Resumen de Contribucion a la deteccion de frecuencias propias laterales de rotores de turbomaquinas hidraulicas
In hydraulic turbo machines of new design, the excitation forces usually have significant amplitudes. In such machines the study of the dynamic behaviour is a must to avoid excessive vibrations and cracks. So, apart from the excitations, the dynamic response of the rotor and the runner has to be determined as well.
Regarding the rotor response, the most important is to study the modal behaviour of its lateral natural frequencies which are the easiest to be excited.
The rotor dynamic modelling is possible by means of a finite element numerical model (FEM).
Nevertheless, these models are very complex and they show too many uncertainties except if they have been carefully calibrated with experimental data.
The current thesis presents a research work about the methodology to detect natural frequencies in large reaction type hydraulic turbines during the transients by the use of time-frequency transforms. In these turbines, the shaft is not easy to access and the impact hammer excitation method is very difficult to apply also due to the large masses to excite. Conversely, during the start-up transients different excitations act on the rotor (hydraulic and electric forces) which are able to excite it since they are very intense with a short duration.
Initially, the classical time-frequency transforms were analysed to determine the most adequate for the current application. It was concluded that the Morlet wavelet shows the best sensitivity-resolution characteristics for the vibratory signals generated during the transients.
Later on, a theoretic-experimental investigation was carried out in a prototype of a Pelton turbine in order to evaluate the validity of this wavelet in a real case. To obtain the natural frequencies and mode shapes of the machine a FEM model was developed and validated with an experimental modal analysis on the machine. This model allows to know the modal shapes of the first rotor lateral natural frequencies and to analyse its dynamic behaviour. After analyzing this model, an experimental investigation of the star-up transients was carried out with the Morlet wavelet to determine the feasibility of natural frequencies detection. The results obtained were compared with those from conventional methods. It is concluded that the time-frequency methodology permits to detect all the natural frequencies of the machine which are calculated with as much or even higher precision than with conventional modal analysis based on impacts or periodic excitation.
For this Pelton turbine it is also demonstrated that the excitation provoked by the impact of the water jet on the buckets can excite adequately the rotor. Therefore, this could be used to determine the first natural frequencies of any machine of this type.
Once the methodology had been validated, a research work was carried out on the two most representative types of reaction turbines which are the Francis and the Kaplan. Reaction turbines are machines with low rotating speed and with vertical rigid shaft. In Francis turbines the periodic excitations are very low (except in low specific speed designs) but they suffer from an important stochastic excitation due to the turbulence. For this research, the results from numerical models, classical experimental methods and wavelets are compared.
It has been proved that by the application of the wavelet transforms on the induced vibrations all the natural frequencies are detected. It can be concluded that the main excitation mechanisms are the initial impulse, part of the run up and the connection to the grid. The forces of hydraulic origin excite the modes with displacement in the turbine and it is observed that the detection of each mode depends on the modal displacement on each measuring point.
Kaplan turbines show slightly different conditions during the start-up in comparison with the machines previously considered. For them, the broad-band excitations prevail during all the star-up process. In this case the proposed method also permits to identify all the rotor natural frequencies.
Finally, it has been demonstrated that it is possible to detect the turbine natural frequencies during operation by the use of impacts and wavelets. This study permits to calculate the variation of the frequency values caused by the operating conditions, which is very useful to understand the rotor dynamic behaviour of any turbine.
