24
$\begingroup$

I ask this question because I read somewhere that SR-71 was designed in 60's with hand calculations and drafters and when engineers later tried to improve the design with modern computers they found that they got original design exactly correct!

I take this with a pinch of salt but I find it intriguing that engineers were able to do this 60 years ago without CFD tools and CAD programs. Can anyone give me an insight on how they got each curve and line correct? Was it purely a lot of hit and trial in the wind tunnels? What technology or math tools did they use?

Can a mechanical engineering graduate with GA design experience succeed in designing jets like F-16 considering available literature, reports and findings,modern engineering tools and computers available in 2021?

$\endgroup$
7
  • 1
    $\begingroup$ This question seems very broad, but check out the answers it generated - definitely worth keeping open! $\endgroup$
    – Ralph J
    Mar 6 at 13:04
  • 1
    $\begingroup$ It's not broad. It says "like the F-16 (...)". It does not ask about the exact design history of all those planes, rather the methodology compared to now. $\endgroup$
    – ymb1
    Mar 6 at 13:29
  • 10
    $\begingroup$ "...exactly correct..." - please define. $\endgroup$ Mar 6 at 21:50
  • 4
    $\begingroup$ I recommend Skunk Works: A Personal Memoir of My Years at Lockheed. It's a memoir, not a technical book, but it will convey a lot of context. For example, the F-117 is made of flat panels because they're easier to calculate than the curves of a modern stealth aircraft with the tools available in the 70s. How did they test it? They made a scale model, put it on a pole, and stuck it on a radar range. $\endgroup$
    – Schwern
    Mar 7 at 23:47
  • 7
    $\begingroup$ "Exactly correct" is certainly an overstatement. Any aircraft design is a compromise among many competing factors. Assuming (which is silly) that you can quantify the factors and their weights the value will be quadratic around the optimum, so it won't fall off too fast if you are close. The availability of computers has made it necessary to compute thousands or millions of cases. Before computers, experienced engineers picked out the half dozen cases worth calculating. Most (but not all) of the time the critical case was in the list. $\endgroup$ Mar 8 at 4:36
43
$\begingroup$

The short answer is that if we can program it into a computer then we can do it manually, it just takes a lot longer.

Back in the day, a "computer" was a clerical professional or mathematician and with paper, pencil, log tables, slide rule and the like on their desk. The data centre of the day was a large hall filled with rows of desks with a computer working industriously at each one, each carrying out a precisely specified routine. Supervisors collected their calculations and passed them on to the next calculator in the chain, who carried out the next routine. It could take a couple of years to fully calculate out a large airframe for production.

The mechanical desktop calculator looked rather like a small old-style cash register and made their lives a lot easier. Electronic desktop calculators raised their game another step.

The computer was one of the first professions to be wholly usurped by the electronics revolution. Programmers long afterwards talked of writing their code in a modular fashion, as routines and sub-routines.

Nowadays we can figure out the basic airframe in a few weeks at worst, while the wind tunnel models are being 3D printed, and iterate until we achieve joy unbounded. However design for manufacturing has become immensely more sophisticated, not to mention standards-bound, and the timescale has if anything increased.

But all computers suffer from one particular phenomenon, known in the trade as GIGO - Garbage In, Garbage Out. You really do need a competent and experienced designer at the top of the food chain. A GA designer turning their hand to a warplane has a steep learning curve; the likes of SR-71 design lead Kelly Johnson do not turn up still wet behind the ears. (It was not always so; UK company de Havilland badgered the Air Ministry for years, were fobbed off and insulted at every turn for being mere civilian designers, so they built their brainchild as a private venture and called it the Mosquito....)

$\endgroup$
14
  • 1
    $\begingroup$ @GuyInchbald Thank you Mr. Inchbald, were these calculations 2D in nature? could you comment on the number of decimals places they were accurate upto? $\endgroup$
    – Mridul
    Mar 6 at 14:41
  • 13
    $\begingroup$ A computer was not a mathematician. They were semi-skilled labourers at best and took orders from mathematicians (or more likely, engineers). $\endgroup$
    – Nobody
    Mar 6 at 20:29
  • 5
    $\begingroup$ " iterate until we achieve joy unbounded" --> 🤣 (Yes I know comments like this get axed, but gotta love the language). $\endgroup$ Mar 7 at 14:44
  • 5
    $\begingroup$ @Nobody In a lot of cases computers were mathematicians and in some cases even physicists and engineers. A lot of university lecturers and professors were drafted into the job of computers in WW2. The Manhattan Project was famous for sending calculations to 3 or more "computers" to check for errors. These calculations were not simple arithmetic problems but were often differential equations that needs solving. The simplest "computation" were often some calculus. Unskilled arithmetic work was done by slide rules $\endgroup$
    – slebetman
    Mar 8 at 6:51
  • 2
    $\begingroup$ Short answer; a computer could be either, depending on the sophistication of their calculations. Basic stuff was given to clerks with a mathematical bent, more advanced equations were given to proper mathematicians. The clerks were cheaper and more plentiful, but the whole game nevertheless horribly expensive and time-consuming, so designing the routines for least cost and time was an important part of setting up the job. $\endgroup$ Mar 8 at 10:35
33
$\begingroup$

