How to detect where a hole is made in a metal rod using neural network?

Same length 5 metal rods has a hole (of same diameter) in different places. Using neural network we have to detect where the hole is by using sound produced when the metal rod is knocked.

Note: Instead of metal rod, a plastic bottle might also b开发者_Go百科e used.

1. make your chimes.
2. Determine using your ears how you'll know where a hole is. Does the pitch change? For example, one chime sounds like an A, another like a B? Or does one ring longer than the other? This tells the characteristic of the sound wave we need to measure.
3. record each chime many times, perhaps 50 or so, you don't get burned by a single bad recording. See if you can pare down the bit width or sample length as it will save time later. We don't need a crystal clear super-oversampled recording - we need recordings that allow us to differentiate 5 chimes.
4. designing the net itself is tricky of course. I'd try a single hidden layer first. I'd make its connections span from wave-peak to wave-peak, assuming the pitch is what's changing. I'd make the net as wide as the number of samples' bit width, and as long as the number of samples.

aside - I thought neural nets had fallen to the wayside, replaced by genetic algorithms and FFTs...