This project presents the application of two methods allowing the extrapolation of short stationary load samples of strain measured in situ on hydraulic turbine blades. The extreme value theory and the concept of cyclostationarity are the theoretical basis for both proposed methods. The purpose of these techniques is to generate coherent signals representative of the long term expected load history. Such load history could then be used as an input for fatigue behaviour estimation of the turbine. The simulated long term data generates a wider range of fluctuations based on observed strains but of limited length.
The important criteria identified for the modeling are the determination of an appropriate threshold for extreme values and the selection of the stochastic model parameters of the random component which has been obtained by cyclostationarity.
The two proposed approaches can be applied to various operating conditions and demonstrated satisfactory results for speed no load, partial load and full load operating conditions. Hence, it is essential that the reference signal is recorded for a period long enough to contain sufficient information for its modeling. Also, this signal should not have important anomalies. On that point, it is important to analyse in detail the characteristics of the random part of the signal as they significantly affect the simulation methods.
| Date | 6 Dec 2013 |
|---|
| Original language | French |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Antoine Tahan (Supervisor) & André Coutu (Co-supervisor) |
|---|
Poirier, M. (Author),
Tahan (Supervisor) & Coutu (Co-supervisor),
6 Dec 2013Student thesis: Master's thesis › Master in Engineering: Mechanical Engineering