Slashdot Mirror


Musical Wings Reduce Aircraft Stall Risk

notwrong writes "The Sydney Morning Herald is reporting that a Qantas engineer has found a way to help small aircraft avoid stalling at low speeds: pumping sound through the wings. He found that music also works, having tested Spiderbait and Radiohead (nice choices; Spiderbait apparently works better)."

9 of 235 comments (clear)

  1. Re:Obscure reference? by lightyear4 · · Score: 2, Informative

    reference to the Radiohead song 'fake plastic trees.'

  2. Re:Obscure reference? by aaron_ds · · Score: 2, Informative
  3. Re:From the article by Hecatonchires · · Score: 2, Informative

    Aussie. As in Aus tralia. And I don't own any tie-dye.

    --

    Yay me!

  4. Re:As a pilot by drgonzo59 · · Score: 5, Informative
    I think Boeing's flight control computer is based of Integrity-178B by Green Hills that uses a separation kernel. In fact there is LynxSecure, AESecure, VxWorks and LynxOS-178 but Integrity-178B is by far the one with the smallest separation kernel thus the more secure one. Because it can be mathematically proved that it is correct (does what it is supposed to do and nothing more or else),so anything with 500,000 lines (think Linux kernel) is no good for that, need something that is no more than a couple of thousand lines and it still can take up to 2 years to complete the verification process.

    So what do you do if ya want complex and sofisticated system calls that the Integrity-178B doesn't provide? Well, use another real-time os on top of Integrity-178B or make it part of Integrity-178B but run it in user mode. So all the drivers are really in user mode in such a system. This all is needed so that no single program if corrupted can hang the system. (Trust me you don't want an airliner's computer to freeze with a BSoD or with a Oops!-Kernel Panic while in mid-air).

    Another side note, FAA actually has a concrete limit on the failure due to software. So something like no more than once out of tens of millions of flight hours a plane full of people is allowed to completely crash and burn because of a software problem and have everyone on board die a horrible and painfull death and that would be perfectly "ok" with FAA. So the requirements to certify a system (OS) to fly a plane are very stringent. Linux doesn't even come close. It might be good enough to play music though...

  5. Re:As a pilot by ZarkOmicron · · Score: 2, Informative

    The idea is not to predict what the system can be used to do, but rather what each individual system call can do. There would not be a way to invoke a system call to do arbitrary things. If a user level program implemented something that ran into the halting problem, that would not keep the kernel from servicing other user level programs. I believe the primary trade-off of this type of kernel is probably performance, which in this scenario is certainly worth sacrificing (up to a point) for stability.

  6. Re:As a pilot by drgonzo59 · · Score: 4, Informative
    It is not quite the halting problem.

    Imagine that you have a 3 line program that computes the absolute value of some input x say {if(x>=0) abs=x; else abs=-x; return abs;}. Ok do you think you'll be able to write another program that will verify that this program works correctly?

    You would look at it and perhaps see that it has one branch. We give it inputs such that both paths in the branch are taken and then you look at the output and see if it is what you want it to be, and _also_ you look at all the rest of the memory and make sure that it didn't change. Maybe give it some extreme values, like the maximums and minimums and then also look at output and the _whole_ memory.

    Then can you do the same for a 'for' loop that computes a dot product. You just give the program the known input then at every iteration look for some loop invariants and also check the rest of the memory that shouldn't be affected that it is indeed not affected and then check the output.

    Also this means that the code itself has to be written in a certain way in order for its correctness to be checked easily. That means that a lot of nested 'if's are not a good idea, so they try to reduce the branching as much as possible and modularize the program. It is the burden of the software developers to submit their code for certification and pass before it is accepted by FAA or DoD.

    This actually can be reduced to the SAT problem, which runs in exponential time (but there are ways to take shortcuts in some of the cases).

    I would think that some actual application code that runs on it would define what the system does.

    First though before you even let any application code run you have to make sure that no single appliation will ever take control of the memory and cpu for more than it's allowed share. That is what the separation kernel does. You run this small provem and scrutinized piece of code (note: you also need specialized hardware to make sure it will work) that makes absolutely sure (up to a margin of error) that no application will take more than its share of resourses. So if one application crashes it will not crash the system, instead the rest of the machine will continue to work. So that is why sometimes they will run two RTOSes on top of each other with the first being Integrity-178B that will make sure the other OSes on top are partitioned and separated and allocated only a given share of resourses.

  7. Re:As a pilot by drgonzo59 · · Score: 2, Informative

    The point of having a kernel that is proven to work right (up to a margin of error) is for it to prevent other programs to hog or take control of the cpu or other resourses. Here even drivers are considered external programs. Then of course each application provider, which might be different than the OS developers, will have to submit its application to be certified.

  8. Re:Probably a similar phenomenon by Chris+Snook · · Score: 3, Informative

    Blowers aren't half as good as vacuum pumps. If you microperforate the upper leading edge and attach a vacuum pump underneath that section of wing, you can get a wing performance boost as high as 50%. Still experimental, but don't be surprised if you see it before too long.

    The significance of this has nothing to do with adding kinetic energy to the flow, and everything to do with adding small-scale irregularities to it. It's the same reason why golf balls are dimples, why putting a little sand in the top coat of paint on your racing yacht will make it go faster (they have more sophisticated techniques for this now), and why sharks' bumpy skin actually helps them glide more smoothly through the water. The irregularity creates a thicker boundary layer, though I admit I don't know why, which makes it easier for flow to stay attached.

    --
    There's no failure quite as dissatisfying as a complete and total solution to the wrong problem.
  9. Re:Other bands to test... by greylion3 · · Score: 2, Informative

    I'm guessing you watched 'Con Air' recently.
    (it's slightly a slightly rewritten line from that movie):

    "Define irony: a bunch of idiots dancing around on a plane to a song made famous by a band that died in a plane crash."

    --
    Privacy begins with ..