In compression systems, the stable operating range is limited by rotating stall and/or surge. Two distinct types of stall precursors can be observed prior to full scale instability: the development of long-wavelength modal waves or a short-wavelength, three-dimensional flow breakdown (so-called “spike” stall inception). The cause of the latter is not well understood; in axial machines it has been suggested that rotor blade-tip leakage flow plays an important role, but spikes have recently been observed in shrouded vaned diffusers of centrifugal compressors where these leakage flows are not present, suggesting an alternative mechanism may be at play. This paper investigates the onset of instability in a shrouded vaned diffuser from a highly loaded turbocharger centrifugal compressor and discusses the mechanisms thought to be responsible for the development of short-wavelength stall precursors. The approach combines unsteady 3D RANS simulations of an isolated vaned diffuser with previously obtained experimental results. The unsteady flow field simulation begins at the impeller exit radius, where flow is specified by a spanwise profile of flow angle and stagnation properties, derived from single-passage stage calculations but with flow pitchwise mixed. Through comparison with performance data from previous experiments and unsteady full-wheel simulations, it is shown that the diffuser is accurately matched to the impeller and the relevant flow features are well captured. Numerical forced response experiments are carried out to determine the diffuser dynamic behavior and point of instability onset. The unsteady simulations demonstrate the growth of short-wavelength precursors; the flow coefficient at which these occur, the rotation rate and circumferential extent agree with experimental measurements. Although the computational setup and domain limitations do not allow simulation of the fully developed spike nor full-scale instability, the model is sufficient to capture the onset of instability and allows the postulation of the following necessary conditions: (i) flow separation at the diffuser vane leading edge near the shroud endwall; (ii) radially reversed flow allowing vorticity shed from the leading edge to convect back into the vaneless space; and (iii) recirculation and accumulation of low stagnation pressure fluid in the vaneless space, increasing diffuser inlet blockage and leading to instability. Similarity exists with axial machines, where blade-tip leakage sets up endwall flow in the circumferential direction leading to flow breakdown and the inception of rotating stall. Rather than the tip leakage flows, the cause for circumferential endwall flow in the vaned diffuser is the combination of high swirl and the highly nonuniform spanwise flow profile at the impeller exit.

In compression systems the stable operating range is limited by rotating stall and/or surge. Two distinct types of stall precursors can be observed prior to full scale instability: the development of long-wavelength modal waves or a short-wavelength, three-dimensional flow breakdown (so-called “spike” stall inception). The cause of the latter is not well understood; in axial machines it has been suggested that rotor blade-tip leakage flow plays an important role, but spikes have recently been observed in shrouded vaned diffusers of centrifugal compressors where these leakage flows are not present, suggesting an alternative mechanism may be at play. This paper investigates the onset of instability in a shrouded vaned diffuser from a highly loaded turbocharger centrifugal compressor and discusses the mechanisms thought to be responsible for the development of short-wavelength stall precursors. The approach combines unsteady 3D RANS simulations of an isolated vaned diffuser with previously obtained experimental results. The unsteady flow field simulation begins at the impeller exit radius, where flow is specified by a spanwise profile of flow angle and stagnation properties, derived from single-passage stage calculations but with flow pitchwise mixed. Through comparison with performance data from previous experiments and unsteady full-wheel simulations, it is shown that the diffuser is accurately matched to the impeller and the relevant flow features are well captured. Numerical forced response experiments are carried out to determine the diffuser dynamic behavior and point of instability onset. The unsteady simulations demonstrate the growth of short-wavelength precursors; the flow coefficient at which these occur, the rotation rate and circumferential extent agree with experimental measurements. Although the computational setup and domain limitations do not allow simulation of the fully developed spike nor full-scale instability, the model is sufficient to capture the onset of instability and allows the postulation of the following necessary conditions: (i) flow separation at the diffuser vane leading edge near the shroud endwall; (ii) radially reversed flow allowing vorticity shed from the leading edge to convect back into the vaneless space; and (iii) recirculation and accumulation of low stagnation pressure fluid in the vaneless space, increasing diffuser inlet blockage and leading to instability. Similarity exists with axial machines, where blade-tip leakage sets up endwall flow in the circumferential direction leading to flow breakdown and the inception of rotating stall. Rather than the tip leakage flows, the cause for circumferential endwall flow in the vaned diffuser is the combination of high swirl and the highly non-uniform spanwise flow profile at the impeller exit.

