Inlet Conditions, Cavitation and Aeration in Hydrostatic Pump Circuits

2. The inlet governs what the outlet can do

2.1 What a piston actually requires

During the suction stroke of an axial piston pump, the piston withdraws and the volume behind it increases. Liquid fills that volume only if something pushes it in. In almost all mobile installations that something is atmospheric pressure acting on the reservoir surface, assisted or opposed by static head, and reduced by every loss between the reservoir and the port. The pump itself does not draw; it creates a space, and the circuit either fills it in the time available or does not.

The time available is short and it shortens with speed. At 2,200 rev/min each bore's suction stroke lasts half a revolution, about 14 ms, however many pistons the unit carries, and the time the bore spends fully open to the inlet port is shorter still. If the pressure difference driving fill is inadequate, the bore does not fill completely with liquid, and what occupies the remainder is vapor, released air, or nothing at all until the fluid arrives late and at velocity.

2.2 The suction balance

The margin available is conventionally expressed as net positive suction head, the amount by which the absolute pressure at the inlet exceeds the vapor pressure of the fluid, stated as a head of that fluid:

where p_atm is the absolute pressure over the reservoir, p_v the vapor pressure of the fluid at operating temperature, ρ the density, h_static the elevation of the pump above the fluid surface, and h_loss the total hydraulic loss of the suction line. The pump manufacturer states a required value, NPSH_r, below which the unit is not warranted to fill; the steady-state test methods by which such figures are established and presented are those of ISO 4409². The design question is whether the first exceeds the second under the worst condition the machine will meet, which is almost never the condition at which it was commissioned.

Two terms in that expression are commonly treated as fixed and are not. Vapor pressure rises with temperature, and it rises steeply once the fluid carries dissolved water. Atmospheric pressure falls with altitude: a machine commissioned at sea level and worked at 2,000 m has lost roughly a fifth of the only term that was pushing fluid into the pump, without anything on the machine having changed.

3. Suction-line loss, and why cold oil is the governing case

The loss term is given by the Darcy–Weisbach relation, which applies to the suction line as it does to any other:

with f the friction factor, L the developed length, D the internal diameter and v the mean velocity. The apparent implication is that velocity dominates, since it enters squared. In the suction line of a mobile machine on cold oil, it usually does not, and the reason is the friction factor.

Suction velocities are low by design — the general practice published in installation literature is to hold them to a few meters per second at most — and hydraulic oil is viscous. Grade is assigned by the classification of ISO 3448⁶ and the kinematic viscosity itself is determined by ISO 3104⁵, which is what makes the viscosity term in the relation below a measured quantity rather than an assumed one. The Reynolds number is therefore frequently below the transition, and in laminar flow the friction factor is not an empirical constant but f = 64 / Re. Substituting Re = vD/ν collapses the velocity dependence from square to linear and introduces kinematic viscosity as a direct multiplier:

This is the Hagen–Poiseuille result, and it changes where the engineering attention belongs. Velocity is linear, so halving it halves the loss. Viscosity is also linear, and viscosity is the term that moves by more than an order of magnitude between a machine at working temperature and the same machine started at dawn in winter.

The figure states the case. For a 50 mm line two meters long carrying oil at 1 m/s, the loss at a warm-oil viscosity of 46 mm²/s is a small fraction of the available margin. At the viscosity of the same oil well below freezing it is not a small fraction, and the margin that existed in the specification has been spent before the pump has done any work. This is the mechanism behind the familiar observation that inlet damage is a cold-start phenomenon, and it is also why a strainer that is merely dirty rather than blocked can be sufficient to cause it: the added loss is proportional, and it is added to a term that is already large.

4. Vapor cavitation

If absolute pressure at any point in the inlet falls to the vapor pressure of the fluid, the liquid boils. The vapor cavities so formed are carried into the bore and then into the delivery region, where pressure rises by two orders of magnitude within a fraction of a revolution. The cavities do not gently recondense; they collapse, and they collapse asymmetrically when near a surface, driving a microjet at the boundary.

The damage that results is characteristic and is not readily confused with anything else once seen. It is highly localized, it is directional, and it has the appearance of material having been removed rather than worn: a pitted, matte, almost sintered texture on valve plate lands, port plate kidneys, and the leading edges of the bore entries. Its location is diagnostic, because it marks where pressure recovered rather than where load was highest.

Two features of the mechanism deserve emphasis because they govern what an inspection can conclude. First, the energy that does the damage is supplied by the delivery side, not the inlet, so the severity of the marking says nothing about how far below vapor pressure the inlet fell. Second, the process is self-accelerating in a way that is easy to misread: eroded lands increase leakage, leakage lowers volumetric efficiency, the operator or the control compensates by demanding more speed or displacement, and higher speed shortens the filling window that was already inadequate.

5. Air release, and why it is not the same failure

5.1 Solubility

Mineral hydraulic oil holds air in true solution, and the quantity it holds is proportional to absolute pressure. The tendency of a given oil to surrender that air once released is itself a specified property, measured as an air release value by ISO 9120⁴ or by ASTM D3427⁹. For the mineral oils used in mobile equipment the Bunsen coefficient is conventionally taken as approximately 0.09, which is to say that at one bar absolute the oil holds of the order of nine per cent air by volume, dissolved and invisible:

Dissolved air is not a problem. It does not compress separately, it does not scatter light, and it does not affect stiffness. The problem is what happens when the pressure that was holding it in solution is removed.

