STRUCTURE STUDY
Triangulation On Double Curved Profile
Hexagonal packing on double curved surface can cause distortion, we can get a fine packing of hexagons on a circular tube, but not no a double curved surface. the problem we cannot keep the area, the angle and the edge length even at the same time. Keeping the angle as constant results in uneven in the area of cells, keeping the area constant results the bow-tie shape in the concave surface, while keeping the area results in variation in shapes, from the pentagon to heptagon.
To further investigate about the packing of polygons, two approaches are tried, first is the circle packing on double curved surface, second is the meshing machine in grasshopper, rhino.
We got the function called 'mesh dual', which can transform the triangulated mesh to a collection of polygon cells. The function allows certain freedom is meshing, to follow the rule for the even angle or even edge length, the importance of following the profile which contrasts with the importance of keeping the cell size; all these results in the control of the final meshing.
The problem of the meshing machine is losing control in array patterns of the cells, we can hardly predict the position of mullions, we can not array the cells according to the floor height, every cell would be unique, which means the standardization will not be possible. The angles among cells would be irregular, which means calculations will be difficult and the assembling will be difficult, every component needs to be unique. Although the
method gets high adaptivity into shape, it is not practical.
Structural Optimisation
Another approach is getting a structural optimisation tool for different shell grids, to see if we can get a tool to examine the exoskeleton behaviour, which parts are referring to which kinds of loadings. The tool we chose is the plugin Millipede in grasshopper, rhino.
The Millipede is a relatively serious optimisation tool for design, which results in a requirement in detail component setting and yet, it still gets some critical weaknesses that limit the tool as just a design reference.
The system is defined by material, cross section and boundary connection, the pre-set material include steel, iron, aluminium, glass and most of the common materials, we can define the component cross section profile and all the dimensions. The boundary conditions, which means the connector to the fixed ground, can be controlled by six boolean options, RX RY RZ TX TY TZ. RX stands for rolling through x-axis while TX stands for transition through x-axis. By controlling the six parameters, we can create fix joints, hinge joints and roller joints through the XYZ axis.
The optimisation method can further be defined as the 2d plane optimisation, the shell optimisation, stress line pattern optimisation, frame optimisation and finally a solid optimisation.
At first, the 3d optimisation is used as the tool of form finding, but the trial is not successful as the fineness of the simulation relates inversely with the complicatedness of the simulation method, the form extracted is quite rough and gets a humble amount of reference value.
For the another extreme, the planar optimisation can show a nice pattern and with a fast optimising speed, since the force can only stay on the working plane, the shearing, torsion, and bucking can hardly be reflected in the planar optimisation.
While the methods left, the frame and the shell optimisation, the frame optimisation shows a more relevant result.
For the optimisation of frames, a matrix can be set, with the two shell pattern shapes, the rectangle and the hexagon, for each shape, the patterns are further divided into 1, 2 and 3 degrees of structure. For the type of forces, we mainly investigate the self-weight, the lateral push, the inward pressure onto the shell and torsion onto the shell. To keep the test fair, the number of iteration of the shells are kept constant, which is 50 in the tests.
From the tests, the multi-layered structures show a general tendency, that is the structure would generally converge to the member with the higher level. The another observation is, the single-layered structure takes more iterations to deform. The even distribution of structural member matches with the structural need of shell structure more, as the loading can be evenly distributed.
There is a limitation on the structural analysis. First, the nature of the nodes are pre-set as the fixed joint, no further nature can be defined, second, the thermal factor can not be simulated by the plugin, buckling within the component is neglected, and the factor that concentration of the loading causing partial structural deformation and failure can not be simulated. That is the reason of the tendency on the thickening of the main structural members.