In large modern turbochargers, compressors often constitute the main source of noise, with a frequency spectrum typically dominated by tonal noise at the blade passing frequency (BPF) and its harmonics. In transonic operation, inflow BPF noise is mainly generated by rotor locked shock fronts. These and the resulting acoustic fields can be predicted numerically with reasonable accuracy. Outflow noise, while also dominated by BPF tones, is linked to more complex source mechanisms. Its modal structure and the relationships between sources and modal sound pressure levels (SPL) are less well understood. Perhaps this is linked to the intrinsically non-axisymmetric geometries, which results in the need for full stage simulations if high accuracy is of paramount importance.
In order to shed some light on outflow noise generation, a transient simulation of a 360° model of a radial compressor stage, including a vaned diffuser and a volute, was carried out using state-of-the-art CFD. Additionally, experimental data was gathered at a multitude of data points downstream of the volute exit for post processing and modal analysis. The sources and the propagation were calculated directly. Optimized values for tempo-spatial acoustic wave resolution and buffer layer design were chosen, based on extensive studies on simplified models. Two grid refinement levels were used to check grid convergence and time step size independence of the results was ensured.
Numerical and experimental data match within 1% for total pressure ratio, volume flow and exit total temperature for the studied operating point. Both show the same modal content at the 1st BPF and indicate the presence of the same single dominating mode. The numerical results underpredict overall sound power levels (PWL) at the 1st BPF by 6.6dB. This difference is expected to decrease with further grid refinement and improved accounting for numerical damping. At the 2nd BPF, the experimental data show a significant broadening of the modal content with homogeneous modal PWL distributions. The multitude of modes leads to the generation of complex interference patterns, which shows that single-point acoustic measurements are often inadequate for component noise qualification and should be substituted by modal techniques. The dominating dipole sound sources are found in narrow areas around the vane leading edges and the rotor blade trailing edges. Because of the non-axisymmetric geometry, vane dipole source strengths become a function of circumferential position. The unsteady shedding of vortices from the vane suction surfaces is identified as a further possible source mechanism. However, the contributions of structural vibrations and mode scattering due to small manufacturing imperfections remain unclear.