The A-12 and SR-71 were flat like an ironing board for a reason. This way, the pitching moment at Mach 3+ was easy to calculate. At the time of the design, Mach 3 wind tunnels were rudimentary and had super small test sections, so their results were unreliable. Indeed, most work was done with intuition and slide rules. Calculations were done by rooms full of human computers because digital computers were rare and complicated to use (did you ever try to sort a deck of punch cards after you accidentally dropped the metal box which held them to the floor?). Granted, these computers were miles ahead of the digital computer that was used in the wing design of the Messerschmitt 262, but there was little existing software, so if you needed a quick result, those human computers were the way to go.

While today much computation removes around iteratively solving large systems of equations, where each equation describes the state in a tiny speck of a large, there-dimensional grid, back then much was done with differential equations which were solved once for a complete system, be it heat in an engine or lift on a wing. Much data was tabulated and only had to be looked up, but this worked only when you stayed in the realm of tried and tested parameters. For new materials and Mach numbers, development was a combination of basic research and trial and error. To give you an impression what was available for flow parameter calculations, look at NACA report 1135, published in 1951. I'm sure papers like this piled high on the desks of the Lockheed engineers of those days.

But those engineers had much more time to try and learn the new stuff. Meetings were short and to the point, bean counters and lawyers still saw themselves as supporting staff and not (yet) as the center of the company, so much of the bullshit that is keeping today's engineers from being productive wasn't around yet. If you read the biography of people like Ben Rich, you will learn that years of experience from wind tunnel data enabled Kelly Johnson to guess the peak temperature of a shock to a few degrees. I have worked with engineers who could tell you the lift curve slope of an arbitrary planform to two digits after the decimal point simply from looking at it.

With the F-16 much had become different already. Powerful and flexible computer codes allowed to replace wind tunnel research and made complex curved surfaces possible. If you look at the supersonic drag of the clean F-16, you see that the shape of the airplane has been optimized for trans-sonic flow. While the drag coefficient of older designs had a clear peak between Mach 1.0 to 1.2, the drag coefficient of the F-16 stays roughly constant over the whole supersonic range. This is the result of careful tweaking that would had been impossible in the times of only wind tunnels. Also, the belly intake is the result of lots of experience, not only from aerodynamics, but also from operations: Earlier generations of engineers were too afraid of FOD to accept a belly intake.

While today's mechanical engineering graduates have again vastly better tools available to them than the team around Robert Widmer, they will lack all of the experience that comes with having contributed to the development of several airplanes. The result will be respectable, but fall short in many details where experience made the engineers of the 1970s pick the overall best solution.

$\endgroup$
14
  • 6
    $\begingroup$ "did you ever try to sort a deck of punch cards after you accidentally dropped the metal box which held them to the floor?" Well, I didn't drop them, someone else did. The cards had a section for sequential numbering, against just such an event. Turned out I was the only one who had filled in the numbers, everybody else picked through the box in utter dismay. Seriously, the big problem was that the mainframe computers of the day were just not powerful enough. $\endgroup$ Mar 6 at 12:40
  • 15
    $\begingroup$ @GuyInchbald Most of the people I knew ran a black marker diagonally across the edge of the card deck to simplify getting them back in correct order. It also helped protect against pranksters that added an extra card or two with a PRINT command to your deck. And to be fair, analyzing a differential equation with an analog computer was quite fun once you got the patch panel done. $\endgroup$
    – Gerry
    Mar 6 at 13:42
  • 2
    $\begingroup$ @Gerry The coloured marker was fine if you had a freshly debugged stack of cards. But debugging messes up the line and numbers are easier to update. First time round I always coded a lot of bugs. $\endgroup$ Mar 6 at 13:54
  • 13
    $\begingroup$ A basic engineering rule was just as valid with 1950s technology as it is now: Don't design things you can't (or don't know how to) analyze. (And the fact that your latest supercomputer cluster can produce pretty pictures of something doesn't necessarily mean you know what you are doing!) $\endgroup$
    – alephzero
    Mar 6 at 18:19
  • 6
    $\begingroup$ @alephzero I was tempted to try to write an answer along the lines of your comment. When we marvel at the old designs, it is important to remember that behind all those technical designs was a business engine and an industrial complex which centered around being able to identify good goals which were actually within the realm of achieving! $\endgroup$
    – Cort Ammon
    Mar 6 at 20:28

Not the answer you're looking for? Browse other questions tagged or ask your own question.