To meet the challenge of reducing material-consumption and managing a growing population, the construction industry has to think of new ways. The Cluster of Excellence Integrative Computational Design and Construction (IntCDC) at the University of Stuttgart develops methods and solutions to address this. Besides timber buildings, one concept is to use fiber-polymer composites with high strength for long-span structures and lightweight building systems. Robots are mainly used in manufacturing for automated fabrication, and a co-design framework is developed to design and analyze the coreless filament wound (CFW) structures taking into account design, simulation, and fabrication. For that, small-scale and large-scale specimens under different loads and test scenarios were created and investigated in [1]. In this context, it is still challenging to set up accurate numerical models to correctly predict the behavior due to the many deviations and uncertainties in the built components compared to the design and simulation assumptions, see Pérez et al. [1]. These issues result in high safety factors and reduced material efficiency. To circumvent this, an approach is presented to quantify these uncertainties, e.g., the cross-section in general and its shape, which vary due to the winding sequence of the robotic assembly, the interaction between fibers, and the winding tension. In the initial stage, we utilise descriptive statistics, to organise the data gathered from experimental testing, and draw correlations and inferences to measure the degree of uncertainties (see e.g. [2]). Furthermore, we want to reduce assumptions in the structural model to a minimum. Our aim is to increase predictability of the behavior and the structural performance of this particular type of structure. The key issue will be transitioning from a probabilistic model to a decrease in safety factors.