Numerical Investigation of IGV–Rotor Flutter of a Modern Transonic Compressor: On the Assessment of the Nonlinear Harmonic Method

In the present work, two components of a modern transonic compressor are studied for flutter risk. Blade flutter is a significant problem for manufacturers of turbomachinery since it often leads to catastrophic structural failure and is one of the most dramatic physical phenomena in the field of aeroelasticity. This way, it is of paramount importance to have accurate and fast predictions of the flutter risk, enabling more efficient design and analysis processes. Thus, the flutter risk of these components is studied considering the non-linear harmonic (NLH) method. In the NLH method, the flow variables are decomposed into time-mean value and a sum of periodic perturbations — which are then Fourier decomposed in time into several harmonics — and the URANS equations are cast into the frequency domain. This way the problem is simplified to a time-mean flow equation and several transport equations for harmonics, allowing the capture of the unsteady flow physics at an affordable computing cost. Furthermore, these components are also considered for the evaluation of the NLH method capabilities in comparison with full unsteady Reynolds-averaged Navier–Stokes (URANS) simulations. This evaluation is conducted across the entire operating range of the engine (including off-design points). This comparison is important for validating the NLH method against trusted standards, particularly at several off-design points (where the complexity of the flow field is increased) showing if NLH can provide fast and accurate results but also to identify any limitations or areas where the method does not perform well, guiding its appropriate use in industry and research. This assessment is conducted considering single component and stage computations. Overall, it was found that both components are flutter-free across the entire operating range and that the NLH method is able to provide accurate and fast predictions (up to 40X), under some conditions. More concretely, large variations in the stability of the components are verified across the operating lines. The differences between both methods are dependent on the mode shape, on the operating line, on the mass flow, and on the maximum displacement imposed. While the mode shapes with a small torsional component reveal negligible differences across the entire operating map, when the torsional component is significant, these differences are more sensitive to the mass flow rate and can reach up to 9%. For the combination of torsional mode shapes with a small mass flow rate, attention should be paid to the maximum displacement imposed for the NLH simulations to avoid unphysical predictions. In this comparison, the NLH simulations were conducted considering only one harmonic of the mode shape, illustrating the viability of this method for flutter predictions. Increasing the number of harmonics (up to 3) was verified to not significantly influence the predictions by increasing substantially the computational cost. Additionally, the stage effects on the blades’ aerodynamic damping are only relevant when shocks travel across the rows.