Turbopump

Hydrodynamic design The pump side of turbopumps consist of impellers that spin at very high speeds (thousands of RPM) in order to pump liquid propellants. Impellers are mounted on a central shaft, which also has a turbine mounted to it (or in some cases geared off on a different shaft). The turbine supplies shaft power, which is then consumed by the impellers in order to impart energy to the liquid propellants. Impellers mostly impart this energy by accelerating the liquid to a high velocity. However the ultimate goal is not a fast liquid, but a high pressure one; so surrounding the impeller is either a volute or a diffuser - these are specially shaped housings to decelerate the flow which then consequently dramatically increases its pressure (via Bernoulli's principle). The liquid is then discharged to the rest of the rocket engine, or in some cases to a second impeller and volute/diffuser stage which increases the pressure even further. Turbopumps on liquid rocket engines virtually always have inducers as well, upstream of the impellers. Inducers are spiral shaped pumping elements that serve to gently raise the pressure of the incoming fluid enough to prevent it cavitating when it reaches the impeller. In many cases the impeller and inducer are manufactured as a single component, with a gradual transition between the axial spiral and the radial blades. Aerodynamic design The turbine side of turbopumps consist of one or more stages, where each stage has a stator and a rotor. Individual rotor discs in a turbine are more commonly referred to as wheels in the modern day. These turbines are virtually always of the axial type, because of the very high gas flow (volumetrically) needed to supply enough shaft power for a liquid rocket engine. Contrast this with turbochargers, which usually feature radial turbine designs because of their much lower gas flow. Upstream of the turbine is the turbine manifold, which collects gas from whatever source that rocket engine's cycle has upstream of it, and then disperses it circumferentially along the rim of the turbine. It then flows from the manifold axially downwards to the stages of the turbine. Stators are typically bladed, though it is also quite common (where pressure drop is particularly high, as in gas generator cycles) to forgo blades and drill angled nozzles directly off of the manifold itself to then impinge on the turbine wheel. Downstream of the turbine varies based on cycle - in closed cycles it leads to the main injector of the engine, where (depending on whether the turbine is fuel-rich or ox-rich), one of the propellants can be injected into the main combustion chamber as a gas which can be very advantageous for promoting propellant atomization and mixing. In open cycles it is dumped to atmosphere. This can either mean dumped overboard directly off the side of the engine, or it can also lead to a manifold on the rocket engine nozzle which then injects it in the main flowpath, far downstream of the throat where ambient pressure is much lower than the chamber. The purpose here is to provide extra film cooling to the nozzle, since the hot gas leaving the turbine is nevertheless much cooler than the gas in the main combustion chamber. the latter option is common in vacuum optimized open cycle engines because they have much larger nozzles (with correspondingly large areas that need cooling, often without a regen jacket at its furthest extremes). It is important to note that the dumped gas from the turbine can still provide a non-negligible portion of the engine thrust. For this reason even if it is dumped overboard directly, there will usually still be a housing and a mild converging-diverging nozzle downstream of the turbine to take full advantage of the extra thrust opportunity. There is also an opportunity to extract waste heat from the flow at this point via heat exchangers; useful for heating up repressurizing gas for the tanks, for example. Cycle design Turbomachinery / engine cycle design looks very different in liquid rocket engines compared to air-breathing engines (turbojets) for essentially one main reason: turbine materials cannot survive combustion chamber temperatures. Rocket engine cycles are all various workarounds to this fundamental problem. The first ever turbopump designs (by Goddard, early 1930s) did partially put the turbine into the main combustion chamber flow with regenerative cooling. By the time of actual builds in the later 1930s, he had moved on to a rudimentary expander cycle, and then ultimately a gas generator cycle. The turbine of a turbopump is always driven by high pressure gas. The exact source of this gas is the primary differentiator between the various rocket engine cycles. Air-breathing engines (turbojets and similar) mount their turbine downstream of the burner