0

IN THIS VOLUME


Turbomachinery

1990;():V001T01A002. doi:10.1115/90-GT-011.
FREE TO VIEW

This paper describes two new developments in the numerical analysis of linearized unsteady cascade flows, that have been motivated by the need for an accurate analytical procedure for predicting the onset of flutter in highly loaded compressors. In previous work, results were determined using a two-step or single-pass procedure in which a solution was first determined on a rectilinear-type cascade mesh to determine the unsteady flow over an extended blade-passage solution domain and then on a polar-type local mesh to resolve the unsteady flow in high-gradient regions. In the present effort a composite procedure has been developed in which the cascade- and local-mesh equations are solved simultaneously. This allows the detailed features of the flow within the local-mesh region to impact the unsteady solution over the entire domain. In addition, a new transfinite local mesh has been introduced to permit a more accurate modeling of unsteady shock phenomena. Numerical results are presented for a two-dimensional compressor-type cascade operating at high subsonic inlet Mach number and high mean incidence to demonstrate the impact of the new composite- and local-mesh analyses on unsteady flow predictions.

Commentary by Dr. Valentin Fuster
1990;():V001T01A003. doi:10.1115/90-GT-012.
FREE TO VIEW

The general theory of three-dimensional flow in subsonic and supersonic turbomachines (Wu, 1952) is extended to the three-dimensional rotational flow in transonic turbomachines. In Part I of the paper, an approximation that the S1 stream filaments are filaments of revolution is made. Then, the three-dimensional solution is obtained by an iterative solution between a number of S1 stream filaments and a single S2 stream filament. Recently developed relatively simple and quick method of solving the transonic S1 flow is utilized. The complete procedure is illustrated with the solution of the 3D flow in the DFVLR rotor operating at the design point. The solution is presented in detail, special emphasis being placed on the fulfillment of the convergence requirement. The character of the three-dimensional field obtained is examined with the three-dimensional structure of the passage shock, the relative Mach number contours on a number of S1 surfaces, S2 surfaces, and cross surfaces, and the variations of the thickness of S1 and S2 filaments. Comparison between the calculated three-dimensional field with the DFVLR measured data shows: the character of the flow field and the streamwise variation of the flow variables in the middle of the flow channel are in good agreement. It is recommended that the method presented herein can be used for three-dimensional design of transonic turbomachines.

Commentary by Dr. Valentin Fuster
1990;():V001T01A004. doi:10.1115/90-GT-013.
FREE TO VIEW

The general theory for three–dimensional flow in subsonic and supersonic turbomachines has recently been extended to transonic turbomachines. In Part II of the paper, quasi– and full three–dimensional solutions of the transonic flow in the CAS rotor are presented. The solutions are obtained by iterative calculation between a number of S1 stream filaments and, respectively, a central S2 stream filament and a number of S2m stream filaments. Relatively simple methods developed recently for solving the transonic flow along S1 and S2 stream filaments are used in the calculation.

The three–dimensional flow fields in the CAS rotor obtained by the present method are presented in detail with special emphasis on the converging process for the configuration of the S1 and S2 stream filaments. The three–dimensional flow fields obtained in the quasi– and full 3D solutions are quite similar, but the former gives a lower peak Mach number and a smaller circumferential variation in Mach number than the latter. A comparison between the theoretical solution and the Laser–2–Focus measurement shows that the character of the transonic flow including the 3D shock structure is in good agreement, but the measured velocity is slightly higher than the calculated one over most of the flow field.

Commentary by Dr. Valentin Fuster
1990;():V001T01A005. doi:10.1115/90-GT-014.
FREE TO VIEW

The development of an analysis to predict the unsteady compressible flows in blade boundary layers and wakes is presented. The equations that govern the flows in these regions are transformed using an unsteady turbulent generalization of the Levy-Lees transformation. The transformed equations are solved using a finite difference technique in which the solution proceeds by marching in time and in the streamwise direction. Both laminar and turbulent flows are studied, the latter using algebraic turbulence and transition models. Laminar solutions for a flat plate are shown to approach classical asymptotic results for both high and low frequency unsteady motions. Turbulent flat-plate results are in qualitative agreement with previous predictions and measurements. Finally, the numerical technique is also applied to the stator and rotor of a low-speed turbine stage to determine unsteady effects on surface heating. The results compare reasonably well with measured heat transfer data and indicate that nonlinear effects have minimal impact on the mean and unsteady components of the flow.

Commentary by Dr. Valentin Fuster
1990;():V001T01A006. doi:10.1115/90-GT-015.
FREE TO VIEW

An inviscid-viscous interaction technique for the analysis of quasi-three-dimensional turbomachinery cascades has been developed. The inviscid flow is calculated using a time-marching, multiple-grid Euler analysis. An inverse, finite-difference viscous-layer analysis, which includes the wake, is employed so that boundary-layer separation can be modeled. This analysis has been used to predict the performance of a transonic compressor cascade over the entire incidence range. The results of the numerical investigation in the form of cascade total pressure loss, exit gas angle and blade pressure distributions are compared with existing experimental data and Navier-Stokes solutions for this cascade, and show that this inviscid-viscous interaction procedure is able to accurately predict cascade loss and airfoil pressure distributions. Several other aspects of the present interaction analysis are examined, including transition and wake modeling, through comparisons with data.

Commentary by Dr. Valentin Fuster
1990;():V001T01A007. doi:10.1115/90-GT-017.
FREE TO VIEW

A rotating cavity with an axial throughflow of cooling air is used to provide a simplified model for the flow that occurs between adjacent corotating compressor discs inside a gas-turbine engine. Flow visualization and laser-Doppler anemometry are employed to study the flow structure inside isothermal and heated rotating cavities for a wide range of axial-gap ratios. G. rotational. Reynolds numbers, Reφ, axial Reynolds numbers, Rez, and temperature distributions.

For the isothermal case, the superposed axial flow of air generates a powerful toroidal vortex inside cavities with large gap ratios (G > 0.400) and weak counter-rotating toroidal vortices for cavities with small gap ratios. Depending on the gap ratio and the Rossby number, ε (where ε ∝ Rez/Reφ), axisymmetric and nonaxisymmetric vortex breakdown can occur, but circulation inside the cavity becomes weaker as ε is reduced.

For the case where one or both discs of the cavity are heated, the flow becomes nonaxisymmetric: cold air enters the cavity in a “radial arm” on either side of which is a vortex. The cyclonic and anti-cyclonic circulations inside the two vortices are presumed to create the circumferential pressure gradient necessary for the air to enter the cavity (in the radial arm) and to leave (in Ekman layers on the discs). The core of fluid between the Ekman layers precesses with an angular speed close to that of the discs, and vortex breakdown appears to reduce the relative speed of precession.

Topics: Cooling , Air flow , Cavities
Commentary by Dr. Valentin Fuster
1990;():V001T01A008. doi:10.1115/90-GT-018.
FREE TO VIEW

A methodology is presented for simulating turbomachinery blade rows in a multistage environment by deploying a standard 3D Navier-Stokes solver simultaneously on a number of blade rows. The principle assumptions are that the flow is steady relative to each blade row individually and that the rows can communicate via inter-row mixing planes. These mixing planes introduce circumferential averaging of flow properties but preserve quite general radial variations. Additionally, each blade can be simulated in 3D or axisymmetrically (in the spirit of throughflow analysis) and a series of axisymmetric rows can be considered together with one 3D row to provide, cheaply, a machine environment for that row.

Two applications are presented: a transonic compressor rotor and a steam turbine nozzle guide vane simulated both isolated and as part of a stage. In both cases the behaviour of the blade considered in isolation was different to when considered as part of a stage and in both cases was in much closer agreement with the experimental evidence.

Commentary by Dr. Valentin Fuster
1990;():V001T01A009. doi:10.1115/90-GT-019.
FREE TO VIEW

The extension of a well established three dimensional flow calculation method to calculate the flow through multiple turbomachinery blade rows is described in this paper. To avoid calculating the unsteady flow, which is inherent in any machine containing both rotating and stationary blade rows, a mixing process is modelled at a calculating station between adjacent blade rows. The effects of this mixing on the flow within the blade rows may be minimised by using extrapolated boundary conditions at the mixing plane.

Commentary by Dr. Valentin Fuster
1990;():V001T01A010. doi:10.1115/90-GT-020.
FREE TO VIEW

Measurements of the mean and turbulent flow field have been made in a cascade of high turning turbine rotor blades. The inlet turbulence was raised to 5% by a grid placed upstream of the cascade, and the secondary flow region was traversed within and downstream of the blades using a 5 hole probe and crossed hot-wires. Flow very close to the end wall was measured using a single wire placed at several orientations. Some frequency spectra of the turbulence were also obtained.

The results shows that the mean flow field is not affected greatly by the high inlet turbulence. The Reynolds stresses were found to be very high, particularly in the loss core. Assessment of the contributions to production of turbulence by the Reynolds stresses show that the normal stresses have significant affects as well as the shear stresses. The calculation of eddy viscosity from two independent shear stresses show it to be fairly isotropic in the loss core. Within the blade passage, the flow close to the end wall is highly skewed and exhibits generally high turbulence. The frequency spectra show no significant resonant peaks, except for one at very low frequency, attributable to an acoustic resonance.

Commentary by Dr. Valentin Fuster
1990;():V001T01A011. doi:10.1115/90-GT-021.
FREE TO VIEW

A measuring technique based on multisensor hot-wire anemometry has been developed to determine the unsteady three-dimensional velocity vector and the structure of turbulent flows. It then has been applied to the passage and the exit flow of an annular compressor cascade, which is periodically disturbed by the wakes of a cylinder rotor, located about 50 percent of blade chord upstream. In part I of this paper the decay of the rotor wakes will be described first without stator and secondly through a stator passage. The time-dependent turbulent flow field downstream of this stator is discussed in Part II.

The rotor wakes have a major influence on the development of three-dimensional separated regions inside the compressor cascade, and this interaction will be addressed in both parts of this paper.

Commentary by Dr. Valentin Fuster
1990;():V001T01A012. doi:10.1115/90-GT-022.
FREE TO VIEW

A measuring technique based on multi-sensor hot-wire anemometry has been developed to determine the unsteady three-dimensional velocity vector and the structure of turbulent flows. It then has been applied to the passage and the exit flow of an annular compressor cascade, which is periodically disturbed by the wakes of a cylinder rotor, located about 50 percent of blade chord upstream. In Part I of this paper the decay of the rotor wakes will be described first without stator and secondly through a stator passage. The time-dependent turbulent flow field downstream of this stator is discussed in Part II of this paper.

The rotor wakes have a major influence on the development of three-dimensional separated regions inside the compressor cascade, and this interaction will be addressed in both parts of this paper.

Commentary by Dr. Valentin Fuster
1990;():V001T01A013. doi:10.1115/90-GT-036.
FREE TO VIEW

The complex three-dimensional flow fields in a mixed-flow pump impeller are investigated by applying the incompressible version of the Dawes’ 3D Navier-Stokes code. The applicability of the code is confirmed by comparison of computations with a variety of experimentally measured jet-wake flow patterns and overall performances at four different tip clearances including the shrouded case. Based on the computations, the interaction mechanism of secondary flows and the formation of jet-wake flow are discussed. In the case of large tip clearances, the reverse flow caused by tip leakage flow is considered to be the reason for the thickening of the casing boundary layer followed by the deterioration of the whole flow field.

Commentary by Dr. Valentin Fuster
1990;():V001T01A014. doi:10.1115/90-GT-037.
FREE TO VIEW

For predicting the tip clearance loss of turbomachines, different equations are published in the literatures based on differnt principles. In 1986 the present author posturated a new theory where the pressure loss consisted of two parts, one was the pressure loss induced by the drag force of the leaked flow and the other was the pressure loss to support the axial pressure difference without blades in the tip clearance zone. There were comments such as the two losses were the same loss looked from two different view points, or at least a part of the former was included in the latter or vice versa.

