Ultrasonic transducer

[rosse]

Hello.

I'm interested in measuring ultrasonic signatures in reciprocating internal combustion engines. I want to detect engine faults like leaking valves and piston rings, fuel injection, exhaust blowdown, valve closings,...

I want to measure this events at very high frequencies (35-50kHz approx, then I will need an ultrasonic transducer.

I have a data acquisition system and I would like to know what type of sensor is the better for ultrosonics signatures (high frequency microphones, or accelerometers with a very high frequency range?)

What is the better method of mounting a microphone in the engine? (magnetic, stud mounting, airborne, ...?)

Thank you very much.

 

[GregLocock]

I've never got any data off an engine over 5 kHz that was worth a damn.

Good luck.

I'd use accelerometers, but then I've never used ultrasonics.


Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Warpspeed]

I have never tried anything like this myself to any serious extent, but from what I have read about the practical difficulties of interpreting the onset of combustion detonation ultrasonically, you are going to have a very serious challenge ahead of you.

There is just such a vast amount of general engine noise, modified by acoustic paths and resonances. You will certainly be able to hear it all, but trying to interpret anything meaningful from it may be rather a tall order.

An example of this is listening to someone talk in a pub or disco where the ambient noise level is incredibly high. You may be able to do it with very great concentration, the human brain is wonderful, especially when visual and other clues are added. But trying to interpret the same total racket recorded from a single microphone, with just a spectrum analyser or some clever software, will probably get you nowhere.

 

[BayernHase]

Hmmm!

From an embedded controls perspective....

You can use an ultrasonic transducer in a receive-only mode, yeah, a fancy microphone. How about quantification?

4-cylinder engine, overhead cam, up to 7K rpm ignoring aspiration method.

To determine what data we are looking for one would have to attach an acoustical transfer media to a valve stem; say a poly tube containing a liquid with a specific acoustical index. We attach our sensor to the end of this. Next, we drill into the head and, doing the same for the valve seat; perhaps a metal rod this time.

To establish an acoustical base sample we will need to over-sample the event, odd numbers work better than even, so say nine times over. Thus, one valve and seat, to accommodate max RPM, needs sampled nine times ~fifty-eight events/second each. With incident timing leading things, anticipated by crank position, well … that’s two A/D channels across perhaps ten degrees of rotation – simultaneously – or about a sample rate of 30 usecs ( caused by latency) of at least 12 real bits.

Well that’s a test setup for one valve position in a single design of engine.

To know what a leaking valve sounds like you’ll have to create the base test situation. And, the RPM oriented criteria. Leaking rings? You’d have to test against every ring material composition and then type of oil. Did I mention the valve float phenomena at higher RPMs?

Extrapolation: You seek to create a piece of shop equipment that can detect problems with any off-the-street vehicle. To test the equipment one must be very intimate with the process itself and its’ environment, the methods of testing will come from this intimacy. Current stock engine production does not truly allow for all the required testing. As an example: the sensor(s) may not diagnosis that acoustically created wave backpressure, which helps to seat the exhaust valves, is inappropriate and thus a leaking exhaust valve is indicated when all that’s really required is a new muffler.

Hey, project scope.