Multi-objective optimal design and assessment framework of freeform timber structure oriented by robotic automation construction

Although robotic fabrication has been successfully applied to non-standard geometry structure forms using various materials, there has not been enough research on detailed design guidance for design under the use of robotic arm for construction based on the impact of the technique on design with a higher level of rationality and reliability. And freeform structures are one-of-a-kind, which means that specific working space, tools, and tool paths can only be used for the required design. These are the primary impediments to robotic automation in mass production. As a result, the purpose of this study is to describe the benefits and limitations of robotic timber automation techniques in order to create a comprehensive design system that includes geometry form generation, mesh generation, material, structural performance, modularisation, fabrication, and assessment. Six hypotheses are proposed in this study and will be tested in the following sections. This study transformed a complex system into a multi-objective optimisation system. The impact factors of robotic automation in fabrication and construction are identified through a review of the literature and the use of the self-organizing maps (SOM) text clustering method to provide reference for the indicators. Fuzzy-Analytic Hierarchy Process (F-AHP) and Fuzzy Comprehensive Evaluation (FCE) methods are chosen based on these factors to determine the weights of these impact factors and to evaluate the impact level of robotic technology on design, fabrication, construction, and management aspects using a Multi-input and Multi-output (MIMO) assessment framework. Taking the impact factors into consideration, this thesis chooses complex geometry & particle method, biometrics & reverse engineering method, and machine learning method to generate rational and complex geometry forms. Non-Uniform Rational B-Spline (NURBS) are used to extract variables to optimise the strain energy and mass of the structure describe by numerical finite element analysis method. The mesh of the optimal surfaces is generated and optimised in triangular, quadrilateral or hexagon forms to achieve planar and homogeneous grids. After the whole building information model is complete, the working space for the robotic automation construction system would be set, and the initial trajectories for the fabrication tasks would be generated and optimised in operating time and travel distance. Indicators are selected to build the assessment framework for geometry rationality, structure robustness, automation automaticity, and modular standardisation, information management criteria. 12 plans are put forward by combing different form-finding and mesh modularisation methods. The assessment framework applies the kernel principal component analysis (KPCA) method to determine the weights and grey relation analysis (GRA) together with grey clustering (GC) assess the relations of different indicators on different design plans. In the end, fuzzy c-means clustering (FCM) predicts the level of different plans to support decision making. The results show the initial geometry forms generated by three design methods are more rational in material aspect. The NURBS geometry models provided detailed geometry information on control points, weight factors to optimise the strain energy and robustness by adjusting the morphology. The structural performance of the structure is remarkably improved while giving more substantial rationality to the freeform surfaces. The optimisation results for robotic trajectory planning show that the travel distance and operating time are shortened significantly to enhance motion control stability and to improve the automaticity of fabrication. The assessment framework for the impact level of robotic automation technique provides quantitative results and the assessment framework for design system performance gives qualitative and quantitative evaluation for the plan decision-making. The final conclusion section discusses the six hypotheses and summarise the accomplishments and limitations. Followi g that, further research topics are discussed briefly.