and take direct advantage of the full flow and pressure of the engine. Rocket engines have never been able to do this because their mixture ratios are much closer to stoichiometric (since oxidizer comes at a premium; it must be carried with the rocket) and thus the flame temperature in the combustion chamber is dramatically higher. They are so high that nearly all possible materials would melt, and even the few that do have very little structural strength left at these temperatures. For this reason, rocket engine cycles are all various schemes to circumvent this and supply hot gas to the turbine that is nevertheless much cooler than the main combustion chamber gas. Gas generator and staged combustion cycles do this by mounting an entirely separate and smaller combustion chamber to the engine, termed the gas generator (whose gas is ultimately dumped overboard) or the preburner (whose "pre-burnt" gas eventually reaches the main combustion chamber after passing through the turbine). These smaller chambers run very far from stoichiometric, either with way too much fuel or way too much oxidizer. Hence, one can have "fuel rich" and "ox rich" gas generator and staged combustion cycles. One could also have two preburners, one fuel rich and one ox rich, which is termed "full flow staged combustion". Beyond these, there are also expander cycles, where liquid propellant is heated (usually fuel) in the regenerative cooling loop of the main combustion chamber, to the point of boiling, and then fed as gas to the turbine. The last major cycle is the tap off cycle, where a portion of the main combustion gas is "tapped off" and routed to the turbine. Because of the aforementioned temperature problem, tap off cycles require large dedicated heat exchangers to rapidly cool the re-routed gas before it reaches the turbine. Mechanical design The collection of all rotating components in a turbopump (i.e. the impellers, inducers, wheels, shaft, parts of the seals, and various spacers) are collectively known as the "rotor". The rotor is spinning at extreme angular velocities: shaft speeds in the tens of thousands of RPM are common. Nominally the only mechanical connection between the rotor and the rest of the turbopump is via the bearings. Most common by far are ball bearings, with some modern exceptions pivoting to hydrodynamic bearings. The goal of bearing selection is to minimize friction - both because high friction can wear out the bearing, and also because any frictional energy losses are dissipated as heat that must be carried away rapidly to not destroy the bearing. The extra challenge in turbopump design is that the local environment in the pumps is very often at cryogenic temperatures, which virtually all greases and oils normally used to lubricate bearings are not compatible with (they freeze). Therefore, turbopump bearings do not use lubricants at all in the traditional sense. Rather, they are installed as bare metal, and some amount of cold propellant is intentionally routed through them (i.e. where the balls are) to dissipate the heat generated by their friction. This bearing cooling circuit is a secondary flow that the hydrodynamic designer must also design in addition to the primary flow of the propellant through the inducer/impeller/volute. Turbopumps can be very sensitive to the exact placement of components and the loads/stresses developed in them. Hydrodynamic considerations typically demand very tight clearances between the impellers / inducers and the pump housings, as well as aerodynamic considerations demanding tight clearances between turbine wheels and stators / manifolds. Furthermore, rotordynamics demands a high stiffness coupling of the rotating components with the shaft, especially when it comes to the turbine wheel. These considerations and more demand high precision and high stiffness mechanical design. Bolted joints are generally the default method by which to join parts; some turbopumps have welded joints as well but require more careful consideration and analysis because of their generally lower stiffness, potential for thermally induced warpage of the parts during the welding process, as well as increased risk of fatigue over the life of the turbopump. In order for the rotor to act structurally as one rigid object, all of the components are stacked into one long stackup that envelops the entire shaft and then is preloaded onto it from both ends. This moderately loads the ball bearings, which are usually of the angular contact type, which increases their stiffness. Typically the preload is supplied one end by a bolt clamping onto the nose of the inducer and threaded into the end of the shaft below it. Depending on the exact configuration of the turbopump, the other end could be another inducer (for the other propellant), or a turbine wheel which will