Validation of a Forced Response Prediction Method for Design of Axial Compressors
Konferensbidrag (offentliggjort, men ej förlagsutgivet), 2015
The linearized time-harmonic Navier-Stokes solver LINNEA is used to evaluate the forcing of a rotor blade in the transonic compressor Hulda, at part speed. Hulda is a 1½ stage demonstrator compressor consisting of a variable inlet guide vanes (VIGV), rotors, stators and outlet guide vanes (OGV). The forcing of the rotor is evaluated for different setting angles of the VIGV. The mesh study shows that, for this computational setup, a mesh independent solution was obtained in RANS simulation using CFX and in the linearized Navier-Stokes solver LINNEA for the same grid resolution.
The results obtained using LINNEA are compared to results from three URANS simulations using CFX. The first URANS case uses scaled geometries and rotational periodic boundary conditions. The second and third case use the true geometry in conjunction with rotational periodic boundary conditions: “profile transformation” and with time-lagged periodic boundary conditions: “time transformation” in CFX’ nomenclature, which corresponds to time-inclined computational planes. LINNEA is run with two different boundary conditions at the inlet to the rotor domain, obtained from a RANS and URANS simulation, respectively.
The forcing is evaluated at engine orders 15, 30 and 45, corresponding to VIGV passing frequency and multiples thereof. The results show that the amplitude of the tangential force variation calculated with the different methods and boundary conditions, is within a 1.3% interval based on the total tangential force for the nominal (0 degree) VIGV setting and within a 3.3% interval for the VIGV setting angle of 30 degrees at engine order 15.
Using a linear time-harmonic Navier-Stokes solver to calculate forcing of a rotor blade interacting an upstream component is shown to be a possible and attractive alternative to URANS calculations. A linear harmonic simulation is less computationally demanding than a URANS simulation. The mean flow solution required by the linearized solver does not incur any additional computational effort since it is usually already available from general performance calculations. Further is shown that for URANS analyses a scaling of the model can be neglected since results obtained for a scaled model are in good agreement with results obtained for an unscaled model using profile transformation (PT) to handle the variation of pitch between components.