The glass Sashimono joint

Abstract

Glass is gaining more and more popularity as a structural material and has become a big share of the building industry. Unfortunately, the building industry is the largest polluter in terms of industrial waste and is responsible for 40% of Europe’s energy demand (CIB, 1999), and glass is also playing a significant role in this number. One strategy to reduce the amount of pollution from a product or industry is by making it circular (PBL, 2019). Meaning the amount of waste is minimized and the energy needed to produce new material is also decreased. In the Netherlands, buildings have to be completely circular from
2050 (Rijksoverheid, 2016). In order for glass to contribute to this goal, elements have to be able to be reused or recycled, taking demountability into account during the design of a structure.
One way of creating such a demountable joint, is by making use of portal frames, which require a rigid connection between the columns and beams. Currently, this connection is designed using either mechanical connections or adhesives. Rigid mechanical connections are visually not aesthetically pleasing and cause impurities right at the points where stresses are highest. This makes the joint more sensitive to failure. Rigid adhesive connections are very prone to execution and design errors, and are uncertain regarding their long-term strength. Currently, there is no efficient way to properly remove adhesives, making them non-demountable joints. This research will therefore design a demountable and rigid joint using contact pressure, taking inspiration from traditional Japanese joinery. To develop this joint, first, theory is studied, followed by the design and lastly by experimental testing.

From literature, the Kanawa and Gooseneck joints are selected, because they have the capacity to take up both shear and a bending moment. These joints are then further optimized to determine the optimal geometry for a rigid glass joint. This means a geometry that minimizes tensile stresses in the glass, decreasing the chance an existing flaw will tear and cause the material to fail. This optimization is done using analytical and numerical analyses, followed by full-scale experiments. To determine the optimal force transfer the geometries were first schematized, and the relevant parameters were determined
for later variation. From hand calculations, it follows that the optimal geometry finds a balance between the stresses resulting from normal force and the stresses resulting from the eccentricity of the internal line of force.

Using a parametric Grasshopper model, the geometries are further optimized by varying dimensions and curvature. Several designs are imported into DIANA FEA and Abaqus to acquire numerical values for the expected stresses of these set parameters. The models are set up as two 2D glass panes with a polymer interlayer in between them. In DIANA FEA a lot of difficulties arose with the combination of complex geometry and multiple contact surfaces. Therefore all designs were mitigated to Abaqus, because this software is more suitable for complex contact surfaces. Comparing the heavily simplified hand
calculations to the FEA, there was a constant increase of peak stresses with a factor of 4.

The Gooseneck design was manufactured using a CNC milling machine and afterwards, its edge was polished, resulting in optimal edge quality. Due to the nature of the geometry, the Kanawa design had to be manufactured using a waterjet. There was a large difference between the accuracy of the two production methods, resulting in the Kanawa joint having a lot more space between the glass plates. This strongly influences the placement of the interlayer materials, but also the stiffness of the joint during the experiments.

Before these models could be validated using experiments, a suitable interlayer to place between the edges of the glass panes was researched. POM, PVC, Surlyn, PA6 and PU85 are deemed suitable and are examined. Eventually, only PU85 could be fitted between the glass panes, which seemed to have the least favourable mechanical properties. The angle of the geometry was too small for most materials to bend them into, even after heating the plastics. The other issue lay with the tight tolerances of the polished glass panes. These had to be additionally polished by hand and the PU85 was treated with
silicone spray, in order for the whole joint to fit. The disadvantage was that this manual polishing damaged the edge quality, increasing the probability of failure at a lower strength.

Experiments were then conducted to validate earlier analytical and numerical calculations, using full-scale single-pane annealed glass. The joints were tested in pure tension and a bending moment, using polarizing filters to visualize the stress trajectories. For the Gooseneck model, the stress trajectories and expected stiffness corresponded well with the models. The samples failed at an average force of 6.0 kN. The model predicted peak stresses of 300 N/mm2 and stiffness of 2.8 N/m at this point, the experiments displayed a stiffness of 3.0 N/m. This means the model turned out to be 5.7% less stiff than the
experiments.

The stress trajectories coincided less clearly with the model for the Kanawa tension model. The samples failed at an average force of 4.1 kN. The expected peak stresses at this point were 150 N/mm2 and the stiffness 0.73 N/m based on the model. The experiments showed a stiffness of 1.1 N/m. The model underestimates the stiffness of the experiments by 33%.

Interestingly, because the tolerances in the Kanawa joint were larger, there was more movement possible in this joint. This influenced the force transfer and therefore resulted in different peak stresses than expected. The Gooseneck model turned out to be almost 3 times as stiff as the Kanawa model. This has two likely reasons. First of all, the geometry of the Kanawa joint is not designed to take up pure tension in the direction that it was tested. Therefore, the geometry itself was a lot less stiff than that of the Gooseneck joint. Secondly, the tolerances were of large influence. Because the Kanawa samples were produced using a waterjet, with quite large tolerances, there was a lot of movement possible in the joint. This meant little force was necessary to displace the joint, resulting in a lower stiffness.

The Kanawa design was tested under a bending moment, because the force transfer is very different compared to pure tension for this design. The full beam had dimensions of 2400mmx 400mmx 10 mm. Locations of peak stresses were similar to the models. The samples failed at an average of 4.0 kN, which corresponds with a moment of 1.1 kNm. The force-displacement graph of the experiments was not linear, but showed varying stiffness with plateaus where the stiffness was around 0. Most likely, this was caused by a combination of the plastic deformation of the PU85 and movement and/or sliding in the
machine itself. It was attempted to calibrate the model to the experiments, by increasing the stiffness of the interlayer. This did not result in sufficient stiffness, which implies the stiffness originates from another element in the setup. The rotational stiffness was 611 kNm/rad, which is 9.1% of the stiffness compared to a solid beam of the same dimensions.

This means the designed joint is not fully rigid, but this exploratory study shows there is great potential for such a system. Further optimizing the geometry and finding a more suitable interlayer could result in a rigid and demountable glass joint, as part of a portal frame.