Because of the relative movement of the rotor's blades in relation to those of the stator, the flow is very unsteady. To reduce the size of the machines, the axial distances between successive rotors and stators have to be reduced, which tends to intensify the unsteady effects. It is essential that we understand these phenomena in order to master the design of machines. Continuous improvements in computing power, combined with the development of numerical methods, have made the calculation of 3D unsteady flows accessible. Thus, the Turbomachinery team has suitable tools for simulating, analyzing and better understanding the complex mechanisms that come into play in turbomachinery.
The ADTurB project
It is still difficult to simulate the flow in all the channels of a multi-stage machine because of the computing resources necessary. However, the flow's spatial-temporal periodicity properties can be used to design solutions for getting over this difficulty. Thus, it is possible to only simulate one channel of each rotor (chorochronic calculation) or only a part of the channels of the fixed and mobile rotors (RNB technique - Reduction of the Number of Blades). Within the framework of the ADTurB project, which aims at improving the prediction of the aeroelastic phenomena in axial turbines, 3D [video] unsteady simulations were performed and then compared to the measurements provided by the DLR (fig. 1).
Figure 1 : relative Mach number field against time
(above: RNA computation, bottom: LDV measurements)
Processing of unsteady signals
The unsteady phenomena that take place in turbomachines are not always periodic in time, notably in the transient states (start-up, stopping) or the entry into unsteady states. The use of a Fourier Transform to analyze the signals is then no longer sufficient and the implementation of better adapted methods is essential. Under these conditions, the time/frequency representations (wavelets, sliding window Fourier) are techniques that provide more complete information and give a better understanding of the flow's structure (figure 2).
Figure 2 : Time/frequency plot of an unstedy signal of pressure
(computation on the sur CME2-Lemfi compressor bench)
