Centrifugal pumps driven by asynchronous motor are widely used in industrial applications because of their low cost, high performance and robustness. However theses pumps can be damaged by wear and the performances of the machine can be altered.This research work deals with fault detection and diagnosis of fault that may appear in a pump. Usually, a vibratory or noise monitoring can be established, but it’s often really costly and can be cumbersome. A different approach based on the analysis and processing of the stators’ current is proposed here in order to highlight theses faults. This method can detect a torque drop in the pump shaft, and coupled with a flow measure it can determine if a cavitation is occurring or not. This provisional maintenance gives the health state of a pump in real time and only requires acquiring the current and using RMS and spectral analysis.

For 2 years, Schneider Electric and International Office for Water have work together on this subject.

For now, the methodology researches have been done and a first set of experiments has been led on an existing pedagogical platform at IOW (multi-stage centrifugal pump with an asynchronous motor, current, voltage, pressure and flowrate measurement).

The next objective is now to develop a new facility with additional features:

• Torque and speed measurement
• Test on a synchronous motor
• Use of a mono-stage centrifugal pump

With this new pumping loop, we will be able to test this new detection method on a different pumping structure (different pump, different motor), and to compare it with mechanical measurements.