Pump displacement of 4500 gph against a 350 gph load
The quantity that decides this case is the ratio of pump displacement to actual demand. The direct-drive gear pumps deliver about 4500 gph at 1200 rpm (about 75 gpm, or roughly 0.0625 gal/rev, derived from the two figures). The load is five 1000 kVA generators at 70 gph each, so normal demand is 5 x 70 = 350 gph. The 1400 gph figure is that demand multiplied by a safety factor of 4 (70 x 4 x 5) to refill the day tanks quickly.
The consequence is that the pump is sized about 3.2 times above the safety-factor flow and about 12.9 times above normal flow. The internal relief valve has to recirculate the difference:
| Operating point | Downstream flow | Flow through internal relief | Share of displacement |
|---|---|---|---|
| Normal demand (5 x 70 gph) | 350 gph | 4150 gph | about 92 % |
| Day-tank refill (safety factor 4) | 1400 gph | 3100 gph | about 69 % |
At 350 gph a 2 in swing check valve sits barely cracked off its seat, because the flow is far below what it needs to open fully. A swing disc held at a small opening in a pulsing gear-pump stream flutters against the seat. The pump also sees a discharge pressure that hovers near the relief setpoint, so the relief valve opens and closes as the disc moves. Those two pressure-sensitive elements in series, disc and relief poppet, are the interaction to isolate.
The installation details that bound the problem: suction is flooded (pump below liquid level, strainers found full), the strainer-to-pump suction run is about 18 in, and the discharge train is gauge, check valve, ball valve in that order. Pump ports are 3 in on 2 in pipe. The pump ports were not noticed until installation, and the supplier accepted 2 in pipe with a floor of 2 in. The old 1 in gear pumps are gone, and so is an elevated 250 gal tank on the downstream side that was filled by manually opening a ball valve.
Check valve spacing carries no pressure
The 8 in between pump discharge and check valve does not set back pressure, and no minimum distance applies. Pressure at the pump discharge port equals the downstream head, plus friction, plus the check valve cracking pressure. Diesel is effectively incompressible, so pipe length between the port and the valve changes only the trapped liquid volume, not the pressure. Moving the valve further downstream adds no back pressure to the pump.
The test that showed the noise disappearing with the swing arm removed still points at the valve, but through its dynamic behavior. A disc that opens only slightly at low flow, and that closes when the relief valve dumps pressure, produces a pressure-and-flow interaction at the valve and pump. Without the disc, flow passes freely and that interaction cannot occur. This result does not show that the distance is wrong.
Keep a check valve on each pump. Without one, an idle pump becomes a reverse-flow path from the common header and the running pump's discharge.
Candidate mechanisms and what each predicts
| Mechanism | Physics | Prediction at start | Discriminating check |
|---|---|---|---|
| Insufficient discharge back pressure (free-wheeling gears) | Pump delivers into an empty, unloaded line after the elevated tank was removed; gear teeth lightly loaded chatter against each other | Grinding for roughly the first seconds, fading as flow and pressure rise | Start against a nearly closed manual valve near the discharge, then open it gradually over 5 to 10 s; noise should drop |
| Relief valve and swing check interaction (hammer) | About 92 % of flow recirculates at normal demand; check disc cracked off seat while relief opens and closes | Noise tied to relief setpoint and flow; changes when the setpoint is moved | Watch the discharge gauge during start; raise or lower relief setting within piping limits and compare |
| Suction starvation | Insufficient liquid entering the pump at start | Persistent noise, often with vacuum-gauge reading and no dependence on the check valve | Verify suction valves open, strainer clean, flooded head; less likely because strainers are full and the pump is below the liquid level |
| Old valves passing, new valves sealing | Previous check valves leaked back, so the line stayed partly primed and pressurized; new ones seal and expose the empty-line start | Noise appears only with the new sealing valves | Compare with the old valve behavior; check whether the downstream line drains between starts |
Grinding that clears as flow builds is a hydraulic transient. Grinding that persists at steady flow with rising casing temperature is mechanical or a lubrication fault, and the pump then goes back to the supplier. The pump representative already found the pump itself fine and shaft alignment was checked.
