Fault detection of electric motors by sound

Описание: This video demonstrates laboratory setting for two electric motors. One of the motors has a bearing fault and another is a normal functioning motor. Idea is to listen which one has a bearing fault. Answer to the question is in bottom of this description.

Detecting obvious faults by listening is easy task even if you're not expert on this field. But for machine task is more challenging. Normally in industrial environment operators detect faults in machinery by hearing an abnormal sound. Detecting faults is important because failure in some critical component might cause the entire plant to halt. Production losses of downtime can be easily millions of euros.

It is possible to detect faults by measuring the vibration of the motors shaft or acceleration. Advantage of sound based detection is cheap cost of microphones. They do not require direct contact with the motor and are therefore more easier to install. However, loud background noise typical in industrial plants can cause problems to sound detection.

Sound of the motors was recorded and sound processing was continued with using MATLAB. Purpose was to try find out if faulty sound could be recognized automatically using sound processing. There's scientific publications available from this topic. Most of them conclude that sound based fault detection method is feasible.

I'd like to thank ABB for providing me the faulty electric motor. I contacted ABB and asked if they would happen to have extra faulty motor for this test. 51 % of the electric motor failures are caused by faulty bearing. It's the most common failure type. They didn't have this type of failure but they did have some old unused extra motors left in storage. So in the name of science they broke one bearing and sent the motor for me. I really appreciate their efforts.

Answer to the question: motor number 2 has a bearing fault. It can be clearly recognized from it's scratchy noise.