent-based modelling and planning approach first takes place as a generally applicable planning methodology for digitally produced precast concrete elements before the method is validated and evaluated in a specific case study. A new, essential contribution of this project is the optimization of agent behavior through reinforcing learning.In the future, the method is not only relevant as an integrative design approach, but also as a control and regulation method within the framework of a sensor-integrated CPPS.

• DFG Programme: Priority Programmes
• Subproject of: SPP 2187: Adaptive modularized constructions made in a flux: precise and rapid building