also have preloaded bolt(s) onto the end of the shaft. Design of the shaft itself is driven by the need to carry high torque; the more torque it can carry the more power can be transferred from the turbine to the pump(s). Shaft power is the product of shaft speed and shaft torque. This high torque requirement drives the designer to maximizing the polar moment of inertia of the shaft. It is not uncommon for shafts to be hollow, as this maximizes this polar moment of inertia for a given weight of material. Shaft also need to transfer torque to the components of the rotor stackup. This can be accomplished via keyways, which carry less torque but are easier to manufacture, splines which generally carry higher torque but more difficult to manufacture, or shear pins, which are common for components attached to the circular face of the shaft (i.e. turbine wheels). Seal design Turbopumps need to keep fuel and oxidizer apart from each other; otherwise there is high risk of ignition in the turbopump that will cascade into a total failure of the rocket engine. Secondarily they also need to keep propellants out of the turbine cavity; to avoid wastage and also to avoid changing the conditions of the gas flow through the turbine. They especially want to keep oxidizer out of a turbine running fuel-rich, and fuel out of a turbine running ox-rich. This is because leakage in this case would push the oxidizer/fuel (OF) ratio of the working gas closer to stoichiometric, increasing its flame temperature which may be too much for the turbine materials to handle. For these reasons turbopumps always have dynamic seals around their shafts, where one part of the seal is attached to the rotor and corotating with it, while the other part is statically attached to the housing. The dynamic seals in turbopumps have quite specialized requirements compared to seals in most systems. They must support very high shaft speeds on shaft of significant diameter, meaning rubbing velocities are very high. They usually need to be cryogenically compatible as well, and oxygen-compatible on the seals exposed to the oxidizer side. This eliminates the possibility of elastomer based seals, which will embrittle (and cannot hold up at these speeds anyways). Spring loaded and other compression type seals are also not practical at these speeds. In practice, turbopumps primarily use three seals: labyrinth seals, face seals, and carbon ring seals. Labyrinth seals are a non-contact type where the fluid is routed through a circuitous path that minimizes the seal's discharge coefficient, and thus minimizes leakage through it. Labyrinth seals leak the most of the three types, and so are seldom used in isolation. Face seals consist of two metal sealing faces that are lapped to a very smooth finish and are pressed together during assembly. These face seals are typically of the non-contact "lift-off" variety, where they develop a thin microfilm of leakage fluid between them during operation that minimizes friction between the static and rotating face. Carbon ring seals are contact seals that consist of multiple carbon static segments around the shaft. They are pressed tightly around the shaft and during operation will intentionally "wear in" to provide a precision sealing surface with minimal leakage. In practice, all three of these seal types will leak to some extent. A large part of seal design is providing safe flow paths for this leakage. Most imperative is that the interpropellant seal (IPS), which the vast majority of engines have at least one of, does not leak fuel and oxidizer together. This is often accomplished by having a central cavity that is continuously purged with inert gas (e.g. helium, nitrogen) at a higher pressure than the propellants on either side, so that the IPS will leak that inert buffer gas outwards from the cavity instead of propellants inwards to the cavity. The only engines that are able to forgo an IPS entirely are full flow staged combustion cycles, because they have one entirely fuel rich turbopump and one entirely ox rich turbopump that do not interact with each other. Rotordynamic design A major driver of turbopump mechanical design and shaft speed selection (which by extension affects hydrodynamic and aerodynamic design), is its rotordynamics. At high speed the rotor can start precessing in its bearings, which can induce large stresses and cause failure. This phenomenon is referred to as whirling. Of paramount importance is to avoid running the shaft near to critical speeds, defined as speeds that will excite natural frequencies in the design. Beyond tuning the speed, one can mitigate whirling by avoiding cantilevering large masses (e.g. impellers or turbine wheels) being cantilevered away from bearings. From a rotordynamic perspective, the ideal turbopump has bearings at the extreme ends of the shaft and all the rotor components between them (the RD-180 turbopump gets somewhat close to doing this). In most engines this kind of design is usually impractical; for example it creates rather complex flow paths. Instead the engineer's task is to minimize the cantilevered length, and increase the stiffness of the load path between the cantilevered components and the bearings - hence the desire to preload rotor components onto the shaft. This is done in a strategic way to reduce the moment of inertia of the rotor below some set value. == Impellers ==