In this paper the pressure loss due to the tip clearance is examined based on a macroscopic balance of forces and the two kinds of loss are derived. Furthermore, it is made clear that the former comes from the induced drag which is parallel to the blade while the latter comes from the missing blade force normal to the blade in the clearance zone. Because these two forces are mutually perpendicular, the two losses are entirely different in nature and they do not even partially overlap to each other. It is also made clear quantitatively, how the loss of the kinetic energy of leaked flow is related to the induced drag of the clearance flow.

Commentary by Dr. Valentin Fuster
1990;():V001T01A015. doi:10.1115/90-GT-038.
FREE TO VIEW

The mechanism of mixing in axial flow compressors has been investigated in two low speed machines. For reasons of length this is described in two parts. Results in a 4-stage compressor are described here in Part I and show that the mixing coefficients across the first and the third stators are of similar magnitude. Part I also describes the background and experimental facilities and techniques used in both parts together with the nomenclature and all the references. Part II describes the results from a large single stage compressor. It also presents measurements of mixing in a simple two-dimensional duct, and presents conclusions for the whole investigation.

Commentary by Dr. Valentin Fuster
1990;():V001T01A016. doi:10.1115/90-GT-039.
FREE TO VIEW

This paper follows directly from Part I which contains not only the description of the facilities and the results for the C106 four-stage compressor but also the background, list of nomenclature, acknowledgement and references. The discussion and conclusions for Part I and Part II are given here. The single stage compressor results show the significant effects of inlet guide vane (IGV) wakes on mixing across the stage in the so called ‘free stream’ region; in the casing region tip clearance flow is shown to play an important role in mixing. Explanations for these results are given. Investigations were also carried out in a two-dimensional rectangular duct flow to reveal the mixing mechanism in the corner region similar to those formed by blade surfaces and endwalls in a compressor. Turbulent diffusion has been found to be the dominant mechanism in spanwise mixing; anisotropic inhomogeneous turbulent diffusion is mainly responsible for the non-uniform mixing in the corner region. The larger spread of tracer gas in the tangential direction than in the radial direction is mainly caused by the wake dispersion and relative flow motions within the blade wakes as well as secondary flow contributions in the end-wall regions.

Commentary by Dr. Valentin Fuster
1990;():V001T01A017. doi:10.1115/90-GT-049.
FREE TO VIEW

The detailed three-dimensional velocity distributions, corresponding to design and off-design operation, were measured in two different circumferential cross sections of a volute by means of LDV. It is shown that the swirl has a forced vortex type velocity distribution and that the location of the swirl center is changing with mass flow. The through flow velocity distribution is primarily defined by the conservation of angular momentum. A strong interaction between the through flow and swirl velocity is observed.

Flow visualisation in the tongue region reveals a reversal of the velocity at the volute inlet with increasing mass flow. The pressure drop between volute outlet and inlet at low mass flow pushes extra fluid through the tongue gap and increases the mass flow in the volute. The abrupt pressure rise at high mass flow results in local return flow perturbing the flow in the outlet pipe.

Commentary by Dr. Valentin Fuster
1990;():V001T01A018. doi:10.1115/90-GT-050.
FREE TO VIEW

Two dimensional potential flow was used to determine the velocity field within a laboratory centrifugal pump. In particular, the finite element technique was used to model the impeller and volute simultaneously. The rotation of the impeller within the volute was simulated by using steady state solutions with the impeller in 10 different angular orientations. This allowed the interaction between the impeller and the volute to develop naturally as a result of the solution. The results for the complete pump model showed that there are circumferential asymmetries in the velocity field, even at the design flow rate. Differences in the relative velocity components were as large as 0.12 m/sec for the radial component and 0.38 m/sec for the tangential component, at the impeller exit. The magnitude of these variations was roughly 25% of the magnitude of the average radial and tangential velocities at the impeller exit. These asymmetries were even more pronounced at off design flow rates. The velocity field was also used to determine the location of the tongue stagnation point and to calculate the slip within the impeller. The stagnation point moved from the discharge side of the tongue to the impeller side of the tongue, as the flow rate increased from below design flow to above design flow. At design flow, values of slip ranged from 0.96 to 0.71, from impeller inlet to impeller exit. For all three types of data (velocity profiles, stagnation point location, and slip factor) comparison was made to laser velocimeter data, taken for the same pump. At the design flow, the computational and experimental results agreed to within 17% for the velocity magnitude, and 2° for the flow angle. The stagnation point locations coincided for the computational and experimental results, and the values for slip agreed to within 10%.

Commentary by Dr. Valentin Fuster
1990;():V001T01A019. doi:10.1115/90-GT-055.
FREE TO VIEW

Three linear cascades of highly loaded, low aspect ratio turbine blades have been tested in order to investigate the mechanisms by which blade lean (dihedral) influences loss generation. The blades in all three cascades have the same section but they are stacked perpendicular to the endwall in the first cascade, on a straight line inclined at 20° from perpendicular in the second and on a circular arc inclined at 30° from perpendicular at each end in the third cascade.

Lean has a marked effect upon blade loading, on the distribution of loss generation and on the state of boundary layers on the blade suction surfaces and the endwalls, but its effect upon overall loss coefficient was found to be minimal. It was found, however, that compound lean reduced the downstream mixing losses, and reasons for this are proposed. Compound lean also has the beneficial effect of substantially reducing spanwise variations of mean exit flow angle. In a turbine this would be likely to reduce losses in the downstream bladerow as well as making matching easier and improving off-design performance.

Topics: Turbines , Blades
Commentary by Dr. Valentin Fuster
1990;():V001T01A020. doi:10.1115/90-GT-056.
FREE TO VIEW

This paper presents an exhaustive experimental documentation of the three–dimensional nature of the flow in a one–and–one–half stage axial turbine. The intent was to examine the flow within, and downstream of, both the stator and rotor airfoil rows so as to delineate the dominant physical mechanisms. Part 1 of this paper presents the aerodynamic results including: (1) airfoil and endwall surface flow visualization, (2) fullspan airfoil pressure distributions, and (3) radial–circumferential distributions of the total and static pressures, and of the yaw and pitch angles in the flow. Part 2 of the paper presents results describing the mixing, or attenuation, of a simulated spanwise inlet temperature profile as it passed through the turbine. Although the flow in each airfoil row possessed a degree of three–dimensionality, that in the rotor was the strongest.

Commentary by Dr. Valentin Fuster
1990;():V001T01A021. doi:10.1115/90-GT-057.
FREE TO VIEW

This paper presents an exhaustive experimental documentation of the three–dimensional nature of the flow in a one–and–one–half stage axial turbine. The intent was to examine the flow within, and downstream of, both the stator and rotor airfoil rows so as to delineate the dominant physical mechanisms. Part 1 of this paper presented the aerodynamic results. Part 2 presents documentation of the mixing, or attenuation, of a simulated spanwise inlet temperature profile as it passed through the turbine including: (1) the simulated combustor exit–turbine inlet temperature profile, (2) surface measurements on the airfoils and endwalls of the three airfoil rows, and (3) radial–circumferential distributions downstream of each airfoil. Although all three rows contributed to profile attenuation, the impact of the rotor was strongest.

Commentary by Dr. Valentin Fuster
1990;():V001T01A022. doi:10.1115/90-GT-063.
FREE TO VIEW

There were carried out the experimental investigations of the gas dynamic features and the analysis of the thermodynamic characteristics of the advanced engine turbines, designed with allowance for effects of the contra-rotating rotors. The investigations were performed on 12 rectilinear cascades with the different fluid deflection and meridional opening. The comparison of the obtained characteristics and the analysis of the flow pattern show the cascades for contra-rotating rotors have specific features which are necessary to take account while its designing.

Topics: Engines , Rotors , Turbines
Commentary by Dr. Valentin Fuster
1990;():V001T01A023. doi:10.1115/90-GT-064.
FREE TO VIEW

A steady, three-dimensional viscous “average passage” computer code is used to analyze the flow through a compact radial turbine rotor. The code models the flow as spatially periodic from blade passage to blade passage. Results from the code using varying computational models are compared with each other and with experimental data. These results include blade surface velocities and pressures, exit vorticity and entropy contour plots, shroud pressures, and spanwise exit total temperature, total pressure, and swirl distributions. The three computational models used are inviscid, viscous with no blade clearance, and viscous with blade clearance. It is found that modeling viscous effects improves correlation with experimental data, while modeling hub and tip clearances further improves some comparisons. Experimental results such as a local maximum of exit swirl, reduced exit total pressures at the walls, and exit total temperature magnitudes are explained by interpretation of the flow physics and computed secondary flows. Trends in the computed blade loading diagrams are similarly explained.

Commentary by Dr. Valentin Fuster
1990;():V001T01A024. doi:10.1115/90-GT-073.
FREE TO VIEW

Within the scope of a compressor research program a highly loaded two stage transonic compressor was investigated.

The submitted paper presents global information on the performance characteristics of the compressor as a whole and of the two stages as a function of the variable Stator and inlet guide vane settings. For the investigated two stage compressor a variable stator 1 is able to improve the performance to the same extent as a variable inlet guide vane.

Subsequently on the basis of L2F velocimetry, hotfilm gauges and probe traverses a more detailed view of the flow inside of the compressor is given. The main results can be summarized as follows:

In spite of high supersonic relative Mach numbers in front of rotor 1 the incidence angles change evenly from hub to tip depending on the back pressure. This is in contrast to plane cascade experience and is attributed to the bow shocks.

In the downstream rotor the shock pattern is strongly influenced by the relative position between rotor 1 and rotor 2 blades. As the glue on hotfilms on the stator 2 suction side show, the influence of the rotor 1 wakes can still be recognized in the laminar separation bubble and the turbulent boundary layer. Downstream of the compressor the mixing of the rotor wakes causes strong pitchwise gradients of total temperature.

Topics: Compressors , Stators
Commentary by Dr. Valentin Fuster
1990;():V001T01A025. doi:10.1115/90-GT-074.
FREE TO VIEW

A detailed investigation of three-dimensional flow has been carried out in a low speed rear axial compressor stage with aspect ratio of 1 at the extreme off-design condition-turbine regime. Measurements were performed by means of both stationery and rotating pressure probes. The mechanism of flow in the rotor and stator blade row in the turbine regime is analysed. Comparison is made with flow mechanism at the design condition.

Topics: Compressors , Turbines
Commentary by Dr. Valentin Fuster
1990;():V001T01A026. doi:10.1115/90-GT-075.
FREE TO VIEW

The results of Laser Doppler Velocimetry (LDV) measurements, in particular, turbulent stresses in radial turbine guide vanes are presented in this paper, in order to provide experimental data for the numerical predictions. The flow velocities were measured at upstream, inside and downstream of the guide vanes for two different mass flow rates (0.2 lb/s “0.0907 kg/s” and 0.3 lb/s “0.1361 kg/s”) using a two-component LDV system. The results are presented as contour plots of turbulent stresses. The LDV system consists of a 5 watt argon-ion laser, the seeding particle atomizer, the optical and the data acquisition systems. The optical components were arranged in the backward scatter mode to measure two orthogonal velocity components simultaneously. Frequency shifts were used on both components to determine the flow direction. The results indicate a significant transport of higher turbulence fluid into the suction surface-end wall corner by the end wall cross flows inside the passage. High turbulent stress gradients show that there is considerable flow mixing downstream of the flow passages. Turbulence was found to be locally anisotropic everywhere.

