Analyse modale opérationnelle des roues de turbines hydroélectriques par l’étude de régimes transitoires

Abstract

Since the early 1990’s, the use of hydroelectric turbine has significantly changed to meet the power grid requirements. The energy consumption intermittency and the instability of certain sources of energy have prompted users to operate turbines in a way that balances the grid. Indeed, turbines are particularly well suited for such a task, as they can easily modulate their production and are effortless to start or stop. This new use compels hydroelectric turbines to operate across a wide operating range often away from their best efficient point, and to withstand non-stationary loads during transients. Consequently, the hydroelectric turbine ability to play the role of power buffers must be weighed against early damages and failures. The fatigue damaging process can come with a numerical reliability loss, due to the difficulty of numerical computations to simulate transient and part load regimes. To compensate for the loss of numerical performances, one recent idea in to take benefit from experimental measurements conducted during turbine commissioning or refurbishments. Such data contains information about the real behavior of the device. Properly handled, signal processing should be able to extract runner mechanical characteristic estimates in some specific regimes. At HydroQuébec, this research area has been investigated for years in two wide research projects : Damage Prediction and Integrated Diagnosis of Turbines (PREDDIT), and Hydroelectric Turbines Suited for their Real Use (THAUR). In collaboration with IREQ and Andritz, the research introduced in this thesis is a continuation of these projects. The scope of the work is to investigate data recorded during transient regimes, particularly harmful for the structure. The main objective is to estimate the structure’s modal response, in order to extract modal parameters and related uncertainties. The methodology relies on the characterization of synchronous resonances induced by the harmonics of the rotating speed. Such resonances are extracted using advanced order tracking algorithms, then characterized using operational modal analysis. It appears that Bayesian identification tools are well suited to the processing of our data. Also, a discussion was made about the use of identification results in industrial contexts, to obtain more accurate hybrid models. Results were used to validate numerical models, and for calibrating finite element routines, in order to predict more accurate critical speeds. The final goal is to make better load level assessments, in order to improve the runner fatigue health predictions.

Date	15 Dec 2021
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Antoine Tahan (Supervisor) & Jérôme Antoni (Co-supervisor)