The annulus flow area in a multistage compressor is designed based on the calculated density rise along the axis of the machine operating at the design point. The assumption of constant axial velocity, for example, implies that the channel area shrinks inversely proportional to the density rise, to conserve mass, namely,
This behavior is schematically shown in Figure 8.73.
In the starting phase of a compressor, the mass flow rate is initially small, which means a higher loading for the front stages. The picture for the aft high-pressure stages is just the opposite of the front low-pressure stages. Initially, the compressor does not develop the density rise that it was designed to produce, hence with lower than design densities in the aft stages the axial velocity is increased to satisfy the continuity equation. Higher than the design axial velocity leads to a lower loading of the aft stages. Actually the aft stages would be windmilling and their blades operating in a separated flow. A comparison of the velocity triangles in the starting phase and the design point will help us understand the (starting) problem in a high-pressure compressor. Figure 8.75 shows the velocity triangles for the front and aft stages of a compressor. Note that the flow coefficient is lower than the design value for the front stages, and the flow coefficient is higher for the aft stages. We remember from the compressor performance map that with a reduction of flow coefficient the compressor can go into stall for a given shaft speed. With front stages not producing the design density ratio, the axial flow in the aft stages accelerates to conserve mass. Hence, the two ends of the compressor see the opposite flow fields and are subjected to opposite flow coefficients, that is, lower and higher than the design, respectively. However, the tendency for both front and aft stages is to stall in the starting phase of a high-pressure compressor.
The loading mismatch between the front and aft stages in the starting phase of a high-pressure axial compressor may be solved in several ways. Here we introduce three distinct approaches to the starting problem.
Proposition 1: Split shaft (or multispool shaft system)
Let the aft stages operate at a higher rotational speed than the front stages. This proposition aims at matching the relative flow angle to the rotor blades in the aft stages. Pratt & Whitney spearheaded the development of the split shaft concept in the United States in the 1950s. Today, all modern high-pressure compressors employ a two-to-three spool or shaft configuration to alleviate the problem of starting and improve compressor efficiency and stability. This method primarily attacks the aft stages, which would have been windmilling without a higher shaft speed, as in a single-shaft gas turbine engine. Rolls-Royce is the only engine company that has manufactured a production three-spool commercial engine, known as the RB 211, with several successful derivatives.
Proposition 2: Variable stators
This proposition aims at adjusting the absolute flow angle (α) through a stator variable setting to improve the mismatch between the flow and the rotor relative flow angle β. GE-Aircraft Engines spearheaded the development of high-pressure compressors with the variable stator approach in the United States, also in 1950s. Today, modern high-pressure compressors use variable stators for the front several stages to help with the starting/off-design operation efficiency. This method primarily attacks the front stages. Although a variable stator benefits all stages of a compressor, its use is limited to the front few (say six) low-pressure stages as the sealing of hot gases through the variable stator seal poses an operational and maintenance problem on the high-pressure end of the compressor.
Proposition 3: Intercompressor bleed
The windmilling operation of the aft stages is the result of high axial velocities through those stages. To cut down on excessive axial velocities, we may bleed off some mass flow in the intermediate stages of the compressor, the so-called intercompressor bleed. Therefore in the starting phase, bleed ports need to be opened to help the high-pressure end of the compressor to operate properly. This proposition represents a relatively low-cost method of starting compressors that does not employ either a split shaft or the variable stators. Stationary gas turbine power plants employ the intermediate bleed solution as a method of starting the compressor. This method attacks the aft as well as the front stages, as the bleed ports are located in the middle. Therefore, mass withdrawal causes a reduction in the axial velocity to the aft stages as well as lowering the backpressure for the front stages, hence increasing the flow speed in the machine. The ability to tailor the flow through the engine to improve overall efficiency, component stability, and provide cooling to accessories and engine components all speak in favor of intercompressor bleed. In fact, all modern gas turbine engines today employ intercompressor bleed/flow control. These methods are summarized in Figure 8.75.