A few criteria are used when sizing and designing impellers. The first is specific speed - this is a dimensionless parameter characterizing the impeller discharge, for which certain ranges of values are empirically known to indicate different impeller designs would be optimal. N_s = \frac{N_{RPM} \cdot \sqrt{Q_{gpm}}}{H_{ft}^{0.75}} \,\,\,\,\,\,\,\, n_q = \frac{N_{RPM} \cdot \sqrt{Q_{m^3/s}}}{H_m^{0.75}} \,\,\,\,\,\,\,\, \omega_s = \frac{\omega_{rad/s} \cdot \sqrt{Q_{m^3/s}}}{(g \cdot H_m)^{0.75}} \omega_s = \frac{N_s}{2733.00} = \frac{n_q}{52.9186} • H = head • Q = volumetric flowrate • N = shaft speed N_s is the imperial version, common in US literature. n_q is common in European literature. \omega_s is the dimensionless version, but is not yet commonly seen in pump literature. The second parameter is similar: the suction specific speed. This characterizes the impeller's inlet (suction) conditions, and is used to quantify the required inducer and tank pressures upstream of the impeller. N_{ss} = \frac{N \cdot Q^{0.5}}{(NPSH_R)^{0.75}} NPSH is net positive suction head; NPSHR is the amount of head required to be generated in the fluid before it reaches the impeller inlet in order to not excessively cavitate in the impeller. "Excessive" is often defined as the level of cavitation that would degrade the pump's discharge head by 3% – hence it is common to see NPSHR defined as NPSH3%. Another key parameter is the impeller's head coefficient \psi. This characterizes how effective a given tip speed u_{out} is at generating head. Head coefficient is typically selected (for a given specific speed) from empirical curves generated by previous industry experience. \psi=\frac{2gH}{u_{out}^2} in some sources; Centrifugal (radial) impellers Centrifugal impellers are optimal on a range of 500 N_s < 2500 (numbers are approximate and vary by source). Compared to centrifugal impellers, axial impellers trade lower head generation for higher volumetric flowrates of propellants. For this reason they can be attractive for pumping liquid hydrogen, because of its significantly lower density than essentially all other propellants which use centrifugal pump designs. The only pure axial pump to have ever flown operationally was on the J-2 engine of the Saturn V upper stages. This is of course not counting inducers, which are ubiquitous, and in some uncommon cases is the sole pumping element (e.g. the RS-25 low pressure fuel turbopump). == Inducers ==

Virtually all turbopumps since about 1955 The inducer is an axial, spiral design that raises the fluid pressure enough to prevent cavitation when it reaches the entrance to the impeller. The head pressure that the fluid rises over the length of the inducer is termed the NPSHA (NPSH available). This must be above the NPSHR of the impeller, quantified with suction margin: NPSH_{margin} = \frac{NPSH_{A}}{NPSH_{R}} Turbopumps also require a certain NPSH before it even reaches the inducer, again termed the NPSHR (so the inducer and impeller both have their own individual NPSHR). This is achieved by pressurizing the propellant tanks to some extent; a few bar is typical to provide enough suction margin for the inducer itself (separate to the suction margin for the impeller). Inducers for cryogenic propellants usually cannot be designed to have zero NPSHR because a rocket usually fills cryogenic propellants at their saturated state, meaning NPSHA in the tank is zero. This gives no margin and thus cavitation at the inducer blades becomes likely. This can be improved with subcooled / densified propellants (e.g. Falcon 9), but even here vapor margin is usually low compared to what an inducer requires. Regardless, some tank pressure is often desirable for structural stability of the rocket itself, and so increases the NPSHA, reducing the NPSHR of the inducer (and so probably its axial length) as a side benefit. Suction performance is usually the main limiting factor for shaft speed in turbopumps, even more than bearing limits / turbine burst / seal limits / etc. For this reason an initial sizing of a turbopump often begins with selecting desired inducer performance parameters (flow coefficient, hub-tip ratio, suction margin, flow margin) and then finding shaft speed from this as a derived quantity. == Volutes and diffusers ==

