Slashdot Mirror
Hand-held "Sound Camera" Shows You the Source of Noises
Zothecula writes "If you work with machinery, engines or appliances of any type, then you've likely experienced the frustration of hearing a troublesome noise coming from somewhere, but not being able to pinpoint where. If only you could just grab a camera, and take a picture that showed you the noise's location. Well, soon you should be able to do so, as that's just what the SeeSV-S205 sound camera does."
Old but cool mechanic's trick: use a screwdriver. Place the metal against a running engine, put the ( plastic or wood ) handle against your ear. Hear amazing things inside of the running engine.
Religous speak to God. Insane are spoken to by God. When all shut up, one can finally hear Shostakovich in peace
Don't know about this particular project, but back when I did my PhD, I open-sourced my sound localization algorithm. Tracks up to ~4 moving sound sources in real-time using 8 microphones.
Opus: the Swiss army knife of audio codec
Because that's really the only time when it's impossible to know where the hell the sound is coming from in my experience.
A ball inside a ball-bearing race typically fails by "spalling": a tiny flake breaks off of the surface of the ball.
As it rolls around the race, the ball makes a periodic "tick" sound whose frequency is related to its rotation.
So... if you record the sound coming from an engine, and you have an index mark input (when the flywheel reaches TDC, for instance) and you know the gearing ratios of all the shafts, the inner race and outer race diameter of the ball bearing races, and the number of balls &c you can relate the frequency to a particular bearing which is going bad before it fails.
You can do the same thing for the races: the inner and outer races rotate with a particular speed relative to the balls, so a crack or spall on a race will also make a sound at a particular frequency.
Essentially, look for energy in the particular frequency that a particular failure in a particular bearing would make based on the engine RPM, and repeat for all races. If you find enough energy (ie - audio volume), you know which bearing is going bad and the nature of the problem.
A bad gear typically starts with a broken tooth: a crack forms at the base of the tooth, resulting in a tooth which doesn't push as hard against the mating tooth in the next gear. This causes the driving shaft to speed up slightly as the cracked tooth mates, and slow down for the next tooth due to inertia.
If you continuously monitor an accelerometer attached to one of the engine shafts you can see this speedup/slowdown signature, and if you know the gearing ratio you can figure out which gear is going bad within the engine. The crack tends to mature over time, so an individual tooth will first become "wobbly" before complete failure.
A Journal Bearing typically wears when the "hole" becomes bigger than the shaft (the oil and mating shaft grind the hole bigger over time). When this happens, the mating shaft and attached mechanics will "wobble" within the hole, causing a noticeable shift in the mass of the engine.
If you continuously monitor an accelerometer attached to the engine block, you can index this wobble to the shaft speed based on the engine RPM and tell if any bearings are failing and how bad they are.
In all cases you can determine the nature and extent of the damage while it is relatively minor - before it damages other parts of the engine (scored shafts, pieces breaking off, catastrophic failure in flight, &c.)
At the time this was figured out the technology was expensive to implement, so it was only appropriate in select situations - aircraft maintenance, for instance.
Nowadays with the rise of high-power microprocessors and personal phone displays, perhaps some enterprising hobbyist will figure out a way to implement this for automobile maintenance.
You're supposed to put the blunt end to your ear.