Audio recording analysis as diagnostic method?

I was working on a 2011 Mini Countryman this past weekend which had an awful engine clatter. After my initial inspection I suspected it was not a rod knocking (it had been driven 6 months in this condition and there was no debris in the oil.) I ordered a belt and water pump pully as I could see with a borescope that both were worn out and had head that both can produce a rod knock type clatter. Short story, the pully and belt fixed it.

I decided to take a video as I had not come across a good before/after recording of this clatter noise. When editing the video I was surprised that even the crude audio level chart of the video editing software clearly showed the difference between the two sounds samples (before and after fix.) I then exported the audio and brought it into Audacity to get a better look (image attached.) The microphone was in the same place during both recordings so the audio level between the recordings is true to life (I did not normalize the samples.) Top trace is before, bottom trace is after. [If you are up for a laugh I'll post a link to the video when I'm done editing it. Just a shade tree mechanic in his driveway. ScannerDanner has nothing to worry about ]

While I have used the frequency of a squeak or other noise to help track down which rotating component was making the sound I had never recorded and analyzed the sound before. It seems to me that if one one audio channel there was a reference, such as crank position then it would be possible to narrow down what was making the noise by it frequency relative to crank position. This type of diagnoses may not be useful in practice but I thought it was interesting nonetheless.

As I am not a professional auto tech/mechanic I'm curious if this type of audio recording analysis is done, is pointless, etc. I'm picturing a mechanics stethoscope type probe connected to a computer to analyze the sound and provide some feedback. Maybe a silly idea. I still use a long screwdriver with the handle pressed to my ear [it worked when my grandpa was wrenching on Model T's and still works today ]

I, personally, like the idea. I had an Idea similar to this 20 years ago, before the internet exploded. I would see customer's come in ( I was just a shop helper, back then ) and try to describe the noise their car was making. I got the idea of compiling audio recordings of different mechanical problems/failures, that you could let the customer browse through, until they found their noise. I thought it was a good idea, at the time. But, I never did anything more than talk about it.

Fast forward 20 years: I use a piezo sensor (from an old smoke detector)

that I glued to a magnet to. I stick the magnet to the engine block, and with an ignition sync, I can see the repetition of a noise, as it relates to engine, or any other rotational cycle.

I have not mastered this technique, yet. But, I really like the idea. I can find VERY little information to aid in my quest to improve/expand on this subject.

Great to see how you have explored this idea. I suspect by knowing what is in the belt path one could estimate the frequencies to expect and then make some judgment based on amplitude as you say. I wonder though what would be seen with something like a bad idler. I would guess harmonics about the fundamental frequency of the idler. Sometimes these types of signal can get mixed together (convolved) and separating them is tricky.

The best practical guide I have found to signal processing is "The Scientist and Engineers Guide to Digital Signal Processing". It covers approachable methods of using the data without getting bogged down with the mathematics theories. You can get the PDF for free: www.dspguide.com/

As threatened my sad Mini repair video

Chad, I played around with the audio recordings I made. A low pass filter of 300hz was applied and then I tried to use the same sort of FFT configuration you describe in your video. I'm using Audactiy so the plot is a bit different than yours with the lower frequencies on the left. You can see the knocking sound dominates the before (B4) sample between 0.000s and 0.005s. Since I was not using a contact pickup like you there is more noise as well.

I think your idea has merit but it will come down to having known good reference waveforms to compare to. I also watched your video on the probe construction, very nice. It might help to include an amplifier in the probe handle to provide a higher signal level to the scope.

We looked at audio range vibration diagnostics for fuel injectors just after the forum moved to this platform. A cheap electronic stephoscope the perfect probe.
www.amazon.co.uk/Gunson-77109-Automotive...id=1638816632&sr=8-5

Link to earlier forum thread

Thanks for the link.

I have ordered an electronic stethoscope and will put the bad pully back on the Mini sometime to do a proper before and after capture. My intuition is that this technique might be most useful for mostly mechanical faults. If you have access to the control signals for an electromechanical device (like an injector) there might be more useful information in the electrical signals.

I do wonder though if the ringing seen on the injector might give some indication of its health. If there is less/different ringing one one injector perhaps it is a sign that it is sticking and returning slower, the return spring is weak, or maybe it is a different brand?