i wrote a paper on the feasibility of analyzing the sound of modern engines via FFT and weighted time domain to see if we could compare the known healthy sound to the bad sounds and diagnose problems, there is a lot of promising results, however seeding the database to make it useful out the gate is a non trivial problem.
turns out it's sorta already used a bit, sensors on fancy industrial equipment are placed near major bearings and if the computer begins to detect too much correlated vibrations to shaft RPM and machine load it will flag the operators that a bearing should be replaced before it actually fails, neat.
i used an FFT app on my laptop once to prove to my land lord that his refrigerator in my apartment was making an insane high pitched noise that i could hear but he couldn't hear.
my friends even complained as soon as they walked in my door.
a nice big spike at about 18-19KHz, i don't recall specifically how high it was, but it was way up there near the limit of people perception, it was modal as well, depending on where you placed your head in the kitchen it would be louder or not noticeable.
i could make the big spike on the computer FFT go away by unplugging the fridge
with this evidence, he replaced the fridge for me, that was the nicest thing he ever did for me, then he wanted to increase rent by 17% the next year so i moved out
anyway, just record the sound of your engine and upload it, maybe we can all take turns running it through different analyzers
-s
