A better understanding of unsteady flow phenomena encountered in rotor-stator interactions is a key to further improvements in turbomachinery. Besides CFD methods yielding 3D flow field predictions, time-resolving measurement techniques are necessary to determine the instantaneous flow quantities of interest. Fast-response aerodynamic probes are a promising alternative to other time-resolving measurement techniques such as hot-wire anemometry or laser anemometry. This contribution gives an overview of the fast-response probe measurement technique, with the emphasis on the total system and its components, the development methods, the operation of such systems and the data processing requirements. A thorough optimization of all system components (such as sensor selection and packaging, probe tip construction, probe aerodynamics and data analysis) is the key of successful development. After description of the technique, examples of applications are given to illustrate its potential. Some remarks will refer to recent experiences gained by the development and application of the ETH FRAP ® system.

In turbocharger applications bleed air near the impeller exit is often used for secondary flow systems to seal bearing compartments and to balance the thrust load on the bearings. There is experimental evidence that the performance and operability of highly loaded centrifugal compressor designs can be sensitive to the amount of bleed air. To investigate the underlying mechanisms and to assess the impact of bleed air on the compressor dynamic behavior, a research program was carried out on a pre-production, 5.0 pressure ratio, high-speed centrifugal compressor stage of advanced design. The investigations showed that bleed air can significantly reduce the stable flow range. Compressor rig experiments, using an array of unsteady pressure sensors and a bleed valve to simulate a typical turbocharger environment, suggest that the path into compression system instability is altered by the bleed flow. Without bleed flow, the pre-stall behavior is dominated by short wavelength disturbances, or so called ‘spikes’, in the vaneless space between the impeller and the vaned diffuser. Introducing bleed flow at impeller exit reduces endwall blockage in the vaneless space and destabilizes the highly-loaded vaned diffuser. The impact is a 50% reduction in stable operating range. The altered diffuser characteristic reduces the compression system damping responsible for long wavelength, modal pre-stall behavior. A four-lobed backward traveling rotating stall wave is experimentally measured, in agreement with calculations obtained from a previously developed dynamic compressor model. In addition, a self-contained, endwall blockage control strategy was employed, successfully recovering 75% of the loss in surge-margin due to bleed flow and yielding a 1 point increase in adiabiatic compressor efficiency.

Experimental investigations on a single stage centrifugal compressor showed that measured blade vibration amplitudes vary considerably along a constant speed line from choke to surge. The unsteady flow has been analyzed to obtain detailed insight into the excitation mechanism. Therefore, a turbocharger compressor stage impeller has been modeled and simulated by means of computational fluid dynamics (CFD). Two operating points at off-design conditions were analyzed. One was close to choke and the second one close to the surge line. Transient CFD was employed, since only then a meaningful prediction of the blade excitation, caused by the unsteady flow situation, can be expected. Actually, it was observed that close to surge a steady state solution could not be obtained; only transient CFD could deliver a converged solution. The CFD results show the effect of the interaction between the inducer casing bleed system and the main flow. Additionally, the effect of the nonaxisymmetric components, such as the suction elbow and the discharge volute, was analyzed. The volute geometry itself had not been modeled. It turned out to be sufficient to impose a circumferentially asymmetric pressure distribution at the exit of the vaned diffuser to simulate the volute. Volute and suction elbow impose a circumferentially asymmetric flow field, which induces blade excitation. To understand the excitation mechanism, which causes the measured vibration behavior of the impeller, the time dependent pressure distribution on the impeller blades was transformed into the frequency domain by Fourier decomposition. The complex modal pressure data were imposed on the structure that was modeled by finite element methods (FEM). Following state-of-the-art calculations to analyze the free vibration behavior of the impeller, forced response calculations were carried out. Comparisons with the experimental results demonstrate that this employed methodology is capable of predicting the impeller’s vibration behavior under real engine conditions. Integrating the procedure into the design of centrifugal compressors will enhance the quality of the design process.

In turbocharger applications, bleed air near the impeller exit is often used for secondary flow systems to seal bearing compartments and to balance the thrust load on the bearings. There is experimental evidence that the performance and operability of highly-loaded centrifugal compressor designs can be sensitive to the amount of bleed air. To investigate the underlying mechanisms and to assess the impact of bleed air on the compressor dynamic behavior, a research program was carried out on a preproduction, 5.0 pressure ratio, high-speed centrifugal compressor stage of advanced design. The investigations showed that bleed air can significantly reduce the stable flow range. Compressor rig experiments, using an array of unsteady pressure sensors and a bleed valve to simulate a typical turbocharger environment, suggest that the path into compression system instability is altered by the bleed flow. Without the bleed flow, the prestall behavior is dominated by short-wavelength disturbances, or so called “spikes,” in the vaneless space between the impeller and the vaned diffuser. Introducing bleed flow at the impeller exit reduces endwall blockage in the vaneless space and destabilizes the highly-loaded vaned diffuser. The impact is a 50% reduction in stable operating range. The altered diffuser characteristic reduces the compression system damping responsible for long-wavelength modal prestall behavior. A four-lobed backward traveling rotating stall wave is experimentally measured in agreement with calculations obtained from a previously developed dynamic compressor model. In addition, a self-contained endwall blockage control strategy was employed, successfully recovering 75% of the loss in surge-margin due to the bleed flow and yielding a one point increase in adiabatic compressor efficiency.

In this paper the three-dimensional inverse design code TURBOdesign-1 is applied to the design of the blade geometry of a centrifugal compressor impeller with splitter blades. In the design of conventional impellers the splitter blades normally have the same geometry as the full blades and are placed at mid-pitch location between the two full blades, which can usually result in a mismatch between the flow angle and blade angles at the splitter leading edge. In the inverse design method the splitter and full blade geometry is computed independently for a specified distribution of blade loading on the splitter and full blades. In this paper the basic design methodology is outlined and then the flow in the conventional and inverse designed impeller is compared in detail by using computational fluid dynamics (CFD) code TASCflow. The CFD results confirm that the inverse design impeller has a more uniform exit flow, better control of tip leakage flow and higher efficiency than the conventional impeller. The results also show that the shape of the trailing edge geometry has a very appreciable effect on the impeller Euler head and this must be accurately modeled in all CFD computations to ensure closer match between CFD and experimental results. Detailed measurements are presented in part II of the paper.