Aircraft gas turbine engines operate downstream of an air intake system. The level of distortion that an inlet creates at the compressor face affects the performance and the stability of the compressor. First, what do we mean by distortion? In simple terms, distortion represents nonuniformity in the flow. The nonuniformity in total pressure as in boundary layers and wakes, the nonuniformity in temperature as in gun gas ingestion or thrust reverser flow ingestion and the nonuniformity in density, as created by hot gas ingestion are some of the different types of distortion. The common feature of all different types of distortion is found in their destabilizing impact on the compressor performance. This means that all distortions reduce the stability margin of a compressor or fan, potentially to the level of compressor stall or the engine surge. The types of distortion are
The most common type of inlet distortion is the total pressure distortion that is caused by separated boundary layers in the inlet. Under normal operating conditions, the boundary layers in the inlet are well behaved and remain attached. However, if the boundary layer management system in a supersonic inlet, as in the bleed system, fails to react to an abrupt change in the flight operation (potentially due to a rapid combat maneuver), the flowfield at the engine face will contain large patches of low-energy, low-momentum flow that could cause flow separation in the front stage(s) of the fan or compressor. In describing the total pressure distortion and its impact on compressor performance, we divide the spatial extent of the spoiled flow according to its radial and circumferential extent, as shown in Figure 8.76.
These inlet total pressure distortion patterns may be simulated by installing screens of varying porosity upstream of the compressor or fan (typically one diameter upstream) in propulsion system ground test facilities. The distortion patterns generated by screens in a test set up in ground facilities represent the “steady-state” component of the distortion and thus lack the “dynamic” or transitory nature of the distortion encountered in real flight environment. The full description of distortion requires both the steady-state and the dynamic components, as in the study of turbulent flow requiring a mean and an rms level of the fluctuation.
The results of a NASA-Glenn 10 × 10 supersonic wind tunnel study on the response of a simple turbojet engine (J-85) to steady-state inlet total pressure distortion are shown in Figure 8.77 (from Povolny et al., 1970).
The normal operating line, the undistorted stall boundary, and two corrected shaft speeds of 100% and 93% design are shown in dashed lines. Four solid lines correspond to the stall boundaries of the four distortion patterns simulated at the compressor face via screens. From having the least to the most impact on the stall margin deterioration, we identify the culprits as (1) radial hub, (2) radial tip, (3) circumferential hub, and (4) the full-span circumferential distortion, respectively. We also note that the full-span distortion at the 100% corrected speed operates at the stall boundary, that is, zero-stall margin! Further research identified a critical circumferential extent of the spoiled sector that causes the maximum loss in the stall pressure ratio of a compressor is at nearly 60°, as evidenced in Figure 8.78 (from Povolny et al., 1970).
The loss in stall pressure ratio with circumferential inlet distortion reaches to ∼10% at 100% corrected speed, as shown in Figure 8.78. At higher shaft speeds, the incidence angle in the spoiled sector is larger than at the lower shaft speed, thus the trend of higher loss of the stalling pressure ratio at higher shaft speeds becomes evident using simple velocity triangle arguments.
The temperature distortion also leads to a reduction in stall margin. In general, static temperature distortion in a flow brings about density nonuniformity, which creates a nonuniform velocity field. Consequently, it is impossible to create a static temperature distortion without creating other forms of nonuniformity, for example, density, velocity, total pressure, in the flow. To quantify the impact of a spatial temperature distortion on engine stall behavior, NASA researchers have conducted experiments with representative data shown in Figure 8.79 (from Povolny et al., 1970). The undistorted operating line, stall limit, and different shaft speeds are shown in dashed lines. Data points corresponding to the effect of temperature distortions of 45–120°F on the stall behavior of a variable geometry turbofan engine high-pressure compressor are plotted in solid lines. The circumferential extents of the temperature distortions were 90 and 100° in different tests. A 100°F distortion of ∼90–100° circumferential extent is seen to stall the high-pressure compressor operating at its 90% corrected flow.
The inadequacy of steady-state distortion simulation in a wind tunnel is best seen in Figure 8.80 showing F-111 flight test data of compressor stall with TF-30 engine. A steady-state distortion parameter KD is graphed at different corrected airflows through the engine, and a band of maximum allowable distortion based on previous wind tunnel tests is shown to miss the mark of in-flight compressor stall data by a large margin (from Seddon and Goldsmith, 1985). The culprit is identified as the dynamic distortion.
