Perhaps take a step back and consider what aspect quality or result of the vibration is important?
if its vibration that moves/shakes the pump on its mounting too much perhaps a limit switch would be applicable.
if its a change in pitch due to perhaps a blockage or the result of running dry after exhausting the source perhaps a microphone would be useful?
perhaps if you hook up your ADXL345 to your pi and attach it to the pump and runs some tests. run the pump under conditions you expect to occur and see if the resulting data collected can reflect the pumps condition clearly.