Hello everyone,
I’ve recently been researching digital form-finding for compression-only structures. While comparing Dynamic Relaxation (DR, via Kangaroo) and Thrust Network Analysis (TNA, via RhinoVault), I’ve run into a few questions regarding their underlying logic. I’d love to get some insights from the structural and computational design experts here:
1. Structural Definitions: Shell vs. Vault
Frei Otto and Heinz Isler frequently used physical hanging models (simulating gravity to find pure tension forms, then inverting them for pure compression) for form-finding. From my understanding, the structural behavior of these inverted forms is compression-only. However, why are their works predominantly classified as “Shells” rather than “Vaults”? Is this simply because the construction materials (e.g., concrete, membranes) are thin and continuous, or is there a fundamental difference in how they are defined regarding their capacity to handle membrane stresses and bending moments?
2. Algorithmic Tolerance and Equilibrium Differences(Fig.2)
Using the exact same initial boundary mesh, I set the spring Length in Kangaroo to 0 and applied a Load, which successfully converged into an inverted compression-only form. However, when I feed this same mesh topology into RhinoVault, it gets stuck during the first phase (Horizontal Equilibrium) with a massive angle deviation. Since DR can successfully find a compression-only solution, why does TNA’s horizontal equilibrium conclude that this topology is invalid or unresolvable?
3. A Coincidence in Unbalanced Forms? (Fig.3)
Following up on the previous point: even though RhinoVault fails to achieve horizontal equilibrium, if I force it to execute Vertical Form-finding anyway, the resulting 3D geometry looks almost identical to the result generated by Kangaroo (Length=0). Does this mean that when TNA fails its horizontal iterations, its underlying force density distribution defaults to the same uniform tension state (like a soap film) used in DR?
4. Practical Terminology
If I ultimately use these Grasshopper mesh form-finding workflows to generate a pure, zero-thickness geometric surface (prior to any volumetric engineering or discrete block tessellation), is it more accurate in an academic or professional context to refer to this output as a “Shell” or a “Vault”?
For context, my setup relies on the NGon and RhinoVault plugins within Grasshopper.
Thank you all in advance for your time and insights!
Fig.1 Kangaroo (Form Finding)
Fig.2 Horizontal equilibrium NOT found! Consider running more iterations.
Fig.3 TNA vs.DR
Shell or Vault.3dm (460.3 KB)
TNA vs. DR.gh (24.6 KB)