The steady-state distortion parameter KD expresses a weighted average of engine face total pressure distortion pattern (circumferential and radial) recorded by an engine face rake system composed of several radial pitot tube measurements. For a mathematical description of KD and other distortion parameters, Hercock and Williams (1974) may be consulted. In addition, Farr and Schumacher (1974) present evaluation methods of dynamic distortion in F-15 aircraft. Schweikhand and Montoya (1974) treat research instrumentation and operational aspects in YF-12. These references are recommanded for further reading.
To meet the real world, engineers need to spend time with the real hardware. Isometric and cutaway views of the hardware provide for a three-dimensional feel but may not be a substitute for the cutaway of a real engine (or its components) in the laboratory. The real engine, albeit old and surplus, shows real fasteners, components assembly, manufacturing tolerances; some operational degradation, for example, wear, on the rotor blade tips, shows labyrinth seals between rotating and stationary parts; it shows signs of erosion in the compressor and turbine and perhaps corrosion in the hot section. A visual inspection and a feel of the compressor and turbine blades show the turbine blades receive heavy deposits from the combustor (and its fuel additives) and give a new meaning to “surface roughness, ” which will not feel “hydraulically smooth.”
Having pointed out some of the advantages of real hardware, let us examine some isometric and cutaway views of axial-flow compressors, designed, built, tested, and flown on aircraft in this section. Figure 8.81 is taken from a manual by Pratt & Whitney Aircraft (1980), whereas Figures 8.82–8.84 are taken from Rolls-Royce’s The Jet Engine (2005). Figure 8.81 shows a two-stage fan and the low-pressure compressor rotor. Mid-span shrouds prevent the first bending mode of the first fan rotor blades. The anti-icing air valve and the actuator point to an operational need of the aircraft (i.e., flight under icing condition) and demands on the compressor air.
In this section, we present design guidelines that are useful in the preliminary design of axial-flow compressors. The approach to turbomachinery design in textbooks is neither unique nor exact. In general, we use the lessons learned from our predecessors and the invaluable contributions of NASA and open literature, which often come from academia and industry. One possible approach to compressor design is outlined in steps that are particularly useful for students who want to learn and practice their turbomachinery design skills.
We propose to design an axial-flow compressor for a design-point mass flow rate and pressure ratio. Here, we first review the steps and in the next section we design an axial compressor at the pitchline. We start the design process by choosing
The centrifugal force is the integral of mv2/r, which is mω2r, that is,
where ρblade is blade material density and A(r) is the blade cross-sectional area as a function of span.
The blade area distribution along the span, Ab(r)/Ah, is known as taper and is often approximated to be a linear function of the span. Therefore, it may be written as
We may substitute Ab(r)/A in the integral and proceed to integrate; however, a customary approximation is often introduced that replaces the variable r by the pitchline radius rm. The result is
The taper ratio At/Ar is ∼0.8–1.0. Therefore, the ratio of centrifugal stress to the material density is related to the square of the angular speed, the taper ratio, and the flow area, A = 2πrm(rt − rh). Equation 8.206 is the basis of the so-called AN2 rule, where the RHS is related to the size (i.e., A) of the machine and the square of the angular speed (i.e., N2), and the LHS is related to material property known as specific strength.
The material parameter of interest in a rotor is the creep ruputure strength, which identifies the maximum tensile stress tolerated by the material for a given period of time at a specified operating temperature. Based on the 80% value of the allowable 0.2% creep in 1000 h for aluminum alloys and the 50% value of the allowable 0.1% creep in 1000 h for other materials, Mattingly, Heiser, and Pratt (2002) have graphed Figures 8.86 and 8.87. They show the allowable stress and the allowable specific strength of different engine materials as a function of temperature
Bending stress is, therefore, proportional to gas pressure, flow coefficient at the tip, stage total temperature ratio, and to the square of tip radius-to-thickness ratio. Bending stress is also inversely proportional to solidity
For blade design, we need to estimate the following angles:
The incidence angle is estimated at the location of minimum loss, from cascade data that is known as ioptimum. Typically, optimum incidence angle is ∼−5° to +5°.
Iteration loop
The deviation angle is first estimated from the simple Carter’s rule:
which is based on the net flow turning and the coefficient “m” in Carter’s formula is taken to be 0.25.