Topics: Turbulence , Stress , Turbines
Commentary by Dr. Valentin Fuster
1990;():V001T01A027. doi:10.1115/90-GT-076.
FREE TO VIEW

Accurate prediction of the flow in turbomachinery requires numerical solution of the Navier-Stokes equations. A two-dimensional Navier-Stokes solver developed at ONERA for the calculation of the flow in turbine and compressor cascades was adapted at SNECMA to run on different types of grid.

The solver uses an explicit, time-marching, finite-volume technique, with a multigrid acceleration scheme. A multi-domain approach is used to handle difficulties due to the geometry of the flow. An H-C grid was used in the calculations. Two turbulence models, based on the mixing length approach, were used.

The flow in a transonic compressor cascade, a subsonic and a transonic turbine cascade were computed. Comparison with experiments is presented.

Commentary by Dr. Valentin Fuster
1990;():V001T01A028. doi:10.1115/90-GT-077.
FREE TO VIEW

The topic of this paper is the computation of transonic turbulent flow fields in high-loaded centrifugal compressor diffusers with a time-marching scheme. A thin-layer approximation is introduced into the time-dependent Navier-Stokes equations and the turbulent quantities are provided by a zero-equation eddy-viscosity model due to Baldwin and Lomax. For solving the governing equations an explicit-implicit MacCormack scheme is applied. The effect of the side wall boundary layer can be employed globally by variable stream sheet thickness.

The present code has been verified by comparison of calculated and measured data. Pressure and velocity fields as well as global results like diffuser efficiency have been considered. The code is very efficient at a CRAY-XMP vector computer. Hence, two-dimensional and quasi-three-dimensional turbulent flow fields can be obtained with a reasonable effort. However, one has to be very careful concerning the modelling of the effect of the side-wall boundary layer by variable stream sheet thickness.

Commentary by Dr. Valentin Fuster
1990;():V001T01A029. doi:10.1115/90-GT-099.
FREE TO VIEW

The authors present an analysis of the flow through a centrifugal compressor rotor. A quasi-3D flow model evaluates the interaction of the meridional and blade-to-blade solution, so as to determine the flow pattern inside an inviscid region.

A further interaction is then considered between the non-viscous flow and the boundary layers which grow along the end-walls and the blade surfaces. This makes it possible both to determine a more realistic flow condition, because of the blockage effects exerted by the boundary layers, and to estimate the total pressure losses related to the momentum thickness.

Examples are presented for a compressor of an aircraft engine. The influence of blade shape on the above described phenomena is analyzed, starting from the actual rotor geometry and making a parametric study of the alterations in flow pattern produced by changes in meridional blade shape, inlet and outlet flow areas, and splitter blades.

The analysis will provide a basis for future activities involving the use of optimizing techniques for the final choice of the blade characteristics.

Commentary by Dr. Valentin Fuster
1990;():V001T01A030. doi:10.1115/90-GT-108.
FREE TO VIEW

A comprehensive basic analysis for various counter–rotating turbines is given with the blade element stage assumption. Similar to the classical analysis of common turbine stages, the appropriate independent variables and evaluation criteria of the counter–rotating turbine stages are first presented and then three typical kinds of rotating blade rows are defined and all possible typical schemes of counter–rotating turbine stages are enumerated. Their performances of specific work, load factor distribution between two counter–rotating shafts and efficiency are analysed and discussed for different shaft rotating speed ratios. This information is useful for the selection and preliminary design of a counter–rotating turbine. From the analysis results, it is concluded that the load capacity per unit engine length of counter–rotating turbines can be much higher than that of common turbines (approximately twice) without efficiency penalty or even with higher efficiency. Some triple counter–rotating turbines suitable for three shaft gas turbine power plants are proposed and analysed briefly too.

Topics: Turbines
Commentary by Dr. Valentin Fuster
1990;():V001T01A031. doi:10.1115/90-GT-122.
FREE TO VIEW

A new method for dynamic control of centrifugal compressor surge is presented. The approach taken is to suppress surge by modifying the compression system dynamic behavior using structural feedback. More specifically, one wall of a downstream volume, or plenum, is constructed so to move in response to small perturbations in pressure. This structural motion provides a means for absorbing the unsteady energy perturbations produced by the compressor, thus extending the stable operating range of the compression system.

In the paper, a lumped parameter analysis is carried out to define the coupled aerodynamic and structural system behavior and the potential for stabilization. First-of-a-kind experiments are then carried out to examine the conclusions of the analysis. As predicted by the model and demonstrated with experiment, a moveable plenum wall lowered the mass flow at which surge occurred in a centrifugal compression system by roughly 25% for a large range of operating conditions. In addition, because the tailored dynamics of the structure acts to suppress instabilities in their initial stages, this control was achieved with relatively little power being dissipated by the moveable wall system, and with no noticeable decrease in steady state performance. Although designed on the basis of linear system considerations, the structural control is shown to be capable of suppressing existing large amplitude limit cycle surge oscillations.

Topics: Compressors , Surges
Commentary by Dr. Valentin Fuster
1990;():V001T01A032. doi:10.1115/90-GT-123.
FREE TO VIEW

Active suppression of centrifugal compressor surge has been demonstrated on a centrifugal compressor equipped with a servo-actuated plenum exit throttle controller. The control scheme is fundamentally different from conventional surge control techniques in that it addresses directly the dynamic behavior of the compression system to displace the surge line to lower mass flows. The method used is to feed back perturbations in plenum pressure rise, in real time, to a fast acting control valve. The increased aerodynamic damping of incipient oscillations due to the resulting valve motion allows stable operation past the normal surge line. For the compressor used, a 25% reduction in the surge point mass flow was achieved, over a range of speeds and pressure ratios. Time-resolved measurements during controlled operation revealed that the throttle required relatively little power to suppress the surge oscillations, because the disturbances are attacked in their initial stages. Although designed for operation with small disturbances, the controller was also able to eliminate existing, large amplitude, surge oscillations. Comparison of experimental results with theoretical predictions showed that a lumped parameter model appeared adequate to represent the behavior of the compression system with the throttle controller and, perhaps more importantly, to be used in the design of more sophisticated control strategies.

Topics: Compressors , Surges
Commentary by Dr. Valentin Fuster
1990;():V001T01A033. doi:10.1115/90-GT-124.
FREE TO VIEW

L2F measurements of the flow at the exit of modern unshrouded centrifugal impellers with backswept blades yield a much more uniform velocity profile compared to former measurements on impellers with radial blading. Further evaluations show that the “classical” jet-wake theory assuming an isentropic jet and a wake flow congruent with the shape of the blade at the impeller exit needs correction in order to obtain meaningful results when interpreting thermodynamic measurements on centrifugal compressor stages.

Commentary by Dr. Valentin Fuster
1990;():V001T01A034. doi:10.1115/90-GT-127.
FREE TO VIEW

Experimental measurements in a linear cascade with tip clearance are complemented by numerical solutions of the three-dimensional Navier-Stokes equations in an investigation of tip leakage flow. Measurements reveal that the clearance flow, which separates near the entry of the tip gap, remains unattached for the majority of the blade chord when the tip clearance is similar to that typical of a machine. The numerical predictions of leakage flow rate agree very well with measurements and detailed comparisons show that the mechanism of tip leakage is primarily inviscid. It is demonstrated by simple calculation that it is the static pressure field near the end of the blade which controls chordwise distribution of the flow across the tip. Although the presence of a vortex caused by the roll-up of the leakage flow may affect the local pressure field, the overall magnitude of the tip leakage flow remains strongly related to the aerodynamic loading of the blades.

Commentary by Dr. Valentin Fuster
1990;():V001T01A035. doi:10.1115/90-GT-128.
FREE TO VIEW

The ability to acquire blade loadings (surface pressure distributions) and surface flow visualization on an unshrouded centrifugal compressor impeller is demonstrated. Circumferential and streamwise static pressure distributions acquired on the stationary shroud are also presented. Data was acquired in a new facility designed for centrifugal compressor aerodynamic research. Blade loadings calculated with a blade–to–blade potential flow analysis are compared with the measured results. Surface flow visualization reveals some complex aspects of the flow on the surface of the impeller blading and hub. In a companion paper, Dorney and Davis (1990), a state–of–the–art, three–dimensional, time–accurate, Navier Stokes prediction of the flow through the impeller is presented.

Commentary by Dr. Valentin Fuster
1990;():V001T01A036. doi:10.1115/90-GT-129.
FREE TO VIEW

The results of two component laser-Doppler velocimeter (LDV) surveys made in the near wake (to one fifth chord) of a controlled diffusion (CD) compressor blade in a large scale cascade wind tunnel, are reported. The measurements were made at three positive incidence angles from near-design to angles thought to approach stall. Comparisons were made with calibrated pressure probe and hot-wire wake measurements and good agreement was found. The flow was found to be fully attached at the trailing edge at all incidence angles and the wake profiles were found to be highly skewed. Despite the precision obtained in the wake velocity profiles, the blade loss could not be evaluated accurately without measurements of the pressure field. The blade trailing edge surface pressures and velocity profiles were found to be consistent with downstream pressure probe measurements of loss, allowing conclusions to be drawn concerning the design of the trailing edge.

Commentary by Dr. Valentin Fuster
1990;():V001T01A037. doi:10.1115/90-GT-130.
FREE TO VIEW

Boundary layer transition has been investigated experimentally under low, moderate and high free-stream turbulence levels and varying adverse pressure gradients. Under high turbulence levels and adverse pressure gradients a pronounced subtransition was present. A strong degree of similarity in intermittency distributions was observed, for all conditions, when the Narasimha procedure for determination of transition inception was used.

Effects of free-stream turbulence on the velocity profile are particularly strong for the laminar boundary layer upstream of the transition region. This could reflect the influence of the turbulence on the shear stress distribution throughout the layer and this matter needs further attention. The velocity profiles in wall coordinates undershoot the turbulent wall layer asymptote near the wall over most of the transition region.

The rapidity with which transition occurs under adverse pressure gradients produces strong lag effects on the velocity profile; the starting turbulent boundary layer velocity profile may depart significantly from local equilibrium conditions. The practice of deriving integral properties and skin friction for transitional boundary layers by a linear combination of laminar and turbulent values for equilibrium layers is inconsistent with the observed lag effects.

The velocity profile responds sufficiently slowly to the perturbation imposed by transition that much of the anticipated drop in form factor will not have occurred prior to the completion of transition. This calls into question both experimental techniques which rely on measured form factor to characterize transition and boundary layer calculations which rely on local equilibrium assumptions in the vicinity of transition.

Commentary by Dr. Valentin Fuster
1990;():V001T01A039. doi:10.1115/90-GT-132.
FREE TO VIEW

A methodology based on wake mixing has been developed that enables more accurate predictions of compressor airfoil pressure distributions when the airfoil is operating downstream of an airfoil row that has strong wakes. The methodology has an impact on through–flow analysis, on airfoil–to–airfoil flow analysis, and on the interpretation of experimental data. It is demonstrated that the flow in the endwall region is particularly sensitive to mixing due to the strong wakes caused by the secondary flow and corner separation that commonly occur in this region. It is also demonstrated that wake mixing can have a strong impact on both airfoil incidence and deviation as well as on loading. Differences of up to 13° and 30% in loading are demonstrated.

