Saturday, September 15, 2007

Design by Concensus


Back when the world was young and I still had hair the Navy hired a gaggle of eggheads to contribute to the design of what eventually became the Spruance-class of destroyers (i.e., DD-963 class). At that time I was the Leading Chief of the computer shop for Pac Fleet's cruiser-destroyer force. I was told to give the eggheads access to anything they wanted in the way of maintenance and repair data, which I did with a cheery aye-aye, sir.

Marvelous stuff, watching those eggheads at work, doing their computerized statistical analysis of equipment failures, tracking everything back to the manufacturer on one hand and the Navy schools on the other.

The product of their work was a list of recommended equipment to go into the new ships; only the best stuff as determined by its failure rate, required maintenance man-hours, mean time to repair and so forth.

Which was all bullshit, unfortunately.

At that time (early 1970's) ComCruDesPac had about 137 ships. The analysis covered such things as electric motors, pumps, air compressors, ammo hoists and so forth, the ancillary systems that are the glue of a modern-day warship. (The hull design and the turbine powerplants were determined by other groups.) The objective of the study was to determine the best of that equipment and on the surface, their methods of analysis appeared valid. But in providing them with data I noticed that while all destroyers had high-pressure air compressors (for example) some of them had never failed. (Not many... four, I think.) Same thing for the other components. All of the ships used a certain type of gear-head motors but a few ships had never reported any problems with them. Which brings up a point worthy of mention.

Even though built to the same plan, vessels within a given class are not identical. The ships are built at different yards and while their specs were identical their equipment came from a variety of manufacturers. In the case of electric motors for example, while most of the ships used motors from General Electric or Westinghouse a few of them had motors from manufacturers I'd never heard of. The key point here is that some ships had never reported any form of failure for certain pieces of equipment.

The bottom line is that the study failed to consider the possibility that some equipment had never failed. Their final report identified only equipment that had failed, giving high marks for designs and manufacturers that failed the least often.

Which completely ignored the Really Good Stuff.

- - - - - - - - - - - - - - - - - - - - - - - - - - - -

So what's all that got to do with airplanes? Quite a bit, when it comes to home-builts.

A fairly common thread on various aviation-related mailing lists and newsgroups is someone polling the subscribers in hopes of determining the ‘best' ...whatever. The best way to paint a spam can; the best brand of tire; the best vacuum pump and so on. Which gets down-right scary at times. (One such poll decided that the ‘best' aluminum was 6061 :-)

Polls and surveys, and the methods of statistical analysis that supports them, are valid tools. But only when your sample is an accurate reflection of the population being polled. Ask a room-full of pre-schoolers to define a balanced diet, don't be surprised if the answer is graham crackers and milk. In a similar vein, wood makes the best fuselage (according to builders of Pietenpol ‘Air Campers'), welding is easy (according to experienced weldors) and flying is inexpensive (according to people earning $100k p/a or more). In the case of the New Ship Design Study Group they failed to include the entire population of ancillary equipment, inadvertently limiting their investigation to equipment having a history of failure. (They were aware of the others but deemed them ‘statistically insignificant.’)

SO WHAT'S THE POINT?

The Internet offers unprecedented access to information but does not provide any means of determining if that information is valid. Indeed, within the field of home-built aviation only a small percentage - - probably less than five percent - - of the available information is valid and even then, only in a particular case. The remainder is either skewed by commercial interest or is a reflection of ‘conventional wisdom,' wherein the poster is simply parroting something they have heard.

Common sense has become remarkably uncommon stuff in modern-day America. Given the risk inherent in rising above the ground on wings I believe the wiser course is to treat all information on the internet as invalid until you can test it yourself. Fortunately, with a technical subject such as aviation the required tests are fundamental and well defined. For the homebuilder, especially those lacking an engineering background, the tricky bit is devising methods of applying such tests to their particular situation.

-R.S.Hoover

PS - - So what happened with regard to selecting failure-prone equipment? I've no idea. By the time the first of the new class slid down the ways I'd been retired for a number of years. But it's interesting to note that several of the Spruance-class have been scraped after barely twenty years service. (Navy ships are designed for a minimum service life of thirty years.)

I identified the Really Good Stuff aboard our own ships and submitted a report on the matter, producing a minor controversy with regard to maintenance. Sailors know what I'm talking about and it really doesn't apply to anyone else.

----------------------------

(The above was originally posted to RAH in 2004. Recent posts to this blog [Chugger's Rib, 4 July 2007 and Chuggers Progress - III on 3 Sept 2007] have generated quite a bit of mail from folks who are upset by my failure to use Sitka Spruce, aviation-grade plywood and T-88 glue. The whole purpose of the Chugger files is to explore the use of less expensive materials that are commonly available. This isn't really a new idea. During World War II aviation was forced to use other woods, different glues and so on. History shows the planes (and gliders) flew just as well. Unfortunately, there isn't a lot of quantified data for those alternative materials. )