Aarhus Universitets segl

Conceptual Design Tool for Structural Layout Optimization in the Early Design Phase

Publikation: Bog/antologi/afhandling/rapportPh.d.-afhandling

  • Lasse Weyergang Rahbek
This thesis develops a conceptual design tool capable of generating optimized structural layout suggestions for building design in the early design phase. The structural layout of a building is the arrangement and design of the load-bearing elements that support the weight of the building and resist external forces. The structural layout in this project solely consists of prefabricated reinforced (RC) elements. The use of prefabricated RC elements is embedded in the Danish construction industry and will likely remain so in the foreseeable future. Therefore, there is great potential for more effective use of concrete in terms of sustainability and decreasing cost. The proposed design tool can help fulfill this potential.
Action Research (AR) is used to create the conceptual framework of the design tool. The AR analysis consists of semi-structured interviews and a co-creation workshop where architects, engineers, and contractors contribute to the development of the design tool to ensure that the final tool conforms to real-world practice. The final design tool is based on this framework and developed using four core principles: optimization, interactivity, dissemination, and automation.
A novel parametric modeling method is developed in the design tool called Adjacent Polygon (APoly) representation. The APoly method creates a dynamic parametric representation of a given building plan to generate diverse yet feasible structural layout suggestions.
The evaluation modules of different structural typologies are constructed using surrogate models in the form of Neural Networks. The surrogate models are combined in a hierarchical structure to create an algorithm capable of predicting the optimized geometry and corresponding cost for a structural element based on the external conditions inputted into the algorithm.
The entire network of prediction models is then combined with a meta-heuristic optimization algorithm in the form of a Genetic Algorithm (GA) to create a surrogate-assisted optimization framework. A repair algorithm is incorporated into the GA to increase the number of valid solutions generated during each optimization iteration to decrease the convergence time.
The performance and reliability of the design tool are validated through two groups of local and global case studies. The first group consists of parameter sensitivity studies on the local approximation modules for each structural typology. The second group of validation studies examines the design tool’s effectiveness across relevant building plans and scenarios. The corresponding results demonstrate that the tool can effectively adapt to these different settings and produce optimized structural layout suggestions. It is also demonstrated that the design tool can conduct multi-objective optimization and produce a front of Pareto optimal solutions.
ForlagAarhus University
Antal sider203
StatusUdgivet - jun. 2023

Se relationer på Aarhus Universitet Citationsformater

ID: 315940447