Commentary by Dr. Valentin Fuster
1990;():V001T01A040. doi:10.1115/90-GT-133.
FREE TO VIEW

Time-resolved radial transport has been measured in a transonic compressor rotor by injecting a thin sheet of tracer gas upstream of the rotor and then surveying the tracer concentration at the rotor exit. The simultaneous, co-located, high frequency response measurements of local tracer gas concentration, total temperature, and total pressure made downstream of the rotor showed that most of the fluid transported radially appears in the blade wakes and that this fluid has considerably higher entropy than the circumferential mean. Both inward and outward fluid transport along the span was observed (3.5% of the total throughflow moved toward the tip while 1.6% moved toward the hub). Tracer concentration and fluid total temperature and pressure varied considerably from wake to wake, even on multiple samplings of the same blade. The time mean spreading rate inferred from these measurements is in general agreement with previously reported studies on multi-stage low speed compressors and is well predicted by the method of Gallimore and Cumpsty. It is suggested that a vortex street in the blade wakes could be responsible for both the observed radial transport and the large wake to wake variability. A quasi-three-dimensional model of a vortex street wake was developed and shown to be consistent with the data. The model predicts all of the inward transport but only 20% of the outward transport. It is hypothesized that outflow in separated regions on the blade suction surface is responsible for the remainder of the transport toward the rotor tip. Since the entropy, as well as the mass of the fluid transported radially, was measured, an estimate of the redistribution of loss in rotor due to radial fluid transport could be made. This showed that the effect of radial transport in this rotor was to move substantial loss from the rotor hub to tip, implying that a conventionally measured spanwise efficiency survey may not accurately represent the performance of individual blade sections.

Topics: Compressors
Commentary by Dr. Valentin Fuster
1990;():V001T01A041. doi:10.1115/90-GT-134.
FREE TO VIEW

An investigation of the “additional” total pressure losses occurring in combining flow through several sharp-edged three-leg junctions has been made. Experimental results covering a wide speed range up to choking are presented for three flow geometries of a lateral branch off a straight duct using dry air as the working fluid. A new theoretical flow model provided results in fairly good agreement with the experimental data obtained. Flow visualisation of the high speed flow using the Schlieren method revealed the presence of normal shock waves in the combined flow about one duct diameter downstream of the junction. The highest attainable Mach number (M3) of the averaged downstream (combined) flow was 0.66 for several of the flow geometries. This value of M3 appears to be the maximum possible and is the result of a combination of flow separation and local flow choking.

Commentary by Dr. Valentin Fuster
1990;():V001T01A042. doi:10.1115/90-GT-135.
FREE TO VIEW

The paper presents finite-difference predictions for the convective heat transfer in symmetrically-heated rotating cavities subjected to a radial outflow of cooling air. An elliptic calculation procedure has been used, with the turbulent fluxes estimated by means of a low Reynolds number k-ε model and the familiar ‘turbulence Prandtl number’ concept. The predictions extend to rotational Reynolds numbers of 3.7 × 106 and encompass cases where the disc temperatures may be increasing, constant or decreasing in the radial direction.

It is found that the turbulence model leads to predictions of the local and average Nusselt numbers for both discs which are generally within ± 10% of the values from published experimental data, although there appear to be larger systematic errors for the upstream disc than for the downstream disc. It is concluded that the calculations are of sufficient accuracy for engineering design purposes, but that improvements could be brought about by further optimization of the turbulence model.

Commentary by Dr. Valentin Fuster
1990;():V001T01A043. doi:10.1115/90-GT-136.
FREE TO VIEW

Numerical solutions of the Reynolds-averaged Navier-Stokes equations have been used to model the influence of cobs and a bolt cover on the flow and heat transfer in a rotating cavity with an imposed radial outflow of air. Axisymmetric turbulent flow is assumed using a mixing length turbulence model. Calculations for the non-plane discs are compared with plane disc calculations and also with the available experimental data. The calculated flow structures show good agreement with the experimentally observed trends. For the cobbed and plane discs, Nusselt numbers are calculated for a combination of flow rates and rotational speeds; these show some discrepancies with the experiments, although the calculations exhibit the more consistent trend. Further calculations indicate that differences in thermal boundary conditions have a greater influence on Nusselt number than differences in disc geometry. The influence of the bolt cover on the heat transfer has also been modelled, although comparative measurements are not available.

Commentary by Dr. Valentin Fuster
1990;():V001T01A044. doi:10.1115/90-GT-140.
FREE TO VIEW

Similarly as in jet engine development, modern design methods are used today to improve the performance of industrial compressors. In order to verify the loading limits, a cascade profile representative for the first rotor hub section of an industrial compressor has been designed by optimizing the suction surface velocity distribution using a direct boundary layer calculation method. The blade shape was computed with an inverse full potential code and the resulting cascade was tested in a cascade windtunnel. The experimental results confirmed the design intent and resulted in a low loss coefficient of 1.8% at design condition and an incidence range of nearly 12° (4% loss level) at an inlet Mach number of 0.62.

Commentary by Dr. Valentin Fuster
1990;():V001T01A045. doi:10.1115/90-GT-141.
FREE TO VIEW

In order to understand overall performance and internal flows of air-cooled turbine blade rows, flows in a model linear cascade were surveyed with secondary air injection from various locations of the blade surfaces. The secondary air interacted with the cascade passage vortices and changed the loss distribution significantly. The cascade overall loss decreased when the air was injected along the mainstream and increased when the air was injected against the mainstream from some locations of the blade leading edge. Effects on overall kinetic energy of the secondary flows and on the cascade outlet flow angle were also discussed in this paper.

Commentary by Dr. Valentin Fuster
1990;():V001T01A046. doi:10.1115/90-GT-146.
FREE TO VIEW

A series of experiments are performed to investigate and quantify the three-dimensional mean flow field in centrifugal compressor flow passages and to evaluate contemporary internal flow models. The experiments include the acquisition and analysis of LDV data in the impeller passages of a low speed moderate scale research mixed-flow centrifugal compressor operating at its design point. Predictions from a viscous internal flow model are then correlated with these data. The LDV data show the traditional jet-wake structure observed in many centrifugal compressors, with the wake observed along the shroud 70% of the length from the pressure to suction surface. The viscous model predicts the major flow phenomena. However, the correlations of the viscous predictions with the LDV data were poor.

Commentary by Dr. Valentin Fuster
1990;():V001T01A047. doi:10.1115/90-GT-151.
FREE TO VIEW

Experiments were conducted to determine the pressure distributions within a multi–cavity compressor drum model for two coolant injection locations and a range of flow conditions. Flow was injected through the upstream conical wall or through the cylindrical wall of the rotating model. The coolant flow, the drum rotational rate, and the model pressure were varied to produce a range of tangential and coolant flow Reynolds numbers, typical of large aircraft engine high pressure compressor drums. The experimental results were used to evaluate analytical procedures for predicting flow characteristics in rotating annular cavities with radially inward flow and for correlating flow characteristics in multiple–rotating annular cavities which are not currently predicted. Swirling flows, radially inward between compressor disks and within rotating annular cavities with no net flow, were analyzed with a procedure which coupled a viscous solution for the rotating core flow with a momentum integral analysis for the boundary layers on the disks. Constant viscosity and variable turbulent viscosity models were used in the analysis. Results from the analysis and the experiments were used to estimate the tangential velocity distribution in trapped cavities for two coolant injection configurations and a range of flow rates.

Topics: Compressors , Coolants
Commentary by Dr. Valentin Fuster
1990;():V001T01A048. doi:10.1115/90-GT-152.
FREE TO VIEW

To assess the possibility of tip clearance loss reduction and to explore the nature and origin of tip clearance loss, blade tip geometries which reduce the roughly 40% of total loss occurring within the gap were studied. The shapes investigated aimed at reducing or avoiding the gap separation bubble thought to contribute significantly to both internal gap loss and to the endwall mixing loss. It was found that radiusing and contouring the blade at gap inlet eliminated the separation bubble and reduced the internal gap loss but created a higher mixing loss to give almost unchanged overall loss coefficients when compared with the simple sharp edged flat tipped blade. The separation bubble does not therefore appear to influence the mixing loss. Using a method of assessing linear cascade experimental data as though it were a rotor with work transfer, one radiused geometry, contoured to shed radial flow into the gap and reduce the leakage mass flow, was found to have a significantly higher efficiency. This demonstrates the effectiveness of the data analysis method and that cascade loss coefficient alone or gap discharge coefficient cannot be used to accurately evaluate tip clearance performance. Contouring may ultimately lead to better rotor blade performances.

Commentary by Dr. Valentin Fuster
1990;():V001T01A049. doi:10.1115/90-GT-153.
FREE TO VIEW

A new approach is presented for analyzing compressor tip clearance flow. The basic idea is that the clearance velocity field can be (approximately) decomposed into independent through-flow and cross-flow, since chordwise pressure gradients are much smaller than normal pressure gradients in the clearance region. As in the slender body approximation in external aerodynamics, this description implies that the three-dimensional, steady, clearance flow can be viewed as a two-dimensional, unsteady flow. Using this approach, a similarity scaling for the cross-flow in the clearance region is developed and a generalized description of the clearance vortex is derived. Calculations based on the similarity scaling agree well with a wide range of experimental data in regard to flow features such as cross-flow velocity field, static pressure field, and tip clearance vortex trajectory. The scaling rules also provide a useful way of exploring the parametric dependence of the vortex trajectory and strength for a given blade row. The emphasis of the approach is on the vortical structure associated with the tip clearance because this appears to be a dominant feature of the endwall flow; it is also shown that this emphasis gives considerable physical insight into overall features seen in the data.

Commentary by Dr. Valentin Fuster
1990;():V001T01A050. doi:10.1115/90-GT-154.
FREE TO VIEW

Existing methods for predicting the tip-leakage losses in turbomachinery are based on a variety assumptions, many of which have not been fully verified experimentally. Recently, several detailed experimental studies in turbine cascades have helped to clarify the physics of the flow and provide data on the evolution of the losses. The paper examines the assumptions underlying the prediction methods in the light of these data. An improved model for the losses is developed, using one of the existing models as the starting point.

Topics: Turbines , Leakage
Commentary by Dr. Valentin Fuster
1990;():V001T01A051. doi:10.1115/90-GT-155.
FREE TO VIEW

A detailed experimental and numerical investigation was carried out to examine the three-dimensional flow field, secondary flows and vortex motion in an annular compressor cascade.

Various flow visualizations near the blade surface and endwalls, wall static pressure and loss measurements, as well as hot-film and hot-wire measurements inside the blade boundary layers were performed at various flow rates to understand the complex flow phenomena. A Reynolds-averaged Navier-Stokes equation was solved to investigate the flow numerically.

The detailed comparison between measurement and numerical prediction indicates that the complex three-dimensional flow phenomena (corner stall, vortex motion, radial mixing, etc.) is very well predicted with the numerical method.

Commentary by Dr. Valentin Fuster
1990;():V001T01A052. doi:10.1115/90-GT-156.
FREE TO VIEW

Stall inception has been studied in two low speed compressors (a single-stage and a three-stage) and in a high speed three-stage compressor, using temporally and spatially resolved measurements. In all three machines, rotating stall was preceded by a period in which small amplitude waves were observed travelling around the circumference of the machine at a speed slightly less than the fully developed rotating stall cell speed. The waves evolved smoothly into rotating stall without sharp changes in phase or amplitude, implying that, in the machines tested, the prestall waves and the fully developed rotating stall are two stages of the same phenomenon. The growth rate of these disturbances was in accord with that predicted by current analytical models. The prestall waves were observed both with uniform and with distorted inflow, but were most readily discerned with uniform inflow. Engineering uses and limitations of these waves are discussed.

Commentary by Dr. Valentin Fuster
1990;():V001T01A055. doi:10.1115/90-GT-159.
FREE TO VIEW

A passive control of an unstable characteristics of a high specific speed diagonal-flow fan has been proposed. It is possible to eliminate the unstable characteristics of pressure-flow rate curve in a low flow region without deterioration of performance at design point. The control action is done naturally (passively) without any energy input. The inlet nozzle of an ordinary diagonal-flow fan was replaced by an annular wing with Göttingen 625 airfoil section. The mechanism of the passive control and the optimum geometrical parameter are discussed on the basis of the performance tests and internal flow measurements.

