Layout Optimisation

   
Once a desired surface geometry has been found, some sort of structural grid will be required to support it. There are many possible approaches to this problem.
     

Subdivision Grid

An obvious solution is to use the subdivision mesh itself (triangles or quadrelaterals) to define a structural grid. Different levels of subdivision can greate grids of different density, and can be iteratively subdivided down to produce panels or members of the required size.
Low density sub-div mesh High density sub-div mesh
Alternatively, subsets of the subdivision-mesh can be included in the structural model, either based on geometric patterns or on some sort of tree-growth algorithm. Rather than simply deciding whether an edge becomes a structural member or not, the members sizes can be adjusted to give a hierarchy of primary and secondary structure. Select members based on patterns Select members based on algorithms

Selective Removal

To optimise this even further, the geometry can be automatically linked to a finite element analysis program and analysed for a given loadcase (such as selfweight). Members which are unstressed can be removed and a new analysis performed. By sequential deletion of under-utilised members the structure can converge on an optimum layout, with sufficent members present to support the load but no more.
Iterative removal of under-utilised members

Draping

New grids can be drawn in-plan and projected onto the limit surface, but this will only work for fairly flat surfaces. A more optimal solution from a construction point of view is to form a new mesh with fixed edge lengths but variable joint angles and 'draping' it over either the base mesh or the limit surface.
Grid draped over base-mesh Grid draped over limit-surface
By attaching the vertices to the surface (base mesh or limit surface) this optimisation can either aim to give equal memebr lengths or equal joint angles.. Equalise Member Lengths Equalise Joint Angles