What kind of specs or features should I look for when choosing a non-invasive sound/vibration sensor that can tell the difference between a 50Hz AC induction motor that is stalled vs running normally. A stalled motor typically makes a loud humming sound with lots of harmonics as shown in this video
You could certainly attach a piezo contact microphone to the case, run that trough a 100Hz filter, peak detector and comparator.
But a more reliable method is clamping a current transformer around one of the wires and sensing the current. It always draws an abnormally large current when stalled.
While you are at it you might as well also connect that sensor to a relay that disconnects power from the motor. I'm guessing you are asking this because you have a motor that is likely to stall, and in that case it would be wise to have a automatic system that kills power to it since leaving it stalled for any significant length of time is sure to destroy its windings to charcoal. But a carefully chosen value slow blow circuit breaker should do the job by tripping when experiencing a stall current for too long. A circuit breaker is much cheaper than a brand new motor
My first thought is an FFT. Sample running and stalled motors and look for the difference. Lack of consistency between different motors might be an issue. You might discover common frequencies among stalled motors that you could look for.
But, wouldn't a clamp on current meter be simpler? The non-invasive part might be a bit harder to attain with that.
If you do an FFT with a long enough sample time you should be able to see the 50 Hz line peak plus the 50 Hz - slip rate mechanical frequency of a running motor. A stalled motor will only have the 50 Hz electrical hum. Depending on slip at minimum load you might need several seconds of data to see this clearly so you would have to determine if that was fast enough for your application. You could also look at 25 Hz which will be the rotational frequency of most induction motors but not have a large contribution on a stalled motor unless you are getting asymmetric saturation.
If at all possible I would say do either a phase current monitor or a shaft sensor.