Commentary by Dr. Valentin Fuster
1990;():V001T01A056. doi:10.1115/90-GT-160.
FREE TO VIEW

Two methods are described for calculating pressure distributions and boundary layers on blades subjected to low Reynolds numbers and ramp–type motion. The first is based on an interactive scheme in which the inviscid flow is computed by a panel method and the boundary layer flow by an inverse method that makes use of the Hilbert integral to couple the solutions of the inviscid and viscous flow equations. The second method is based on the solution of the compressible Navier–Stokes equations with an embedded grid technique that permits accurate calculation of boundary layer flows. Studies for the Eppler and NACA–0012 airfoils indicate that both methods can be used to calculate the behavior of unsteady blade boundary layers at low Reynolds numbers provided that the location of transition is computed with the en–method and the transitional region is modelled properly.

Commentary by Dr. Valentin Fuster
1990;():V001T01A057. doi:10.1115/90-GT-161.
FREE TO VIEW

The unshrouded impeller and the vaneless diffuser of a single-stage radial compressor have been investigated at three flow rates. Three-dimensional velocities and pressures were measured at a tip speed of 84 m/s by an L2F-velocimeter, a slanted single hot-wire probe and piezoresistive pressure transducers. The measurements show that upstream the blading the averaged meridional inlet flow angle is about 54 degree and a periodical variation of the meridional flow angle of about 25 degree occurs near the casing wall. Further, an inlet vortex of clockwise direction appears and an initial whirl is induced. The specific work of the initial whirl corresponds to approximately 12% of the enthalpy losses between inlet pipe and diffuser outlet. In the beginning of the passage, the inlet vortex is suppressed and a solid body vortex of counterclockwise direction can be observed. At the outlet, a heavy flow deceleration at the blade suction side with subsequent separation can be seen. Increasing the flow rate decreases the wake and causes a more uniform loss distribution in this area. The measured secondary vortex flow and rotary stagnation pressure gradients are compared with test results from impellers with inducer. The incidence of the investigated impeller is greater than that of the impellers with inducer, but the wake-jet outlet flows are very similar. Inlet losses could be reduced by improving incidence angles by matching the blade angles to the inlet flow angles. Smaller blade angles at the shroud would reduce or eliminate separation at the leading edge, and the resulting reduction in low momentum fluid along the suction surface would help to avoid separation on that surface near the outlet.

Commentary by Dr. Valentin Fuster
1990;():V001T01A058. doi:10.1115/90-GT-187.
FREE TO VIEW

The performance of a radial inward flow gas turbine stage depends on the design of the rotor as well as that of the casing. The paper describes a simple procedure for designing the complete stage comprising a rotor and a single entry nozzle-less volute casing. The rotor design follows a prescribed mean stream velocity distribution in a step by step manner and the casing design uses the radial equilibrium together with the conservation of mass energy and momentum equations.

Commentary by Dr. Valentin Fuster
1990;():V001T01A059. doi:10.1115/90-GT-198.
FREE TO VIEW

A fully three dimensional compressible inverse design method for the design of radial and mixed flow machines is described. In this method the distribution of the circumferentially averaged swirl velocity, or Display FormularV¯θ on the meridional geometry of the impeller is prescribed and the corresponding blade shape is computed iteratively. Two approaches are presented for solving the compressible flow problem. In the approximate approach, the pitchwise variation in density is neglected and as a result the algorithm is simple and efficient. In the exact approach, the velocities and density are computed throughout the three dimensional flow field by employing Fast Fourier Transform in the tangential direction. The results of the approximate and exact approach are compared for the case of a high speed (subsonic) radial-inflow turbine and it is shown that the difference between the blade shapes computed by the two methods is well within the manufacturing tolerances. The flow through the designed impeller is analysed by using three dimensional inviscid and viscous time marching programs and very good correlations between the specified and computed Display FormularV¯θ is obtained.

Commentary by Dr. Valentin Fuster
1990;():V001T01A060. doi:10.1115/90-GT-207.
FREE TO VIEW

Experimental investigations of loss reduction with riblets have been made. Riblets were put on the blade surfaces in different ways in a compressor cascade or on an isolated airfoil. More than 8% drag reduction benefits for an isolated airfoil and more than 10% loss reduction benefits for the compressor cascade have been obtained by using riblets only on blade pressure surfaces rather than on both pressure and suction surfaces. The mechanism underlying these results has been discussed.

Commentary by Dr. Valentin Fuster
1990;():V001T01A061. doi:10.1115/90-GT-208.
FREE TO VIEW

The effects of surface roughness, freestream turbulence, and incidence angle on the performance of a two-dimensional compressor cascade were investigated. The test section consisted of seven NACA 65-A506 airfoils arranged in a linear cascade. Four different surface roughness conditions were applied to the first 25 percent chord on the suction surface of each of the five middle blades in the cascade. Freestream turbulence levels of approximately one and seven percent were used. Incidence angles of −3, zero and +3 degrees were investigated. Of the three parameters tested, freestream turbulence exerted the largest influence on blade performance. The total pressure loss coefficient increased with increased roughness and was reduced for large turbulence. Changes in flow incidence had a lesser effect on the performance of the blade.

Commentary by Dr. Valentin Fuster
1990;():V001T01A062. doi:10.1115/90-GT-209.
FREE TO VIEW

An investigation is presented of the interaction between a compressor and a downstream distorting component, which is of increasing importance for modern aircraft turbine engines with compact size. The prediction is based on a time–marching method for compressible flow and large perturbation. The results show that the flow field of a compressor is modified by the interaction between the compressor and the downstream distorting component. Nonuniformity in the exit static pressure, maximum distortion at the last blade row implying possible critical source for instability are characteristic of the flow field with downstream distortion. A matching map is introduced to describe the whole flow process, giving clearer insight in the physical aspect to the influence of various factors on the propagation of the downstream distortion. The spacing between the compressor and the downstream distorting component is shown to have an influence of exponential type on the distortion at every axial position. The variation of the distortion with distortion coefficient is basically linear. Steeper characteristics of the compressor will produce stronger attenuation of the distortion. Compressors of multistages have an effect similar to the steeper characteristic.

Topics: Compressors
Commentary by Dr. Valentin Fuster
1990;():V001T01A063. doi:10.1115/90-GT-210.
FREE TO VIEW

This paper reports an experimental investigation of the dynamic behaviour of an axial flow compressor when the flow rate is forced to vary quickly. The results are discussed in two aspects: general trends and special points. Generally, there is a deviation of the dynamic characteristic from the steady–state one, depending upon the algebraic sum of the inertial effect and the lag effect. For the present axial flow compressor, lightly–loaded, the dynamic characteristic goes above the steady–state one while closing the valve and below the steady–state one for opening the valve. Physical explanation is given to the general trend and also to the difference between the behavior of the axial flow compressor and that of the centrifugal compressor. Specially, the rapid closing of the valve will end up with an operating point end dynamic breaking into the instability limit while the throttle setting is somewhat away from the instability limit position for steady–state operation; and the rapid opening of the valve may produce an initial transient decrease in flow rate thus may have an initial dynamic breaking into the instability limit. Physical explanation is also given to these two special points, which are of great engineering interest with regard to the instability issue.

Commentary by Dr. Valentin Fuster
1990;():V001T01A064. doi:10.1115/90-GT-211.
FREE TO VIEW

An investigation is presented of basic features of speed induced transient behaviours of axial compression system. A non–linear model is developed for prediction of the transient response due to speed variation. Generally, the operating point forms a loop during an accelerating process. A small B or great rate of change in speed leads to a large loop. In the hysteresis region of the characteristic of the compressor, there are four possible solutions: steady flow with the compressor working on the unstalled part of the characteristic; steady flow with the operating point first getting across the stall limit then going back to the initial point; rotating stall and surge. Two critical values of the B parameter are identified, on which the system response depends. They are: Br, at which transition from rotating stall to surge occurs, and Bs, at which that from surge to steady flow takes place. Br and Bs are found to vary with some factors concerned. Their variation with the initial point, rate of the change in speed and the time lag constant are also studied.

Commentary by Dr. Valentin Fuster
1990;():V001T01A065. doi:10.1115/90-GT-212.
FREE TO VIEW

In order to improve the air inlet engine compatibility, SNECMA has carried out since several years an important effort to predict the effect of distorted inlet flow on compressor stability.

Two different methods are developed:

In the first one, the Euler equations are integrated in 2D blade to blade surface with a distorted inlet flow. This method is used to compare different profiles, in particular influence of the chord length is presented.

In the second one, the aerodynamic behaviour of a multistage compressor operating in distorted inlet flow is calculated with a three dimensional method. This model is based on Euler equations resolution outside the rows.

An actuator disk model is used to represent the response of the blade rows.

The behaviour of a three stage axial compressor has been studied. The loss of surge margin and the pressure distortion transfer are compared with experimental data.

Commentary by Dr. Valentin Fuster
1990;():V001T01A066. doi:10.1115/90-GT-213.
FREE TO VIEW

A three-dimensional, Navier-Stokes analysis is presented for the prediction of viscous flows through centrifugal impellers. Based on the Navier-Stokes rotor/stator interaction procedure developed by Rai, the present analysis uses a zonal grid methodology to discretize the impeller flow field and to facilitate the relative motion of the impeller. A blade surface oriented O-grid generated from an elliptic partial differential equation solution procedure is patched into an algebraically generated H-grid which is used to discretize the inlet, exit and blade-to-blade regions. The equations of motion are integrated using a spatially third-order accurate, implicit, iterative, upwind, finite difference, time-marching technique. Predicted results are presented for flow through a low speed centrifugal compressor impeller operating at design flow conditions. Comparison of these predicted results with experimental data demonstrates the capability of this procedure to predict impeller blade loading and provide insight into the secondary flow structure within the impeller blade passage.

Commentary by Dr. Valentin Fuster
1990;():V001T01A067. doi:10.1115/90-GT-214.
FREE TO VIEW

Encouraging developments have recently been made both in the understanding and the modelling of compressor flowfields. This paper describes an experimental investigation undertaken to assess the validity of the assumptions made in the simple non-linear model for a bladerow operating in an unsteady or non-uniform flowfield. The results show that the basic fluid dynamics of the problem have been correctly modelled and that meaningful predictions may be made. The measurements also indicate that inlet guide vanes may be more sensitive to incidence variations than had been previously thought. The effects of inter-bladerow gaps are also discussed.

Topics: Compressors
Commentary by Dr. Valentin Fuster
1990;():V001T01A068. doi:10.1115/90-GT-215.
FREE TO VIEW

For many years there has been a significant effort to better understand rotating stall/surge phenomena in axial compressors, as well as the effects of inlet flow distortions. Most experimental investigations to date, whether on laboratory compressors or on full scale jet engines, have tended to focus separately on the effects of these two flow disturbances on compressor performance. The purpose of the present study was to experimentally assess the influence of inlet flow distortions on the inception and nature of rotating stall in a full scale engine compressor. This paper reports results obtained for the first stage rotor of a 10-stage compressor subjected to screen-induced inlet pressure distortions. Previous investigations had shown that during part-speed operation, the front stages operated in rotating stall, and hot-film probe measurements made during the present study showed that the presence of the distortion screens did not affect the speed of the rotating stall pattern, but in some instances changed the number of cells present. However, low frequency flow fluctuations characteristic of surge were much more prevalent when the screens were in place. Also, the further the sensing probe was displaced tangentially in the rotor rotation direction from the screens, the more intense the fluctuations, leading one to conclude that the screens had a localized damping effect on the surge cycle.

