ONR Paper idea: "The Need for Design Tools"

From Mckwiki

Jump to: navigation, search

The Need for Design Tools

All of the naval architecture tools developed and discussed in this series of workshop have been essentially analysis tools. They have spanned a wide range, to wit:

  • A range of sciences: Hydrodynamics, hydroelasticity, structural mechanics, etc.
  • A range of levels of complexity / fidelity: From FLUENT to Dr. Noblesse's "Only The Bow" model
  • A range of levels of integration: From a stand-alone use of Star-CCM+, to LEAPS

But what is "Design?" Design is necessarily synthesis, not analysis. In analysis you have a given geometry (for example) and you calculate its performance. In synthesis you have a given performance, and the challenge is to find the geometry. Unfortunately we model design - and we teach design! - as a series of iterative analyses, wherein we try one idea after another, to try to find out which one is closest to the desired performance, and to give some hints as to what we need to change to get even closer.

This is not actually a correct model of design. Even when design is being done by repeated and recursive analysis, it is being directed by a human mind that is attempting to 'run the problem backward' and perform synthesis. In effect, he is performing synthesis mentally, and then using analysis tools to validate or re-direct his synthesis efforts. But the tools that he is using are inherently analysis tools, not design tools.

What would a design tool be? A design tool should be able to START with a level of performance, and GENERATE the physical object that delivers this performance. Let's imagine:

  • A structural design tool would start by being given a geometry, a set of loads and allowable deflections, and would then allocate material to most effectively carry those loads within those deflections. MAESTRO is the nearest thing I know of to this tool. In MAESTRO once you set up the basic FEA analysis you can click on a model that will move some metal from the lightly-stressed areas over to the highly-stressed areas – a ‘synthesis’ function.
  • A CFD design tool would start by being given a flow characteristic - perhaps by being given a desired pressure distribution on the body - and it would then modify the body until that pressure field resulted. Can you imagine a CFD tool where, once you get the initial ‘pretty picture’ of the pressure field, you can then click on some sort of ‘paintbrush’ and spray-paint the colors to how you want them to be (“..let’s just smooth out that orange spot over there…”) and then click on something which runs that panel code ‘backwards’ to generate the geometry that will yield the hand-painted pressure field?
  • An idealized ship resistance tool would be given a power curve, and would generate a ship hull that possessed that amount of drag.
  • A ship dynamics tool would be given a maneuvering or ship motion goal, and would modify the hull to generate it.
  • A machinery design tool....well I can hardly imagine it!

In the sketches above I have attempted to only paint a brief picture intended, not as guidelines for tool development, but more as illustrations of what I mean when I highlight the difference between design and analysis. A few of those word-sketches are realistic. A few of them are probably impossible. I intend them to span that range.

But I purposely started that list with at least one example that shows that synthesis tools are indeed possible. And my purpose in this short essay is to invite this community to consider expending at least some of our effort in developing tools of this sort. We have excellent analytical tools, spanning a broad range of disciplines, precision, and purpose. But we have darn few - too few - tools that work by synthesis.

Personal tools