A review of condition monitoring in Francis turbines for predictive maintenance

Abstract

Hydropower is one of the important sources of clean energy, where the use of the Francis turbine is the most than any other turbine. During off-design operations, turbines undergo significant fatigue loading, pressure pulsations, and vibration, which can be detrimental to the plant if operated for a long period of time. This paper aims to provide a general overview of (PdM) predictive maintenance in the context of Francis turbines showing how different parameters like strain, vibration, pressure, and acoustic emission can be measured in real-time to understand the condition of Francis turbines. Since every fault occurring in a Francis turbine has its own characteristic signature, it can be detected and diagnosed before any mishap occurs. Any deviation in the monitored signal from the reference dataset could be an indicator of fault. Appropriate filters are used to remove any noise or disturbances in the signal coming from sensors. Preprocessed signals are observed in time, frequency, or time–frequency domain for the analysis. Machine-learning has also been applied in several studies for the predictive maintenance. Although predictive maintenance can be an effective method to optimize the cost of operation and maintenance of power plants, only a few cases of its application can be found in hydropower. Moreover, there are very few reviews available on the condition monitoring of Francis turbine. This study presents a comprehensive description on various aspects of the condition monitoring, with strong emphasis on the application of online condition monitoring.