Also, based on the inlet and exit flow angles, we may use the (NACA 65-series) cascade correlation data of Mellor (Appendix I) to identify a suitable 65-series cascade geometry that gives an adequate stall margin. Note that the front and aft stages of a multistage compressor or fan face different stall challenges; for example, the front stages are more susceptible to positive stall and the aft stages are more susceptible to negative stall. Therefore, we may choose the design point to be farther away from the two stall boundaries for front and aft stages. Figure 8.88 is a definition sketch for the design point selection.
The optimum incidence angle is defined (by Mellor) to be at the point of lowest profile loss, although some authors prefer the optimum incidence to correspond to the angle where lift-to-drag ratio is at its maximum. Here, we adopt Mellor’s definition, due to its ease of use. The cascade loss data, as shown in Figure 8.22, can be used to arrive at the iopt.
Johnsen and Bullock (1965) in NASA SP-36 outline optimum incidence estimation based on cascade data correlations for different profile types and profile thicknesses. This reference may be used for more accurate estimation of the incidence angle. From the inlet flow and an estimate of the optimum incidence angle, we calculate the leading-edge angle on the mean camber line (MCL), κ1, and from the exit flow angle and deviation we calculate the blade angle at the T.E.
The iteration loop on the deviation angle may be initiated at this point, since we have calculated a camber angle and we may now use the expanded version of the Carter’s rule:
Although we have introduced a possible iteration loop on the deviation angle, in the preliminary stage of the compressor design it is acceptable to use the simple formula of Carter for deviation.
The supersonic section is designed based on
These designs are described in Cumpsty (1989), Schobeiri (2004), and Wilson and Korakianitis (1998).
The annulus flow area shrinks inversely proportional to the rising fluid density for a constant axial velocity, that is,
We can calculate density rise per stage (at the pitchline) from the pressure and temperature rise and the use of the perfect gas law. This is an approximate method, but is adequate for the preliminary design purposes.
Common practices are
Compressor stall margin may be established based on the use of Koch equivalent diffuser model of the compressor blade passage. This is rather laborious, but produces good results. We start with the equivalent enthalpy rise coefficient Ch (for notation, see section 8.10).
Note that the difference between U1 and U2 across the rotor (in the numerator) is due to streamline shift in the radial direction. We further correct for Reynolds number, tip clearance gap, and the axial blade spacing other than the nominal values that are used in Koch’s stall margin correlations. The correction factors are in Figures 8.64–8.66. Finally, we correct the adjusted enthalpy rise coefficient for the effect of stagger and the velocity vector diagram (wake effect) to arrive at the effective static enthalpy rise coefficient according to
The critical stalling pressure rise is shown in Figure 8.73 as “stall line” or the case of 0% stall margin.
Guidelines on the Range of Compressor Parameters | ||
Parameter | Range of values | Typical value |
Flow coefficient | 0.3 ≤ ≤ 0.9 | 0.6 |
D-Factor | D ≤ 0.6 | 0.45 |
Axial Mach number Mz | 0.3 ≤ Mz ≤ 0.6 | 0.55 |
Tip Tangential Mach Number, MT | 1.0–1.5 | 1.3 |
Degree of reaction | 0.1 ≤ °R ≤ 0.90 | 0.5 (for M < 1) |
Reynolds number based on chord | 300, 000 ≤ Rec | >500, 000 |
Tip relative Mach number (1st Rotor) | (M1r)tip ≤ 1.7 | 1.3–1.5 |
Stage average solidity | 1.0 ≤ σ ≤ 2.0 | 1.4 |
Stage average aspect ratio | 1.0 ≤ AR ≤ 4.0 | <2.0 |
Polytropic efficiency | 0.85 ≤ ec ≤ 0.92 | 0.90 |
Hub rotational speed | ωrh ≤ 380 m/s | 300 m/s |
Tip rotational speed | ωrt∼450—550 m/s | 500 m/s |
Loading coefficient | 0.2 ≤ ψ ≤ 0.5 | 0.35 |
DCA blade (range) | 0.8 ≤ M ≤ 1.2 | Same |
NACA-65 series (range) | M ≤ 0.8 | Same |
De Haller criterion | W2/W1 ≥ 0.72 | 0.75 |
Blade leading-edge radius | rL.E. ∼5–10% of tmax | 5% tmax |
Compressor pressure ratio per spool | πc < 20 | up to 20 |
Axial gap between blade rows | 0.23 cz to 0.25 cz | 0.25 cz |
Aspect ratio, fan | ∼2–5 | <1.5 |
Aspect ratio, compressor | ∼1–4 | ∼2 |
Taper ratio | ∼0.8–1.0 | 0.8 |
Axial-flow compressors are highly evolved machines that provide efficient mechanical compression (of the gas) in a gas turbine engine. The mechanical work is delivered to the medium by a set of rotating blade rows. The rotor blades impart swirl to the flow, thereby, increasing the total pressure of the fluid. In between the rotor blade rows are the stator blade rows that remove the swirl from the fluid, thereby increasing the static pressure of the fluid. The combination of one rotor row and one stator blade row is called a compressor stage. Since the primary flow direction is along the axis of the machine, staging the axial-flow compressor is easy. The flow in an axial-flow compressor is exposed to an adverse pressure gradient environment, that is, a climbing pressure hill that tends to stall the boundary layer. Therefore, the phenomenon of stall is experienced in a compressor. The stalled compressor flow is inherently unsteady and thus may cause a system-wide instability (between the compressor and combustor) known as surge.