Liquid leaving the blades of an impeller has very high velocity but relatively low static pressure. This is problematic because pipes can be very lossy at high mean velocities; head loss scales with v^2 (see Darcy-Weisbach equation). To prevent large losses as propellants travels from the pump to the main injector (especially through a regenerative cooling jacket), it is highly desirable to significantly slow down the flow. This trades the kinetic energy of the flow into potential energy, realized as a much higher static pressure. The goal of the pump housing is to perform this tradeoff as efficiently as possible with minimal energy loss, and then discharge the fluid smoothly to the rest of the rocket engine. The exit fluid passages in a pump housing are usually shaped into a volute design, sometimes with a vaned diffuser between the impeller and the volute. The goal with either is to provide a smooth flow path for fluid to leave the impeller without large pressure variations along its circumference, which can cause large radial loads through the bearings. The goal of the volute itself is not to recover static pressure; this comes later on via a conical diffuser at its discharge. The cross section of a volute smoothly ramps progressively larger along the azimuth of the pump, accepting more and more fluid from the impeller exit along its length. It terminates at a discharge that is piped to the rest of the engine. The separation between the discharge and the rest of the volute is known as the tongue or cutwater. Volutes should be designed around conservation of mass, and conservation of angular momentum so as not to cause non-uniform accelerations in the flow. A common design strategy is to keep a constant mean velocity v_{mean} through the volute passage, which leads to a linear ramp in cross section area A along the circumferential angle \theta. The mean velocity is determined by an inverse relationship from the impeller discharge v_{tang, out} to conserve angular momentum as the radial distance increases. A(\theta)=\frac{Q}{v_{mean}}\frac{\theta}{2\pi} \,\,\,\,\,\,\,\,\,\,\,\,\,\, v_{mean}r_{mean}=v_{tang,out}r_{out} This becomes an iterative process, as the mean properties are found via integration across the cross section. == Turbines ==

Turbopumps, by definition, are driven by gas turbines. Because of the high power requirements, these need very high gas flow, and so are always of the axial type (in contrast with radial inflow turbines common in turbochargers). There is a fundamental tradeoff between the turbine's pitchline radius r , the shaft speed \omega , and the supplied gas flow \dot{m} . Improving one factor necessitates making the other(s) worse to maintain sufficient power output P_{stage} into the shaft and the pump: P_{stage} = \eta\cdot\dot{m}(u_{in}v_{tang,in}-u_{out}v_{tang,out}) = \eta\cdot2\dot{m}u^2 = \eta\cdot2\dot{m}\omega^2r^2 This is a rearranged form of Euler's turbine equation, assuming the ideal (though often impractical) case of zero post-swirl in the working gas leaving the turbine. Turbines are typically either of an impulse design (common in gas generator and other open cycles) or of a reaction design (common in staged combustion and other closed cycles). They can consist of one or more stages, where each stage has both a stator, which can be bladed or nozzles, and a wheel (sometimes referred to as a rotor in older papers and aero focused papers). Impulse turbines for open cycles Open cycles aim to increase efficiency by minimizing mass flow through the turbine, making up for it by maximizing pressure drop. This is because the mass flow is dumped overboard, a performance hit. Comparatively, maximizing pressure drop is easy to do because it dumps to ambient pressure, which will be significantly lower than the gas generator (GG) chamber pressure. This is true even if GG pressure were the same value as the main chamber pressure, which the pumps have to work hard enough to discharge to anyways. These considerations drive the designer towards impulse designs on the turbine, with gas flow expanded via converging-diverging blades or nozzles to supersonic velocities that then impinge on the turbine wheel. Reaction turbines for closed cycles Closed cycles aim to increase efficiency by minimizing pressure drop across the turbine, making up for it by maximizing mass flow. This is because the downstream pressure must be higher than the main chamber pressure. It is often significantly higher because of injector and regen jacket pressure losses. Consequently, the only method for increasing pressure drop is to increase chamber pressure in the preburner much higher than the main chamber. This puts significantly more load on the pumps which must have a high pressure discharge for the preburner. Comparatively, high mass flow is easy to accomplish because none is being dumped overboard - so it is common to route the entire mass flow of one propellant through the preburner and turbine. Full flow staged combustion cycles take this a step further by routing the entire mass flow (hence 'full flow') of both propellants through preburners and turbines, taking advantage of essentially 100% of possible mass flow through the engine to generate shaft power for the turbopumps. These considerations drive the designer towards reaction type designs on the turbine where gas flow is subsonically expelled from, and reacting against, the wheel blades. == Turbopumps as a full system ==