5.2 Release at the inlet

Air comes out of solution when absolute pressure falls, and it comes out at any pressure below saturation — which is to say well above the vapor pressure at which the fluid would boil. This is the crucial asymmetry with vapor cavitation: gaseous cavitation begins earlier, in conditions that are not otherwise remarkable, and it can be occurring continuously in a circuit whose inlet never approaches vapor pressure at all.

Taking the coefficient above, oil saturated at atmospheric pressure and then carried to an inlet at 0.4 bar absolute can retain only about four tenths of what it held. The balance separates as free bubbles, distributed through the fluid rather than formed at a surface. Those bubbles are then carried into the bore and compressed, which heats them intensely — the compression is rapid enough to be closer to adiabatic than isothermal — and hot compressed air in contact with oil at a surface is the condition under which the oil oxidizes locally. Darkened, varnished surfaces in an otherwise undamaged rotating group are the signature.

5.3 Re-absorption is slow

Release is fast and re-absorption is slow; the rates differ by orders of magnitude, because release requires only nucleation while absorption is limited by diffusion across the bubble surface. A circuit that liberates air at the inlet on every revolution does not recover it on the return, and the entrained fraction accumulates until reservoir residence time balances the release rate. This is the mechanistic argument for reservoir volume and for baffling — the subject of SAE J1234¹² — and it is the reason a reservoir that is merely small can present as a pump fault. Where air is being carried rather than dissolved, the foaming tendency of the fluid measured by ASTM D892⁸ becomes the relevant property.

6. What entrained air does to the fluid

Free air in the fluid, unlike dissolved air, changes the property the whole system depends on. The effective bulk modulus of a liquid carrying a volume fraction x of gas is obtained by adding compliances in series, the gas following a polytropic law with exponent n:

The relation carries an immediate and unintuitive consequence. The second term is divided by absolute pressure, so the damage done by a given air fraction is greatest exactly where pressure is lowest — at the inlet, in the reservoir, and during the transient that follows a sudden demand, which is when stiffness matters most.

For the oil bulk modulus of 1.6 GPa typical of published mineral-oil data and an adiabatic exponent of 1.4, one per cent entrained air reduces effective stiffness to a small fraction of the fluid's own value at ten bar, to roughly half at a hundred bar, and to about seven tenths at two hundred and fifty. Those are three different fluids, each carrying one per cent at its own pressure. A single fluid stiffens faster than that, because its air is itself compressed as pressure rises: bubbles that are one per cent of the volume at ten bar occupy about a fifth of a per cent or less at a hundred, and the fluid there has most of its stiffness back. The relation explains a behavior familiar from the field and otherwise puzzling: a machine that is soft, slow to respond and audibly harsh at low pressure, and that firms up and quietens as load comes on. The fluid is not the same fluid at the two conditions.

7. Consequences for pressure control

A pressure-compensated or load-sensing pump control is a closed loop whose plant includes the compliance of the fluid it is acting on. The compensator senses a pressure, moves a spool, and the spool strokes the swashplate to change displacement until the sensed pressure returns to the set margin. The rate at which the sensed pressure responds to a change in displacement depends on the stiffness of the fluid volume between them.

When that stiffness falls by the factor the previous section derives, the loop is no longer the loop the control was designed around. The plant has become slower and softer while the controller's gain is unchanged, and the classical result is a loop that hunts: pressure overshoots, the compensator over-corrects, and the swashplate oscillates at a frequency set by the compliance it is working into. The audible and tactile symptoms — a cycling note at the pump, a shudder in an implement holding position, a margin pressure that will not settle — are commonly attributed to the compensator itself, and the compensator is commonly replaced. Distinguishing a control instability from the pump's own delivery ripple is a measurement problem, and the method for characterizing the latter is given in ISO 10767-1³.

This paper takes no position on any specific control architecture or gain, which are manufacturer-specific and outside what can responsibly be generalized. The narrow claim is the one the relation supports: fluid stiffness is a term in the plant, entrained air is a large and pressure-dependent influence on fluid stiffness, and a control diagnosis that has not established the condition of the fluid has not yet excluded the most likely cause.

8. Implications for inspection and for replacement

Three consequences follow for practice, and none of them is a specification.

The first is that inlet condition is measurable and is rarely measured. Absolute pressure at the pump inlet, taken at the operating temperature and speed at which the complaint occurs rather than at idle on warm oil, converts the argument of this paper from theory into a number. The worst case is a cold start under load, and that is the measurement worth having.

The second is that the damage signatures separate the two mechanisms, and separating them directs the remedy. Localized, directional erosion at pressure-recovery surfaces indicates vapor cavitation and points at the suction balance: line size, line length, strainer condition, fluid temperature, altitude. Distributed varnishing and darkening without erosion indicates air release and points at the return side, reservoir residence, and wherever the circuit is admitting air below the fluid level — and at the suction balance too, because release begins well above vapor pressure. The two share the suction balance but otherwise demand different work, and a unit replaced without that distinction being drawn is a unit installed into the conditions that destroyed its predecessor.

The third is the one that matters commercially and is the reason this subject belongs in a drivetrain library at all. A replacement pump inherits the circuit. Where the inlet was the cause, the specification of the new unit is irrelevant to its survival: the metallurgy, the clearances and the build standard of the replacement have no bearing on whether liquid arrives at its pistons. The distinction between a unit that failed and a circuit that fails units is the distinction that decides whether the second one lasts, and it is established before the order is placed rather than after the second failure.