Topics: Flow (Dynamics)
Commentary by Dr. Valentin Fuster
1990;():V001T01A069. doi:10.1115/90-GT-216.
FREE TO VIEW

Detailed measurements have been made of the stalled flow in a centrifugal pump-turbine model operating in air. Instantaneous velocity, total pressure and flow angle have been measured with hot wire and pressure probes at the impeller inlet and outlet and in the vaned diffuser for several pump operating points, ranging from nominal to very low flow rates.

Three distinct stall phenomena have been found to affect the pump operation at reduced flow rates. The unsteady characteristics of these unstable flows have been analysed both in time and in frequency domain by means of Fourier transforms of the velocity signals. The changes in the flow structure, correlated to the occurrence of the observed stall conditions, have been investigated by means of phase-locked ensemble averages of the instantaneous velocity components at the impeller outlet and surveys of the mean total pressure and absolute flow angle distributions within the diffuser passages.

Commentary by Dr. Valentin Fuster
1990;():V001T01A070. doi:10.1115/90-GT-222.
FREE TO VIEW

The boundary layer profile is observed to be highly distorted by the momentum transfer associated with Taylor-Gortler vortices. This distortion has the effect or increasing the boundary layer growth rate and modifying the start-or-transition momentum thickness Reynolds numbers from the flat plate value. For modest curvatures, the vortices carry turbulence towards the blade which promotes transition, but as the curvature is increased further, the boundary layer profile shape is stabilised and transition is delayed. A model for the distorted boundary layer is presented and is used to predict the boundary layer growth rate and correlate the start or transition results in terms of the momentum thickness and blade radius Reynolds numbers and the free stream turbulence level. The degree or profile distortion needs to be accounted for when predicting the end or transition.

Topics: Boundary layers
Commentary by Dr. Valentin Fuster
1990;():V001T01A071. doi:10.1115/90-GT-223.
FREE TO VIEW

The high temperature level reached at the exit of combustion chambers of modern aircraft engines and the practical limitations of advanced materials, demand efficient cooling of turbine blades. Optimization of the cooling requires an accurate prediction of aerodynamic losses and heat transfer on turbine blades.

A new two-dimensional compressible, aerothermal boundary layer code has been developed. The formulation includes strong viscous-inviscid interaction, which enhances the stability properties of the code. The boundary layer equations associated with the energy equation are solved with an implicit Keller-box scheme. Viscous-inviscid flow coupling is performed by adding an interaction equation which has an elliptic character. The complete system of equations is solved by a multi-pass procedure. This technique contributes to the stabilization of the method and allows the computation of regions with strong adverse pressure gradients, separation bubbles and injections in case of film cooling.

Comparisons between experimental and theoretical results are provided. Flow characteristics including heat transfer were computed for several cases such as flat plates with strong pressure gradients, and turbine blade boundary layers. Good agreement between computation and experiment is observed, demonstrating the high accuracy and robustness of the code.

Commentary by Dr. Valentin Fuster
1990;():V001T01A072. doi:10.1115/90-GT-224.
FREE TO VIEW

Epstein, Ffowcs Williams and Greitzer presented in 1986 arguments that, provided small perturbation stability models were relevant to engine surge and rotating stall, then both could be avoided by active control. The idea was tested and validated on turbocharger compression systems at MIT and Cambridge University. Engine experiments are more difficult, the surge more violent and the signals usually contaminated by noise. This paper describes the early results obtained on a gas turbine engine subjected to the same control strategy. The control system can be switched on when the engine operates in violent surge and smooth operation is restored. This proof that the technique has practical relevance is interesting enough to report in advance of detailed investigations of its scope. This paper reports the experimental details, describes the relevant theoretical model and gives the first positive test results.

Topics: Engines , Surges
Commentary by Dr. Valentin Fuster
1990;():V001T01A073. doi:10.1115/90-GT-226.
FREE TO VIEW

Film cooling of turbine blades by injecting air through holes or slots affects the main stream flow. A numerical model has been developed to predict the resulting three-dimensional flow and the temperature pattern under steady flow conditions. An elliptic procedure is used in the near injection area to include reverse flow situations, while in the upstream area as well as far downstream a partial-parabolic procedure is applied. As first step an adiabatic wall has been assumed as boundary condition, since for this case experimental data are readily available for comparison.

At elevated momentum blowing rates, zones of reverse flow occur downstream of the injection holes resulting in a decrease of cooling efficiency. A variation of the relevant parameters momentum blowing rate m, injection angle α and ratio of hole spacing to diameter s/d revealed the combination of m ≈ 1, α ≈ 30° and s/d ≈ 2 to be the optimum with respect to the averaged cooling efficiency and to the aerodynamic losses. Cooling is more efficient with slots than with a row of holes not considering the related problems of manufacture and service life. The calculated temperature patterns compare well with the experimental data available.

Topics: Film cooling
Commentary by Dr. Valentin Fuster
1990;():V001T01A074. doi:10.1115/90-GT-227.
FREE TO VIEW

A numerical study of the steady, viscous flow prediction capabilities of the three-dimensional turbine stage code R0T0R3 is presented. Computations were performed with RAI3DC, a cascade version of R0T0R3 capable of being run in a planar or annular mode. Computed results are compared with experimental data obtained for Hodson’s cascade, Kopper’s cascade, and United Technologies Research Center’s Large Scale Rotating Rig (LSRR) first-stage stator. The code’s predictive capability is assessed in terms of the accuracy of predicted airfoil loadings, performance (including secondary flows in the LSRR case), boundary layers, and heat transfer. A grid refinement study was conducted in the LSRR case in an effort to more accurately model the boundary layers on the airfoil and endwall surfaces. The effects of the inlet total pressure profile in secondary flow prediction were also assessed.

Topics: Turbines
Commentary by Dr. Valentin Fuster
1990;():V001T01A075. doi:10.1115/90-GT-228.
FREE TO VIEW

Three methods of vortex control over the end wall flow in compressor cascades have been investigated experimentally. The total pressure loss at the exit of a linear compressor cascade is reduced 6.5%, 10.5% and 26.5% respectively by these methods for different incidences over a range of moderate-high values. The physics of these methods has been discussed and some new concepts of vortex control techniques in compressor cascades have been proposed.

Commentary by Dr. Valentin Fuster
1990;():V001T01A076. doi:10.1115/90-GT-229.
FREE TO VIEW

In this paper an extension of our secondary flow calculation method is presented in order to estimate the energy exchange and the secondary losses in axial and radial compressors.

Although it is well known today that end-wall shear layers play a major role in the performance of the compressors, there are only few secondary flow methods able to predict realistically the behaviour of real machines. For this purpose a new theory was developed which extends Mellor’s basic principles and uses our previous work, including the complete form of meridional vorticity transport equation.

A coherent model for the real work exchange between the fluid and the machine has been developed, as well as for the distribution of secondary losses, not only at the compressor’s exit but also at every station through the machine. An expression for the added work due to the secondary field is also given. Finally the influence of secondary losses in the overall efficiency is well estimated.

This work has been successfully applied to a highly loaded compressor cascade, as well as to a transonic axial compressor and to a radial one-stage one. Comparison between theoretical and experimental results is also presented.

Topics: Compressors
Commentary by Dr. Valentin Fuster
1990;():V001T01A077. doi:10.1115/90-GT-230.
FREE TO VIEW

This paper presents the impeller design system developed at Dresser-Rand using Bezier polynomials in cylindrical coordinates. A discussion of the basic techniques utilized in the code is presented as are sample graphic outputs generated to aid the user in the design process. The paper also describes some of the output options and how results may be interfaced with other analytical, drafting, and manufacturing software. Comments are included regarding the increased productivity, accuracy, and quality which resulted directly from use of this code and its support routines.

Commentary by Dr. Valentin Fuster
1990;():V001T01A078. doi:10.1115/90-GT-231.
FREE TO VIEW

A comparison is made between the flow in two impellers, one with radially ending blades and one with blades backswept by 30°. The two impellers have identical inducers. Measurements are made of the three velocity components and total pressures across five measurement stations within each impeller. The flow in the backswept impeller is dominated by a counter-clockwise vortex which reduces the severity of the shroud boundary layer separation and hence leads to a higher impeller efficiency. The wake is consequently smaller in the backswept impeller but adopts a similar position on the shroud surface at the impeller exit. Analysis of the secondary flow generation reveals the mechanisms responsible for the differences in the flow fields in the two impellers.

Commentary by Dr. Valentin Fuster
1990;():V001T01A079. doi:10.1115/90-GT-232.
FREE TO VIEW

To investigate the three-dimensional unsteady flow and the turbulence intensities behind rotating blade rows of turbomachines, a procedure using a fast-response pressure probe has been developed. The integration of the cylindrical miniature pressure transducers into the probe head minimizes the risk of mechanical damage. The dynamic behaviour of the probe was analyzed. The application of the probe to the rotor exit flow of an axial compressor is described and results are presented.

Topics: Pressure , Turbulence , Probes
Commentary by Dr. Valentin Fuster
1990;():V001T01A080. doi:10.1115/90-GT-233.
FREE TO VIEW

The paper describes experimental results obtained using laser velocimetry in a small high speed centrifugal impeller. The formation of wakes and the effect of varying speed and mass flow rate on the flow within the impeller passages are presented. In addition, an indication of the three dimensional nature of the impeller flow is discussed (the three dimensional results being obtained using a novel Doppler anemometer).

Commentary by Dr. Valentin Fuster
1990;():V001T01A081. doi:10.1115/90-GT-234.
FREE TO VIEW

A prediction of the three-dimensional turbulent flow in the NASA Low-Speed Centrifugal Compressor Impeller has been made. The calculation was made for the compressor design conditions with the specified uniform tip clearance gap. The predicted performance is significantly worse than that predicted in the NASA design study. This is explained by the high tip leakage flow in the present calculation and by the different model adopted for tip leakage flow mixing. The calculation gives an accumulation of high losses in the shroud/pressure-side quadrant near the exit of the impeller. It also predicts a region of meridional backflow near the shroud wall. Both of these flow features should be extensive enough in the NASA impeller to allow detailed flow measurements, leading to improved flow modelling. Recommendations are made for future flow studies in the NASA impeller.

Commentary by Dr. Valentin Fuster
1990;():V001T01A082. doi:10.1115/90-GT-235.
FREE TO VIEW

This paper is concerned with the design of a high speed, 5 inch diameter radial-inflow turbine for medium-sized diesel engine turbocharger applications. The turbine was designed by a newly developed fully three dimensional compressible inverse design method, in which the blade shapes are computed for a specified distribution of Display FormularV¯θ. The designed blades had non-radial blade filaments and therefore the impeller was carefully analysed for its structural integrity. This was achieved by the iterative use of a three dimensional structural and vibration analysis program and the design method.

The impeller was made by a casting process. The performance of the new impeller was measured and then compared with three other impellers, one conventional and two experimental. The new impeller performed substantially better than all the baseline turbines and showed a 5.5% improvement in the total-to-static efficiency over the conventional turbine, 2.5% of which was attributable to the aerodynamically superior blade shape computed by the three dimensional inverse design method. The improvement in efficiency was not just confined to the design point and an appreciable improvement could be observed at off-design conditions.

