Build & Correct or Design & Build

2010 February 2

I had a great conversation with a gent who is a multi-talented fabricator/inventor about how to go about designing a machine. He proposed that once you have the basic idea, just start building. As you run into obstacles,simply make changes as you go. If the whole thing doesn’t work the way you want, you can further modify it or start from scratch. This approach is one that I have used successfully. Particularly in cooking, I often start with a basic idea or a couple of main ingredients. From there, I add and improvise until I have a final dish that I am proud of. Sometimes, the dish just does not work for me, so I just throw it out (remembering the problems for future experiments) and start over.

On the farm, we use this iterative approach frequently  for problem solving. We have plenty of tools, inexpensive material, and generous tolerance for mistakes. Our irrigation systems and produce wash area are in a constant state of changing and upgrading, without a grand design drawn in detail at the start. We learn new information as we go and make the needed changes. It is a very effective design methodology for farm infrastructure. We do not have the time, allocated funds or sufficient knowledge at the inception to create a fully detailed grand design.

This approach is NOT how airliners are built. Yes, there is modeling of components and iterative design at some level, but that is not the primary way of designing aircraft. Aircraft are too complex, too expensive, and rely upon too many people to be left to a purely iterative build > correct > build > correct > build approach. A team of designers, engineers, and scientist apply knowledge, physics, chemistry, material science, CAD, and other disciplines to designing the complete aircraft and individual components before much fabrication takes place. It is designed, then built. Iterative design happens within that process.

So which way is best for the roaster project? It seems to me that a combination of the two basic approaches makes sense. For the sample roaster(s), it probably makes sense to do the build and correct iterative method. I will be using smaller amounts of inexpensive materials and very little subcontracting of components (making it in my shop). Futhermore, I do not know everything I would like to know and will need to learn during the construction of the devices. It is likely that the sample roaster(s) will catch fire, seize, warp, discolor and other unpleasant things. This will help me foresee problems as I correct them. One or more sample roaster will end up in the dumpster, I am certain.

Taking these results, my engineering education, plenty of thought, advice of others, and the technical skills of others, I will then design the roaster. The design will use fully detailed mechanical drawings (I will do them by hand) for most components. The design will incorporate an understanding of relative expansion of various materials, static and dynamic loading, thermodynamics and other disciplines. The drawings will then go to subcontractors for fabrication. This approach makes sense for the roaster because: I should have enough collective knowledge to do a complete design, mistakes will be much more expensive, there are many fabricators and components involved, the final product needs to be of top quality (longevity plus fit and finish) AND I it forces me to have considerable mental discipline.

2 Responses
  1. warfin permalink
    February 2, 2010

    Hi, Joe,

    I enjoyed the post. Airplanes, cars, and super colliders are designed and tested on computers these days, but that wasn’t always the case. Since these complex machines are based on a history of trial-and-error, reliable experiments and test criteria. This information can be assembled and modeled in computers and systems can be tested very quickly in a wide variety of configurations far quicker than the prototype methodology. The key is coming up with appropriate experiments to model and develop the specifications of the roaster. A pretty daunting task when you think about it holistically. However, you already know a number of things that can be used to model or prototype individual components of the roaster. Individual experiments and prototypes can be developed to test the roaster piece-by-piece.

    I agree, building a complex machine without modeling the components and deriving specifications is a long shot gamble at best. However, prototyping components (burners, vessel, etc.) of the final assembly is an important and relatively simple part of determining what works.

    Also, a big reason I like to prototype is something left out of modeling equations – it’s fun!

    • February 3, 2010

      Well stated! I do look forward to the prototyping of the sample roaster(s). The inventor in me loves prototyping. The engineer on me says do as much design on paper first. I obviously have internal warfare going on!

Comments are closed.