Turbopumps have a reputation for being difficult to design for optimal performance. Low efficiency may be acceptable in some applications, but in rocketry this is a severe problem. Turbopumps in rockets are important and problematic enough that launch vehicles using one have been caustically described as a "turbopump with a rocket attached". Common problems include: • Excessive secondary flow from the high-pressure impeller discharge back to the low-pressure inlet along the gap between the casing of the pump and the rotor. This problem is greater in unshrouded impellers than shrouded designs. • Excessive recirculation of the fluid at the inlet in the vicinity of the inducer, which can hinder suction performance • Excessive vortexing of the fluid as it leaves the volute casing of the pump • Damaging cavitation to impeller blade surfaces in low-pressure zones • Leakage through seals; especially interpropellant seals • Axial shaft thrust imbalances, which can cause excessive wear and failure of seals (for example non contact face seals can come into contact with enough thrust load) Turbopumps typically require significant testing and refinement to achieve good results. It should be ensured that they can survive a full service life, with significant margin. Some engines also require multiple restarts to be demonstrated; potentially tens to hundreds of restarts in reusable rocket engines. Turbopumps can be susceptible to significant high cycle fatigue (HCF) because they spin so fast - where a resonating turbine wheel is accumulating one or more cycles for every revolution. Even static components can still be susceptible to low cycle fatigue (LCF) over many restarts. ==History==

Invention and early development High-pressure pumps for large missiles had been discussed by rocket pioneers such as Hermann Oberth as early as the 1920's. The turbopump does not have a clear undisputed inventor. Rather they were developed independently in the United States and Germany in the 1930s, where each team had little to no knowledge of the other. The earliest prototype of a turbopump of any kind appears from the historical record to be by the American Robert Goddard and his team, circa 1934. This would evolve by around 1940 into the V-2 rocket design, which used hydrogen peroxide decomposed through a Walter steam generator to power the turbopump The first successful V-2 launch was shortly after on October 3, 1942. Postwar acceleration Turbopump development in the Soviet Union began in earnest in 1943, and achieved a successful first flight by 1947; accelerated towards the end by knowledge gained from the Me-163 and V-2 programs via Operation Osoaviakhim. Similarly, turbopump development in the US accelerated from the experimental designs of Goddard into a much larger industrial effort via expertise gained through Operation Paperclip. Around this time serious effort was put into the development of liquid hydrogen as a propellant - spearheaded by a program for turbopump development at Aerojet led by George Bosco. During the second half of 1947, Bosco and his group learned about the pump work of others and made preliminary design studies. Aerojet representatives visited Ohio State University where Florant was working on hydrogen pumps, and consulted Dietrich Singelmann, a German pump expert at Wright Field. Bosco subsequently used Singelmann's data in designing Aerojet's first hydrogen pump. By mid-1948, Aerojet had selected centrifugal pumps for both liquid hydrogen and liquid oxygen. They obtained some German radial-vane pumps from the Navy and tested them during the second half of the year. It technically featured four separate turbopumps; with both the ox side and fuel side featuring a smaller boost pump before their main pumps, to increase suction performance in a design demanding very high head / shaft speeds. While not technically a turbopump (in that it lacks a turbine), the Electron rocket's Rutherford became the first engine to use an electrically-driven pump in flight in 2018. == Turbopump examples ==

Where two values are given, fuel side listed first and oxidizer side listed second. Notes ==See also==