Commentary by Dr. Valentin Fuster
1990;():V001T01A083. doi:10.1115/90-GT-236.
FREE TO VIEW

Performance calculations for a NASA controlled-diffusion compressor blade have been carried out with a coupled inviscid-boundary layer code and a time-marching Navier-Stokes solver. Comparisons with experimental test data highlight and explain the strengths and limitations of both these computational methods. The boundary layer code gives good results at and near design conditions. Loss predictions however deteriorated at off-design incidences. This is mainly due to a problem with leading edge laminar separation bubble modelling; coupled with an inability of the calculations to grow the turbulent boundary layer at a correct rate in a strong adverse pressure gradient. Navier-Stokes loss predictions on the other hand are creditable throughout the whole incidence range, except at extreme positive incidence where turbulence modeling problems similar to those of the coupled boundary layer code are observed. The main drawback for the Navier-Stokes code is the slow rate of convergence for these low Mach number cases. Plans are currently under review to address this problem. Both codes give excellent predictions of the blade surface pressure distributions for all the cases considered.

Commentary by Dr. Valentin Fuster
1990;():V001T01A084. doi:10.1115/90-GT-237.
FREE TO VIEW

A cascade design-method is presented which complements the meridional through-flow design procedure of turbomachines. Starting from an axisymmetric flow field and the streamline geometry in the meridional plane this simple method produces a solution for the quasi three-dimensional flow field and the blade-element geometry on corresponding stream surfaces. In addition, it provides intra-blade data on loss and turning required for a consistent design and a convenient means of optimizing blade loading. The purpose of this paper is to describe the theoretical basis of the method and to illustrate its application in the design of transonic compressors.

Topics: Compressors , Design , Inflow
Commentary by Dr. Valentin Fuster
1990;():V001T01A085. doi:10.1115/90-GT-238.
FREE TO VIEW

In this paper, the development leading to a 17-stage axial flow compressor (pressure ratio 14.7) for the 25 MW class heavy duty gas turbine H-25 is described. In the course of developing the H-25’s compressor, extensive measurements were carried out on models. Experimental results are compared with predicted values. Aerodynamic experiments covered the measurements of unsteady flows such as rotating stall and surge as well as the steady-state performance of the compressor.

Based on the results of these tests, the aerodynamic and mechanical design parameters of the full scale H-25 compressor were finalized on the basis of two model compressors.

Detailed measurements of the first unit of the H-25 gas turbine were carried out. Test results on the compressor are presented and show the achievement of the expected design targets.

Commentary by Dr. Valentin Fuster
1990;():V001T01A086. doi:10.1115/90-GT-253.
FREE TO VIEW

The influence of periodically unsteady perturbations on the turbulent flow along the suction side of turbine blades is investigated in a test rig. The blade suction side is represented by a flat plate of 550 mm length. The pressure profile typically encountered in a turbine blade channel is generated by a curved wall opposite to the flat plate. The angle of the divergent part of the test channel and hence the pressure can be increased to induce flow separation on the flat plate. For simulation of the wakes from the upstream blade row the incoming flow is periodically disturbed by a wake generator consisting of five flat profiles arranged in front and parallel to the plate rotating with adjustable speed and phase angle.

A LDV with high spatial resolution is used to measure averaged and fluctuating components of the velocity inside the boundary layer flow down to a distance of y = 0,05 mm from the plate surface determining the boundary layer parameters as well as the wall shear stress. By Fourier analysis of the measured time related velocity distributions the stochastic and periodic parts of the overall turbulence are identified.

With a periodic wake flow the separation is shifted downstream as compared to the steady flow situation. This is due to the energization of the boundary layer flow associated with the conversion of periodic in stochastic parts of the turbulence. Conclusions resulting from the experimental findings for the theoretical understanding of the flow in turbine cascades are discussed in particular with respect to turbulence modelling.

Commentary by Dr. Valentin Fuster
1990;():V001T01A087. doi:10.1115/90-GT-258.
FREE TO VIEW

Velocity profiles were measured in the impeller of a centrifugal pump with a two directional laser velocimeter. Blade to blade profiles were measured at four circumferential positions and four radii within and one outside the four bladed impeller. Data is presented herein at two circumferential and three radial locations. The pump was tested in two configurations; with the impeller running centered within the pump, and with the impeller orbiting with a synchronous motion (ϵ/r2). Variation in velocity profiles among the individual passages in the orbiting impeller were found. At design flow rate, these variations ranged from 30 to 60 percent for the radial component, and 15 to 25 percent for the tangential component. Tangential velocity profiles near the impeller exit (r/r2 = 0.973) were near uniform across each individual passage. Differences in the magnitude of the exit tangential velocities among the passages, however, were detected. Systematic differences in the velocity profile shapes of the centered and orbiting impellers were in general not measured, the only exception being at r/r2 = 0.973 at 40% of the design flow rate. At this condition, two distinct radial velocity profiles were measured. Two of the impeller passages of the orbiting impeller contained a recirculation region covering 20–30% of the blade passage while the other two passages contained no recirculation region. The centered impeller also contained this region of reverse flow. Finally, velocity data was numerically integrated to find the forces and stiffnesses due to momentum fluxes on the impeller for the orbiting condition.

Commentary by Dr. Valentin Fuster
1990;():V001T01A088. doi:10.1115/90-GT-259.
FREE TO VIEW

The fluid dynamics of turbomachines are extremely complex, due in part to the aerodynamic interactions between rotors and Stators. It is necessary to acquire fluid dynamic data that reflect the interactive nature of a turbomachine to correlate with the fluid dynamics predicted from modern analyses. The temporal and spatial variations in the midspan aerodynamics of the second stage of a two-stage compressor have been studied with a two-component LDV system. Spatial variations were examined by traversing the LDV probe volume through a dense matrix of both axial and circumferential positions while temporal resolution was achieved by acquiring all data as a function of the instantaneous rotor position. Hence, the data set reveals rotor and Stator wake structure and decay in both the stationary and rotating frames of reference. The data also compared very favorably with extensive pneumatic measurements previously acquired in this compressor. In Part 2 of the paper, the data are used in the assessment of a prediction of the flow in the compressor using a time-accurate, thin-layer, two-dimensional Navier-Stokes analysis.

Commentary by Dr. Valentin Fuster
1990;():V001T01A089. doi:10.1115/90-GT-260.
FREE TO VIEW

A new zonal approach for computation of compressible viscous flows in cascades has been developed. The two-dimensional, Reynolds-averaged Navier-Stokes equations are discretized spatially by a cell-centered finite volume formulation. In order to make the present approach robust, the inviscid fluxes at cell interfaces are evaluated using a highly accurate TVD scheme based on the MUSCL-type approach with the Roe’s approximate Riemann solver. The viscous fluxes are determined in a central differencing manner. To simplify the grid generation, a composite zonal grid system is adopted, in which the computational domain is divided into non-overlapping zones, and structured grids are generated independently in each zone. The zonal boundary between two zones is uniquely defined by cell interfaces of one zone, which ensures the uniqueness of the zonal boundary. The communication from one zone to the other is accomplished by numerical fluxes across the zonal boundary. It should be noted that the complete conservation of the numerical fluxes across the zonal boundary can be satisfied by directly evaluating the numerical fluxes using the finite volume method and by ensuring the uniqueness of the zonal boundary. In order to demonstrate the versatility of the present zonal approach, numerical examples are presented for viscous flows through a transonic turbine cascade.

Commentary by Dr. Valentin Fuster
1990;():V001T01A090. doi:10.1115/90-GT-261.
FREE TO VIEW

A fully elliptic, control volume solution of the two-dimensional incompressible Navier-Stokes equations for the prediction of cascade performance over a wide incidence range is presented in this paper. The numerical technique is based on a new pressure substitution method. A Poisson equation is derived from the pressure weighted substitution of the full momentum equations into the continuity equation. The analysis of a double circular arc compressor cascade is presented, and the results are compared with the available experimental data at various incidence angles. Good agreement is obtained for the blade pressure distribution, boundary layer and wake profiles, skin friction coefficient, losses and outlet angles. Turbulence effects are simulated by the Low-Reynolds-Number version of the k-ε turbulence model.

Commentary by Dr. Valentin Fuster
1990;():V001T01A091. doi:10.1115/90-GT-262.
FREE TO VIEW

A numerical solution procedure, which includes a locally implicit finite volume scheme and an adaptive mesh generation technique, has been developed to study airfoil and cascade flows. The Euler/Navier-Stokes, continuity and energy equations, in conjunction with Baldwin-Lomax model for turbulent flow, are solved in cartesian coordinate system. To simulate physical phenomena efficiently and correctly, a mixed type of mesh, the unstructured triangular cell for the inviscid region and structured quadrilateral cell for the viscous, boundary layer and wake regions, is introduced in this work. The inviscid flows passing through a channel with circular arc bump, and the laminar flows over a flat plate with/without shock interaction are investigated to confirm the accuracy, convergence and solution-adaptibility of the numerical approach. To further prove the reliability and capability of the present solution procedure, the inviscid/viscous results for flows over the NACA 0012 airfoil, NACA 65-(12)10 compressor and one advanced transonic turbine cascade are compared to the numerical and experimental data given in related papers and reports.

Commentary by Dr. Valentin Fuster
1990;():V001T01A092. doi:10.1115/90-GT-263.
FREE TO VIEW

The boundary layers of a transonic turbine blade have been measured in detail. The full velocity profiles have been measured at a number of stations on both the suction and pressure surfaces, at conditions representative of engine operation, using a pitot traverse technique and a large scale (300 mm chord) linear cascade. This information has made it possible to follow the development of the boundary layers, initially laminar, through a region of natural transition to a fully developed turbulent layer. Comparisons with other, less detailed, measurements on the same profile using pitot traverse and surface mounted thin films confirm the essential features of the boundary layers.

Commentary by Dr. Valentin Fuster
1990;():V001T01A093. doi:10.1115/90-GT-303.
FREE TO VIEW

The aim of this paper is to compare viscous predictions for two turbomachinery geometries performed with a simple mixing length turbulence model and a more advanced transport equation approach. The mixing length approach is simple and crude, but cheap; more sophisticated transport schemes are more physical but more expensive. This paper compares the performance of two standard models, in one host code, with experimental data for axial turbine secondary flow development and for axial compressor blade loss-incidence variation. Although the more sophisticated turbulence model does produce detail improvements in the quality of the predictions, as far as loss generation is concerned it is concluded there is little difference between the two models. The reasons for this are discussed in some detail.

Commentary by Dr. Valentin Fuster
1990;():V001T01A094. doi:10.1115/90-GT-304.
FREE TO VIEW

The three-dimensional viscous transonic time marching Denton code L0SS3D, and Dawes code BTOB3D are applied to the first stage rotor of the NACA 5-stage transonic compressor. Computing time per solution on a mini-supercomputer was about 9 hours for a mesh of 65 000 points. LOSS3D predicted pressure ratio and loss distributions reasonably well at design point, but did not quite satisfy the convergence criteria. BT0B3D tended to overpredict the total pressure ratio over the outer half of span due to an underprediction of loss in the complicated separated flow region triggered by shock boundary layer interaction on the suction surface, but prediction was good at 90% speed where shock boundary layer interaction was less severe. The use of a computationally convenient excessively large tip clearance is not recommended when shock-boundary layer interaction is expected, especially at off-design conditions.

Topics: Compressors , Rotors
Commentary by Dr. Valentin Fuster
1990;():V001T01A095. doi:10.1115/90-GT-310.
FREE TO VIEW

A study of the propagation of a Nozzle Guide Vane (NGV) trailing edge shock wave through a transonic turbine rotor passage is presented. The work was based on experimental tests carried out in the Isentropic Light Piston Tunnel in Oxford University using a rotating bar NGV shock wave simulator, together with schlieren photography and wide band width surface pressure and heat transfer rate measurements.

The study identifies a previously unexplained interaction between the incoming wave and the rotor leading edge, which causes the nucleation of a Vortical Bubble. This bubble has been shown to enhance the thermal loading on the early pressure surface of the blade. A method of controlling this bubble and heat loading is also considered.

A previously unseen “Lambda” interaction between the shock wave and the rotor pressure surface is also identified.

