Listening to machines to better predict wear

Cavitation is the formation of vapour bubbles in the liquid, resulting in vibration, loss of efficiency and premature erosion of the turbines. Detecting this phenomenon helps optimise machine service lives and better predict maintenance periods.

Contribution of artificial intelligence

Committed to innovation and the search for new services to improve the maintenance of facilities, HYDRO has joined forces with the HES-SO University of Applied Sciences and Arts – Western Switzerland to come up with a non-intrusive monitoring solution that can be easily deployed in two stages. Initially, the CaVision acquisition module installed on site captures the audio signal from the facility and processes it in real time. The data is then fed into an AI algorithm to come up with a label describing the cavitation. This preliminary level of analysis is used to count the hours of operation under the cavitation conditions, then optimise the machine’s service life.

Detailed analysis over the long term

Long-term data collection and a study of turbine condition can be used to produce a more detailed analysis. Wear caused by cavitation can be estimated based on the number of hours during which cavitation conditions have been present and by observing variations in the noise produced by the facility. 

Watch the video of our predictive maintenance service: