im working at the moment at a project for observing a water pump with an acceleration sensor and hopefully beeing able with the help of maschine learning to predict shutdowns (predictive maintenance). therefore i have to do (a lot of) signal processing and try to find some “markers” where i can say “hey this might be a good indicator” – so i found this website and the forum on the search for ideas and implementions. I tried a lot of tutorials.
the background of the shutdown of the pump is that a filter before the pump! If the filter gets too much blocked the pump will shutdown it self (cause the pressure will go down) – so in my optinion there should be some trend i could find.
now i had the idea of inspect the phase of the signal, this might change – cause the pump has to work a bit harder against the litte blocked filter.
over all – i want to ask if someone might wanna help me with some ideas or even with examples.
here also a few specs of the pump:
5,5kW, 400 V, 11 A and 50 Hz
revolutions per minute: 2940 1/min –> 49 1/sec | Hz
here some signaldata of the sensor: always 60 sec a day over a few days. here i can say that the pressure gone down from 1,8 bar (26,1068 Pound-force per square inch) (new filter) to ~ 1,65 bar (24 Pound-force per square inch) –> pump will auto shutdown at ~ 1,4 bar
Edit: Fs for the signals in the workspace is 6400.