Commentary by Dr. Valentin Fuster
1990;():V001T01A096. doi:10.1115/90-GT-311.
FREE TO VIEW

A simple numerical method for predicting the profile loss of turbine blades in subsonic and transonic flows is presented. A time marching Euler solver is used to obtain the main flow through the blade passages, the loss due to the surface friction is calculated using an integral boundary layer method, the total mixed out loss is evaluated from the mass flow and momentum balances between the trailing edge plane and an imaginary downstream plane where the flow is uniform. The base pressure acting on the trailing edge of the blade is calculated directly from the inviscid calculation without empirical correlations. The spurious numerical loss in the Euler calculation is separated from the real loss. The rationality of the approach is justified by the agreement of the prediction with a wide range of experimental measurements.

Commentary by Dr. Valentin Fuster
1990;():V001T01A097. doi:10.1115/90-GT-312.
FREE TO VIEW

In shroudless axial turbines the flow over the tips of the rotor blades is complex and accounts for significant loss of efficiency. In order to investigate the structure of this overtip flow, a row of high frequency response miniature pressure transducers was mounted in the casing of a cold flow turbine rig in the region swept by the rotor tips.

Commentary by Dr. Valentin Fuster
1990;():V001T01A098. doi:10.1115/90-GT-313.
FREE TO VIEW

The paper presents results from aerodynamic tests on a transonic rotor turbine profile cascade, including interferograms of the flow field and aerodynamic data measured downstream from the cascade. The dimensions and aerodynamic data on the cascade are given in detail. Analysis following the experimental data collection was aimed at investigating the sensitivity of transonic flow in the vicinity of the throat and related conditions leading to the compression effect. Further, the development of the flow structure and the cascade parameters over a wide range of exit Mach numbers, as well as incidence angles are shown. The experimental data are compared with results of calculations based on mathematical models.

Commentary by Dr. Valentin Fuster
1990;():V001T01A099. doi:10.1115/90-GT-314.
FREE TO VIEW

A recent survey of the literature showed a clear need for additional experimental results on the off-design performance of turbines, particularly for airfoils of recent design. This study presents measurements of the low-speed two-dimensional performance of a linear cascade of turbine blades with a turning angle of 87 degrees. The incidence was varied between −25 and +25 degrees in 5 degree steps. The blade surface pressures, total pressure loss coefficients and trailing-edge deviations are presented for all values of incidence. The influence of incidence on the critical Reynolds number is also examined. Surface flow visualization is presented for different values of Reynolds number and incidence to aid in the physical interpretation of the measurements. The measured total pressure losses agree very well with the new off-design correlation introduced by Moustapha et al. (1989).

Commentary by Dr. Valentin Fuster
1990;():V001T01A100. doi:10.1115/90-GT-327.
FREE TO VIEW

The evolution of the secondary flow field in cases of axial turbine bladings is examined in the present work.

Our secondary flow calculation method is successfully used, in an improved form including the complete meridional vorticity transport equation, to predict common viscous flow phenomena in axial turbines, such as high turning of the flow direction and jet-like velocity profiles.

The method used, although close to the parabolized Navier-Stokes approach, preserve several integral approximations thus minimizing the computer time required as compared to equivalent differential approaches. Our method provide fast and reliable tools for industrial applications with results comparable to those obtained by the differential solutions of the Navier-Stokes equations.

Several test cases, covering a wide deflection angle and acceleration ratio range, are analyzed. The results, concerning the secondary flow field, are very encouraging and Justify the choice of using approximate methods to estimate viscous phenomena in axial turbines.

Commentary by Dr. Valentin Fuster
1990;():V001T01A101. doi:10.1115/90-GT-352.
FREE TO VIEW

Extensive numerical analyses and experiments have been conducted to understand mixing phenomena in multistage, axial-flow compressors. For the first time in the literature the following are documented: detailed 3-D Navier-Stokes solutions, with high-order turbulence modeling, are presented for flow through a compressor vane row at both design and off-design (increased) loading; comparison of these computations with detailed experimental data show excellent agreement at both loading levels; the results are then used to explain important aspects of mixing in compressors. The 3-D analyses show the development of spanwise and cross-passage flows in the stator and the change in location and extent of separated flow regions as loading increases. The numerical solutions support previous interpretations of experimental data obtained on the same blading using the ethylene tracer-gas technique and hot-wire anemometry. These results, plus new tracer-gas data, show that both secondary flow and turbulent diffusion are mechanisms responsible for both spanwise and cross-passage mixing in axial-flow compressors. The relative importance of the two mechanisms depends upon the configuration and loading levels. It appears that using the correct spanwise distributions of time-averaged inlet boundary conditions for 3-D Navier-Stokes computations enables one to explain much of the flow physics for this stator.

Commentary by Dr. Valentin Fuster
1990;():V001T01A102. doi:10.1115/90-GT-353.
FREE TO VIEW

A new transient facility for the study of time mean and unsteady aerodynamics and heat transfer in a high pressure turbine has been commissioned and results are available. A detailed study has been made of aspects of the performance and behaviour relevant to turbine mechanical design, and an understanding of the variation of the turbine operating point during the test, crucial to the process of valid data acquisition, has been obtained.

In this paper the outline concept and mode of operation of the tubine test facility are given, and the key aerodynamic and mechanical aspects of the facility’s performance are presented in detail. The variation of those parameters used to define the turbine operating point during facility operation are examined, and the accuracy with which the turbine’s design point was achieved calculated. Aspects of the mechanical performance which are presented include the results of a finite element stress analysis of the loads in the turbine under operating conditions, and the performance of the rotor bearing system under these arduous load conditions. Both of these aspects present more information than has been available hitherto.

Finally, the future work programme and possible plans for further facility improvement are given.

Commentary by Dr. Valentin Fuster
1990;():V001T01A103. doi:10.1115/90-GT-354.
FREE TO VIEW

This paper discusses the mechanics of surge as observed on the high speed axial compressors of modern aero-engines. It argues that the initial stage of the instability consists of a high amplitude blast wave that develops non-linearly from a small scale disturbance and is thus not correctly described by traditional small perturbation stability theories. It follows from this that active control schemes of the global type may be inappropriate, since to be effective, control would have to be applied in a short time and in a very detailed manner, requiring a large number of transducers and actuators. Active control may, though, be effective in controlling the disturbances that grow into the above blast wave and in the control of other phenomena such as rotating stall, given an adequate number of transducers.

Topics: Compressors , Surges
Commentary by Dr. Valentin Fuster
1990;():V001T01A104. doi:10.1115/90-GT-355.
FREE TO VIEW

The purpose of this paper is to describe an investigation of the flow within and downstream of a turbine blade cascade of high aspect ratio. A detailed experimental investigation into the changes in the endwall boundary layer in the cascade (100deg camber angle) and total pressure loss downstream of the cascade was carried out. Flow visualisation was used in order to obtain detailed photographs of the flow patterns on the endwall and for exhibiting the trailing edge vortices. Pressure measurements were carried out using a miniature cranked Kiel probe for three planes downstream of the cascade, with two levels of turbulence intensity of the free-stream. Pressure distribution on the blade were measured at three spanwise locations, namely 4%, 12%, and 50% of the full-span from the wall. Hot wire anenometry combined with a spectrum analyser program was used to determine the frequencies of the flow oscillations.

The change in turbulence level of the free stream has a significant influence on all three pressure distributions. The striking difference between two of the pressure distributions is in the aft half of the suction side where the distribution with the lower turbulence intensity has the larger lift. The oil flow visualisation reveals what appears to be two separation lines within the passage and are believed to originate from the horseshoe vortex. The pitchwise-averaged total pressure loss coefficient increases with the distance of the measurement plane downstream of the cascade blades. A substantial part of this loss increase close to the wall is caused by the high rate of shear of the new boundary layer on the endwall.

Commentary by Dr. Valentin Fuster
1990;():V001T01A105. doi:10.1115/90-GT-356.
FREE TO VIEW

An experimental study reported in this paper was intended to acquire information on the distribution of wall shear stress and surface static pressure in a blade endwall corner. The blade endwall corner region investigated was divided into three sections: 0.4 chord length upstream of the blade leading edge, inside the endwall corner region, and one-chord length downstream of the blade trailing edge. Maximum increase in the values of wall shear stress were found to exist on the endwall, in the corner region, between the blade leading edge and the location of maximum blade thickness (≈140% maximum increase, compared to its far upstream value, at x/D=6). Surface flow visualization defined the boundaries of the vortex system and provided information on the direction and magnitude of the wall shear stress. The acquired results indicated that the observed variations of wall shear stress and surface static pressure were significantly influenced by the interaction of secondary flows with pressure gradients induced by the presence of blade curvature.

Commentary by Dr. Valentin Fuster
1990;():V001T01A106. doi:10.1115/90-GT-358.
FREE TO VIEW

This contribution deals with an experimental aero-thermal investigation around a highly loaded transonic turbine nozzle guide vane mounted in a linear cascade arrangement. The measurements were performed in the von Karman Institute short duration Isentropic Light Piston Compression Tube facility allowing a correct simulation of Mach and Reynolds numbers as well as of the gas to wall temperature ratio compared to the values currently observed in modern aero engines. The experimental programme consisted of flow periodicity checks by means of wall static pressure measurements and Schlieren flow visualizations, blade velocity distribution measurements by means of static pressure tappings, blade convective heat transfer measurements by means of platinum thin films, downstream loss coefficient and exit flow angle determinations by using a new fast traversing mechanism and freestream turbulence intensity and spectrum measurements. These different measurements were performed for several combinations of the freestream flow parameters looking at the relative effects on the aerodynamic blade performance and blade convective heat transfer of Mach number, Reynolds number and freestream turbulence intensity.

Commentary by Dr. Valentin Fuster
1990;():V001T01A107. doi:10.1115/90-GT-359.
FREE TO VIEW

The concept of swept blades for a transonic or supersonic compressor was reconsidered by Wennerstrom in the early 1980s. Several transonic rotors designed with swept blades have shown very good aerodynamic efficiency. The improved performance of the rotor is believed to be due to reduced shock strength near the shroud and better distribution of secondary flows.

A three-dimensional flowfield inside a transonic rotor with swept blades is analyzed in detail experimentally and numerically. A Reynolds-averaged Navier-Stokes equation is solved for the flow inside the rotor. The numerical solution is based on a high-order upwinding relaxation scheme, and a two-equation turbulence model with a low Reynolds number modification is used for the turbulence modeling. To properly predict flows near the shroud, the tip-clearance flow also must be properly calculated. The numerical results at three different operating conditions agree well with the available experimental data and reveal various interesting aspects of shock structure inside the rotor.

Topics: Compressors , Blades
Commentary by Dr. Valentin Fuster
1990;():V001T01A108. doi:10.1115/90-GT-373.
FREE TO VIEW

The prediction of unsteady flow in vibrating transonic cascades is essential in assessing the aeroelastic stability of fans and compressors. In the present work an existing computational code, based on the numerical integration of the unsteady Euler equations, in blade-to-blade surface formulation, is validated by comparison with available theoretical and experimental results. Comparison with the flat plate theory of Verdon is, globally, satisfactory. Nevertheless, the computational results do not exhibit any particular behaviour at acoustic resonance. The use of a 1-D nonreflecting boundary condition does not significantly alter the results. Comparison of the computational method with experimental data from started and unstarted supersonic flows, with strong shock waves, reveals that, notwithstanding the globally satisfactory performance of the method, viscous effects are prominent at the shock wave/boundary layer interaction regions, where boundary layer separation introduces a pressure harmonic phase shift, which is not presicted by inviscid methods.

Commentary by Dr. Valentin Fuster

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In