In analyzing the flow in the rotor, it is most convenient to use the rotating frame of reference that spins with the rotor, which is known as the relative frame of reference. The flow in stator blade row is best analyzed by a stationary observer (i.e., a frame of reference that is fixed with respect to the casing) known as the absolute frame of reference. The velocity vectors in the two frames of reference form a triangle, described by
The Euler turbine equation is known as the fundamental equation in turbomachinery that relates the change of angular momentum Δ (rCθ) across a rotor blade row to the rotor specific work via
If we express the jump in swirl across a rotor in terms of the (rotor) inlet absolute and exit relative flow angles (that remain constant over a wide operating range of the compressor), we deduce from Equation 8.25a that
The significance of this equation is the appearance of U2 in front of the bracket on the RHS, which indicates that the total temperature rise across a rotor is proportional to the square ofthe wheel speed U2. Therefore a high blade Mach number is desirable, which is exactly the argument for the development of transonic compressors. A rule of thumb is that the blade tangential Mach number at the tip is slightly supersonic, MT ∼ 1.2–1.3. The structural limitations currently limit the blade tip Mach number to ∼1.5 in conventional (subsonic throughflow) compressors. A high strength-to-weight ratio material, such as titanium, is desirable for fan blade construction.
Since we have to limit the level and extent of the flow diffusion in a compressor blade row, we define a diffusion factor D for the rotor and stator blade rows according to
The experience shows that diffusion factor is to be limited to ∼.6 about the mid-span of a blade row and to ∼0.4 near the hub or the tip. The lower diffusion imposed on the hub and the tip is due to a complex hub boundary layer/corner vortex formation and the tip clearance flows that dominate the two ends of a blade row. We may view the diffusion factor as a blade-row-specific figure of merit. We define a stage-based figure of merit as well, which is called the degree of reaction °R,
Degree of reaction is the ratio of the rise of the static enthalpy across the rotor to that of the stage. In essence, it speaks to the rotor’s share of the static pressure rise to that of the stator. Therefore, a 50% degree of reaction stage equally divides the burden of the pressure rise to the rotor and the stator. Although 50% sounds desirable for equal burden, since the rotor blades spin their boundary layers are more stable and thus can withstand higher static pressure rise than the stationary stator blades. Consequently, a 60% degree of reaction may be more optimal than the 50%.
Cascade aerodynamics provides a rich background for a two-dimensional (subsonic) blade design. The minimum-loss incidence angle iopt and the cascade (total pressure) loss bucket are used for the preliminary 2D design of compressor blade sections as well as identifying the positive and negative stall boundaries. The limited supersonic cascade data make the preliminary design of the supersonic sections of transonic fans less grounded. The general rules are based in gas dynamics, which indicate the selection of the thinnest possible sections in order to minimize the wave drag. The thinnest (structurally feasible) section is ∼3% thick and the subsonic root is often ∼10% thick.
