OFF-THE-SHELF DATA MODELS: PREDICTIVE MAINTENANCE REVOLUTIONIZES MAINTENANCE IN HYDRO POWER
Salzburg AG tested the use of predictive maintenance for the first time at the "Wald im Pinzgau" hydroelectric power plant, in collaboration with the German AI specialist, LexaTexer. Based on the existing sensor technology and the use of artificial intelligence, a powerful model was created from a data pool. This model can do more than just predict the wear and tear of the turbine impellers. Among the by-products that emerged were digital tools that can be used to detect operational anomalies at an early stage. The results of the pioneering project are so impressive that the application is already being rolled out to other power plants.
In the pilot project at the power plant in Pinzgau, the focus of the investigation was on predicting the wear and tear on the impellers of the vertical-axis Francis turbine. As a rule, the impellers, which are 1,475 mm in diameter, are inspected once a year. A key point was that the project can be implemented without installing additional sensors on the turbine. The measurements should consciously manage with the existing possibilities on the machine set. The centrally recorded values, such as speed, drop height or bearing temperatures, which are necessary for operational management anyway, provide valuable data for the use of predictive maintenance.
Moreover, during data collection and selection, it was possible to draw on 12 years' worth of operating data from the plant, that was automatically stored by the control system. The platform that was developed by LexaTexer enabled the inclusion of written records as well. The experience of the operating staff, who, thanks to decades of practice, have profound knowledge of the intricacies of the plant, was also incorporated into the data pool.
PREDICTIVE MAINTENANCE IS THE FUTURE
Salzburg AG was surprised at how accurately the prediction model was able to forecast the condition of the wheel. Within two years, several series of measurements were taken, and the predicted wear always fell exactly within the measurement tolerance. The predictive maintenance project also produced a broad spectrum of useful digital "by-products". For example, the software is able to suggest that various influencing parameters be adjusted, in order to obtain a preferable maintenance date before too much damage occurs to the impeller. An anomaly analysis system has also been developed. It detects unusual behavior of the system, even before any fault becomes obvious, or before the threshold values that would lead to an alarm being triggered are reached. "Based on the precise forecasts for the Wald power plant, it was decided to also apply the predictive maintenance method to other Salzburg AG plants as well.
The average maintenance cycle for the impellers could be extended by up to 31%. LexaTexer CEO Günther Hoffmann believes that the success of the project is a combination of Salzburg AG's operational knowledge, the right data and a high-performance technical platform (LXTXR) that can easily be integrated into established processes and the existing IT. "It was a complex task, the joint resolution of which ultimately benefited both sides". Salzburg AG is convinced that predictive maintenance will play an even more prominent role in the years to come. Future predictive maintenance projects will focus on the integration of different types of turbines, and on extending the scope to include electrical engineering components (e.g. generators) in Salzburg AG's power plants.
We offer predictive analytics automation to help companies make direct decisions by integrating and analyzing unstructured, structured and alternative data with machine learning technology.
For more information, contact us at email@example.com.