Comparing the four corrective approaches
| Approach | What it changes | Effort and risk | Fit to this case |
|---|---|---|---|
| Move the check valve further from the pump | Trapped volume only | Piping rework with no pressure benefit | Poor; no physical basis |
| Adjust internal relief setting (higher or lower) | Moves the pressure at which bypass starts and detunes the relief and check interaction | Low; must stay within the piping design pressure; pump suitable up to 250 psi discharge | Good first trial |
| Add a back-pressure valve set at about 35 to 40 psi | Opens fully in normal operation but creates instant back pressure at start | Moderate; new component in the discharge train | Good if the throttle test confirms free-wheeling |
| Replace with a smaller pump (model AS4195, 2-1/2 in ports, named as the better match) | Reduces oversizing and recirculation fraction | Highest cost; matches the 2 in pipe better | Best permanent fix if the relief adjustment only masks the noise |
| Accept about 20 s of startup noise | Nothing | Depends on start frequency per day | Only if starts are rare and noise is transient |
Recommendation: run the throttle test and the relief adjustment first, since both are reversible and cost nothing. If the noise is tied to back pressure, add the 35 to 40 psi back-pressure valve. If neither cures it, or the pump runs mostly in bypass, convert to the smaller pump. A spring-assisted, non-slam check valve is a further option for the flutter mechanism, chosen on engineering judgment and the pump supplier's confirmation.
Startup pressure trace and throttle test
- Fit or read the existing discharge gauge, which sits ahead of the check valve, and watch it from the moment of start. Record peak pressure, time to settle, and the moment the noise stops.
- Record the pump speed at the motor nameplate. The pump is direct drive, so the motor speed is the pump speed; the 4500 gph figure assumes 1200 rpm.
- Close a manual valve close to the pump discharge until it is almost shut, start the pump, and open the valve gradually over 5 to 10 s. The internal relief protects the pump during this test. Keep the test short, because the whole displacement recirculates and heats the fuel.
- Repeat with the swing arm installed and the downstream ball valve open, and compare noise and gauge trace with the throttled start.
- Log start frequency per day. It decides whether accepting a brief noise is workable.
If noise drops when discharge pressure rises quickly, the mechanism is insufficient back pressure. If the gauge sits at or near the relief setpoint and the noise follows it, the relief and check valve interaction is dominant.
Relief valve adjustment procedure
- Read the current relief setting from the pump documentation and mark the adjuster position before touching it.
- Confirm the piping, strainer, and fittings design pressure. The pump is suitable to 250 psi discharge, but the system limit is what caps the setting.
- Change the setting in small steps, one direction at a time, and run a start at each step with the gauge trace recorded.
- Stop when the noise disappears at both normal and 1400 gph demand. If moving the setpoint either way changes the noise character, the relief valve is part of the interaction.
- If no setting cures it inside the piping limit, restore the original setting and move to the back-pressure valve or the smaller pump.
Quantities to read and limits
| Quantity | Value or limit | Where to read |
|---|---|---|
| Normal demand | 350 gph (5 x 70 gph) | Generator fuel consumption at full load |
| Refill design flow | 1400 gph (70 x 4 x 5) | Day-tank refill requirement |
| Pump output at speed | 4500 gph at 1200 rpm | Pump curve; confirm speed on motor nameplate |
| Maximum discharge pressure | 250 psi | Pump rating; also verify system piping rating |
| Back-pressure valve setting | about 35 to 40 psi | Valve adjustment, if fitted |
| Relief setpoint | Factory setting, read before change | Pump relief adjuster and documentation |
| Startup discharge pressure | Trace over first seconds | Discharge gauge ahead of check valve |
Acceptance after the change
Accept the fix when three starts in a row show no grinding at the pump casing, the gauge settles without oscillating near the relief setpoint, and all five day tanks refill at the 1400 gph design flow. Run each pump alone through its own check valve to confirm no reverse flow through the idle pump. Leave the swing arm installed; the removed-disc configuration was a diagnostic only. If the throttled start remains quiet but the normal start is loud, the fix belongs on the discharge back pressure and not on the pump internals.
FAQ
Why does the gear pump grind only when it first starts?
At start the discharge line offers almost no back pressure, so lightly loaded gears can chatter until flow and pressure build. The noise stops as pressure rises, which is why a throttled-discharge start over 5 to 10 s is a useful test.
Why does removing the swing check arm stop the noise?
Without the disc there is no flutter at the valve and no interaction with the internal relief valve, because flow passes freely. This points to valve dynamics at low flow, not to a required distance from the pump.
Why does an oversized gear pump make check valve problems worse?
At 4500 gph against a 350 gph demand, about 92 % of the flow recirculates through the internal relief, and the 2 in check valve sits barely cracked. That combination of relief cycling and a low-lift disc produces flutter and pressure pulses.
Why does moving the check valve further from the pump not add back pressure?
Back pressure comes from downstream head, friction, and the valve cracking pressure. Pipe length between the port and the valve only adds trapped volume, and no minimum 8 in or similar distance requirement is documented.
When should I stop testing and call the pump manufacturer?
Stop if the noise persists at steady flow, if casing temperature climbs, or if the noise is unchanged by the relief adjustment and the throttled start. Contact the pump manufacturer's official technical support with the pump model, speed, relief setting, and the recorded startup gauge trace before running the pump further.