Three-dimensional design of blades, in the preliminary stage, is achieved by the so-called vortex design approach. Here we introduce a catalog of swirl profiles, such as free-vortex, solid-body rotation, and others that will be anchored at the pitchline and then give us a swirl profile along the span. We applied radial equilibrium theory to calculate an axial velocity profile. Three-dimensional losses are due to secondary flows, tip clearance, and the blade junction corner vortex. The unsteady flow losses are related to upstream wake chopping (by the downstream blade row) and the subsequent vortex shedding in the wake. Losses due to compressibility are minimized by thin profile designs that form the sections of swept blades.
A single-spool compressor is of limited capability in high-pressure compressor applications. The primary reason for that is the differing rotational requirement of the front versus the aft stages. This is the crux of the classical starting problem for a high-pressure ratio compressor. Multistage compressors are often driven by different shafts in order to spin the high-pressure stages faster than the low-pressure compressor. Variable stators in the front stages are always adjustable to help with the starting problem as well as the off-design operation of the compressor. A fraction of airflow in the compressor is bled at the intermediate and exit sections to cool the HPT, the casing, and the exhaust nozzle. The cooling fraction is ∼10–15% in modern engines.
As an upper bound, multispool, high-pressure ratio compressors can achieve a pressure ratio of ∼45–50. The exit temperature of ∼900 K is the limit of current materials for an uncooled compressor, which limit the compressor pressure ratio. The advances in computational fluid dynamics, parallel processing, and computer memory have elevated the design of compressors to be based on flow physics and with less reliance on empiricism and cascade data. The result has been the appearance of high-efficiency unconventional transonic blades and stages with forward and aft swept blades and lean. The advances in integrated manufacturing technology that involve super plastic forming and diffusion bonding (SPF/DB) are used for the manufacturing of modern (composite) wide-chord fan blades. New materials and manufacturing technology, e.g., BLISK, offer weight savings in compressors with the subsequent improvement on the engine thrust-to-weight ratio.
Dixon [7], Cheng et al. [2], Prince, Wisler and Hilvers [41], Whitcomb and Clark [53] are recommended for additional reading.
Assuming that Um1 = Um2 = 210 m/s and Cz1 = Cz2 = 175 m/s, ρ1 = 1 g/m3 β2 = −25°, and ϖr = 0.03, calculate
Note that the radius of this cut (section) is at r ≅ 0.5 m from the axis of rotation, as shown in the diagram.
Assuming the IGV has a total pressure loss coefficient of ϖIGV = 0.02, calculate
Knowing that the compressor stage has a degree of reaction of °R = 0.5 at the pitchline, calculate
For a rotor total pressure loss coefficient of r = 0.03 at the pitchline, calculate
Assume: γ = 1.4 and cp = 1004 J/kg · K throughout the stage.
for
Also, if the hub radius is constant and for a mass flow rate of 100 kg/s and the inlet total pressure of pt2 = 100 kPa, calculate the tip-to-tip radius ratio (r3/r2)tip
Now, assume that we chose a free-vortex design for the rotor.
Calculate
Assuming that the rotor has a free-vortex design, calculate
Estimate the ratio of bending stress to inlet pressure (σbending/p) in a rotor with axial velocity 165 m/s, the rotor tip radius is rt = 0.75 m, the mean solidity of 1.5, angular speed is ω = 4000 rpm. The maximum blade thickness is tmax = 1 cm. The stage total pressure ratio is 1.6 and the stage adiabatic efficiency is 0.88.
for a linear taper ratio. For a titanium rotor blade of rh = 0.4 m, rt = 0.8 m, and taper ratio At/Ah = 0.8, calculate the acceptable shaft speed ω if the allowable strength-to-weight ratio is ∼9 ksi/slug/ft3, which is equivalent to
allowable creep rupture strength required strength | 120, 560 m2/s2 (or ∼120 kPa/kg/m3) 68, 326 m2/s2 (or ∼68 kPa/kg/m3) |
Calculate this compressor’s stall margin.
Calculate:
[note: the axial velocity at the compressor exit, V3 = V2]
The gas properties are: γ = 1.4 and R = 287 J/kgK.
Calculate:
Assuming that the inlet total temperature is Tt1 = 288 K, γ = 1.4 and cp = 1.004 kJ/kg.K, calculate:
18.190.176.243