From your description I hoped that you would be analyzing audio to solve problems that don't raise codes.
Years ago my daily driver had a noisy wheel hub bearing, so when I heard a similar noise in a different vehicle this year, I assumed it was the same problem. However, when a mechanic investigated, it turned out to be an issue with the breaks instead. I have now learned to discriminate between a bearing and a brake issue, but there are a ton of other conditions I've never heard before that would be obvious to an experienced repair technician.
I bet an app could become as good a seasoned mechanic at identifying mechanical issues from audio samples.
I agree, its an idea I've been playing with for a while also. The difficult part would be isolating the sound, but perhaps with new ai audio models that has become trivial, its definitely worth revisiting
Don't worry about isolating the sound too much. If your human ear can identify a bad bearing even if the radio is on, you can train a model to do the same. Some of this will come just from spectral analysis, which makes it easy to distinguish the rumble of a V8 a belt squeak.
You can also probably guide a user through a couple of system changes to isolate sounds. For instance, a sound related to the alternator may change with both RPMs and electrical load. A sound related to the serpentine belt might be affected by engaging/disengaging a mechanical load like the A/C. A wheel bearing or brake noise will track with the rolling speed, not the RPMs, and may be affected by steering (a hard turn to the right may unload the left side of the vehicle.)
I'm blown away by how much diagnostic info a good repair tech can get just from listening to a car; I would pay to have this kind of guidance in a phone app.
Years ago my daily driver had a noisy wheel hub bearing, so when I heard a similar noise in a different vehicle this year, I assumed it was the same problem. However, when a mechanic investigated, it turned out to be an issue with the breaks instead. I have now learned to discriminate between a bearing and a brake issue, but there are a ton of other conditions I've never heard before that would be obvious to an experienced repair technician.
I bet an app could become as good a seasoned mechanic at identifying mechanical issues from audio samples.