To cope with the intermittent power supply of the new renewable energies and demand fluctuations, Francis turbines are required to operate more and more in an extended operating range, far away from the design point. With this operating behavior, it is very complex to interpret the trend of vibration parameters typically used in Condition Monitoring and to define reasonable alarm and trip levels valid for all the operating range of the unit working in steady conditions. As in the efficiency curves of Francis turbines represented as a function of net head and load (Hill Chart), in this paper we propose to represent the most relevant vibration parameters in surfaces, called Vibrational Hill Charts, which allow a more accurate evaluation of the indicators and their trends and a better classification of abnormal values. To show the potential of Vibrational Hill Charts, a complete database obtained after 2 years of monitoring a large Francis Unit (444 MW rated power) has been used. The mapping of the relevant vibration parameters has been performed by means of Artificial Neural Networks. It is shown that by setting the action levels based on the resulting maps, rather than a constant value, a better diagnosis capacity is achieved as the Receiver Operating Characteristic will be improved. Furthermore, phenomena such as erosive cavitation, which is hard to be detected, could be also assessed with the use of multidimensional analysis based on the Vibrational Hill Chart. As a conclusion, with the Vibrational Hill Chart, the condition monitoring and diagnosis of hydraulic turbines could be improved.