Highly-loaded impellers, typically used in turbocharger and gas turbine applications, exhaust an unsteady, transonic flow that is non-uniform across the span and pitch and swirling at angles approaching tangential. With the exception of the flow angle, conflicting data exist regarding whether these attributes have substantial influence on the performance of the downstream diffuser. This paper quantifies the relative importance of the flow angle, Mach number, non-uniformity and unsteadiness on diffuser performance, through diffuser experiments in a compressor stage and in a rotating swirling flow test rig. This is combined with steady and unsteady Reynolds-Averaged Navier Stokes computations. The test article is a pressure ratio 5 turbocharger compressor with an airfoil vaned diffuser. The swirling flow rig is able to generate rotor outflow conditions representative of the compressor except for the periodic pitchwise unsteadiness, and fits a 0.86 scale diffuser and volute. In both rigs, the time-mean impeller outflow is mapped across a diffuser pitch using miniaturized traversing probes developed for the purpose.Across approximately two-thirds of the stage operating range, diffuser performance is well correlated to the average impeller outflow angle when the metric used is effectiveness, which describes the pressure recovery obtained relative to the maximum possible given the average inflow angle, Mach number and the vane exit metal angle. Utilizing effectiveness captures density changes through the diffuser at higher Mach numbers; a 10% increase in pressure recovery is observed as the inlet Mach number is increased from 0.5 to 1. Further, effectiveness is shown to be largely independent of the timeaveraged spanwise and unsteady pitchwise non-uniformity from the rotor; this independence is reflective of the strong mixing processes that occur in the diffuser inlet region. The observed exception is for operating points with high timeaveraged vane incidence. Here, it is hypothesized that temporary excursions into high-loss flow regimes cause a nonlinear increase in loss as large unsteady angle variations pass by from the rotor.Given that straight-channel diffuser design charts typically used in preliminary radial vaned diffuser design capture neither streamtube area changes from impeller exit to the diffuser throat nor vane incidence effects, their utility is limited. An alternative approach, utilizing effectiveness and vane leading edge incidence, is proposed. INTRODUCTIONModern high pressure ratio centrifugal compressors typically utilize a high speed impeller with backswept blades and a vaned diffuser. As part of ongoing efforts for a systemwide reduction in weight, increase in efficiency and/or reduction of emissions, designers seek to improve the pressure ratio, operating range and efficiency of the centrifugal compressor, which is often limited by the diffuser. Despite the relative simplicity of its geometry, the flow in the diffuser is complex and there is no well-established, reliable approach to gui...

Highly-loaded impellers, typically used in turbocharger and gas turbine applications, exhaust an unsteady, transonic flow that is non-uniform across the span and pitch and swirling at angles approaching tangential. With the exception of the flow angle, conflicting data exist regarding whether these attributes have substantial influence on the performance of the downstream diffuser. This paper quantifies the relative importance of the flow angle, Mach number, non-uniformity and unsteadiness on diffuser performance, through diffuser experiments in a compressor stage and in a rotating swirling flow test rig. This is combined with steady and unsteady Reynolds-Averaged Navier Stokes computations. The test article is a pressure ratio 5 turbocharger compressor with an airfoil vaned diffuser. The swirling flow rig is able to generate rotor outflow conditions representative of the compressor except for the periodic pitchwise unsteadiness, and fits a 0.86 scale diffuser and volute. In both rigs, the time-mean impeller outflow is mapped across a diffuser pitch using miniaturized traversing probes developed for the purpose.Across approximately two-thirds of the stage operating range, diffuser performance is well correlated to the average impeller outflow angle when the metric used is effectiveness, which describes the pressure recovery obtained relative to the maximum possible given the average inflow angle, Mach number and the vane exit metal angle. Utilizing effectiveness captures density changes through the diffuser at higher Mach numbers; a 10% increase in pressure recovery is observed as the inlet Mach number is increased from 0.5 to 1. Further, effectiveness is shown to be largely independent of the timeaveraged spanwise and unsteady pitchwise non-uniformity from the rotor; this independence is reflective of the strong mixing processes that occur in the diffuser inlet region. The observed exception is for operating points with high timeaveraged vane incidence. Here, it is hypothesized that temporary excursions into high-loss flow regimes cause a nonlinear increase in loss as large unsteady angle variations pass by from the rotor.Given that straight-channel diffuser design charts typically used in preliminary radial vaned diffuser design capture neither streamtube area changes from impeller exit to the diffuser throat nor vane incidence effects, their utility is limited. An alternative approach, utilizing effectiveness and vane leading edge incidence, is proposed. INTRODUCTIONModern high pressure ratio centrifugal compressors typically utilize a high speed impeller with backswept blades and a vaned diffuser. As part of ongoing efforts for a systemwide reduction in weight, increase in efficiency and/or reduction of emissions, designers seek to improve the pressure ratio, operating range and efficiency of the centrifugal compressor, which is often limited by the diffuser. Despite the relative simplicity of its geometry, the flow in the diffuser is complex and there is no well-established, reliable approach to gui...