Reflux Pump Cavitation After a Rate Increase
A reflux pump draws liquid from a column accumulator drum at 20 psia. The liquid is at its bubble point, so its vapor pressure is also 20 psia. At the design point of 300 gpm, the static lift from the normal drum liquid level to the pump centerline is 12 ft of liquid, and the suction line friction loss is 3.0 ft. The pump curve gives NPSHr = 7.0 ft at 300 gpm. The seal has failed three times in 18 months, always with a cracked stationary face and a light hydrocarbon weep. Operations wants to run 420 gpm to improve overhead separation, and the drum level is on a controller with a 10-minute reset time. The suction block valve is currently 80% open and accounts for about 1.5 ft of the 3.0 ft friction loss. An engineer proposes opening that valve fully and raising the drum pressure to 30 psia. Liquid specific gravity is 0.62.
1. Compute the available NPSH at the design point and assess the current margin.
Read the worked answer
NPSHa = (P_drum - P_vapor) x 2.31 / SG + static head - suction friction.
Because the liquid is at its bubble point, P_drum - P_vapor = 0, and the pressure term vanishes entirely regardless of specific gravity. That leaves NPSHa = 12 - 3.0 = 9.0 ft.
Against NPSHr = 7.0 ft, the margin is 2.0 ft, or about 1.29 times NPSHr. The usual guidance is to hold 3 to 5 ft of margin, or at least 1.1 times NPSHr with a wide curve. Two feet is inside the range where a level swing, a slightly fouled strainer, or a high-vapor-pressure batch will start cavitation.
That explains the seal history. The 10-minute reset time on the level controller means the drum level can be several inches below normal for minutes at a time, and each excursion shaves the margin further. The seals are the symptom; the suction system is the cause.
2. Evaluate the 420 gpm proposal with numbers.
Read the worked answer
Suction friction scales with the square of flow. At 420 gpm, friction = 3.0 x (420/300)^2 = 3.0 x 1.96 = 5.9 ft. Then NPSHa = 12 - 5.9 = 6.1 ft.
On the required side, NPSHr rises steeply with flow, roughly as the square in the normal operating region: 7.0 x 1.96 = 13.7 ft. Even if you assume a very forgiving 1.3 exponent, you get 7.0 x 1.59 = 11.1 ft.
The comparison is 6.1 ft available against 11 to 14 ft required. That is not a marginal case, it is deep cavitation from the moment the flow increases. Physically, the impeller eye pressure drops below vapor pressure, vapor bubbles form and collapse on the discharge side of the vane, and the resulting erosion and thrust reversal destroy the seal and eventually the impeller.
The proposal must be rejected. If the column genuinely needs more reflux, the fix is a larger suction line, a lower pump elevation, or subcooling the liquid in the condenser, not more flow through the existing pump.
3. Critique the two proposed fixes: opening the suction valve fully and raising drum pressure to 30 psia.
Read the worked answer
Opening the valve fully recovers 1.5 ft, bringing design NPSHa from 9.0 to 10.5 ft, and that gain also applies at higher flows. It is cheap and worth doing, but it does not come close to covering the 4.6 ft shortfall at 420 gpm.
Raising the drum pressure to 30 psia sounds decisive: at 10 psi and SG 0.62 the pressure term would add 10 x 2.31 / 0.62 = 37 ft. That calculation is wrong for this service. The drum liquid comes from the condenser and re-equilibrates at the new pressure, so its vapor pressure rises to 30 psia along with the drum pressure. P_drum - P_vapor returns to zero and NPSH is unchanged. You only gain if the liquid is genuinely subcooled, for example by adding a condenser subcooling pass or a cooler on the reflux line, and then only by the amount of subcooling.
There is a second cost: raising the column pressure changes relative volatility and condenser duty, so it degrades the separation you were trying to improve. The defensible package is a fully open suction valve, a shorter or larger suction line, a raise in normal drum level, and a faster level controller.