Follow the fluid, not the pump curve. Two pumps in series, a cooler between them rated 10 bar, and a full-flow recycle line that returns from the big pump's discharge to a tie-in upstream of that cooler. Block the final discharge, open the recycle, and the question is whether the 10 bar exchanger ends up inside a loop that nothing bounds from above.
Where does the pressure actually stop?
Map every node before touching a relief calculation. The spec break is the only thing that matters, and it is a valve, not a pipe schedule change.
| Node | Normal pressure | Design/rating | What bounds it |
|---|---|---|---|
| Pump A suction | Source pressure | Low-pressure spec | Upstream vessel |
| Pump A discharge = cooler inlet | 5 bar | 10 bar | Pump A shutoff head, 10 bar |
| Cooler outlet = pump B suction | < 5 bar | 10 bar | Cooler ΔP at flow |
| Pump B discharge | 90 bar | High-pressure spec | Pump B shutoff head, 100 bar |
| Recycle valve outlet / tie-in | Loop pressure | 10 bar | Recycle valve trim only |
Two facts fall out of that table. Pump A's shutoff head (10 bar) equals the cooler design pressure exactly — zero margin. And a single control valve separates a 90–100 bar system from 10 bar piping. Both are findings, not assumptions.
Check: walk the P&ID and mark, in colour, every element between the 90 bar node and the 10 bar node. If the count is one valve, the spec break is unprotected by definition.
Does 100 % recycle raise the cooler pressure on its own?
No — and the reason is worth stating precisely, because it is the part people get backwards. In a liquid-full loop with no net inflow or outflow, the pump adds differential head, not absolute pressure. Pump B raises its discharge above its suction by the head it makes at the recycle flow; the recycle valve throws that same ΔP away. Circulate the loop at 100 m³/h and the pressure profile is a closed triangle: it repeats every lap. Nothing inside the loop sets the absolute level.
What sets the absolute level is the boundary. With the discharge blocked, no liquid leaves the loop, so pump A delivers zero flow and rides to shutoff — 10 bar at the cooler inlet, held there by its own discharge check valve. That check valve is a one-way floor: pump A can push the loop up to 10 bar, but it cannot pull it back down. Above 10 bar the check closes and the loop is a blocked-in, liquid-full volume with a running pump inside it and no path to a lower-pressure system.
Add the cooler pressure drop on top. The recycle return enters upstream of the exchanger, so the exchanger inlet sits at pump B suction pressure plus the cooler ΔP at recycle flow. At the full 100 m³/h that ΔP is not negligible, and it stacks directly on the 10 bar floor.
Check: with the plant down, open the recycle valve fully and dead-head pump A against the closed loop. The pressure indicator at the cooler inlet should settle at the certified shutoff head. If it reads above the nameplate value, the impeller, speed, or fluid density is not what the curve assumed.
Which external sources can push the loop above 10 bar?
Internal circulation cannot overpressure itself. Every real overpressure case comes from outside the loop, so enumerate the entry points rather than arguing about pump curves.
| Source | Path into the loop | Driving pressure | Credible? |
|---|---|---|---|
| Backflow from the discharge system | Recycle valve fails open with pump B tripped | Up to 100 bar | Yes, if the recycle takeoff is downstream of the discharge check |
| Pump A shutoff head | Pump A discharge check | 10 bar | Yes — at, not above, design |
| Thermal expansion | Pump B power dissipated into a blocked-in liquid volume | Unbounded, hydraulic | Yes, if cooling medium is lost or isolated |
| Cooling-medium in-leakage | Tube or gasket failure in the cooler | Utility supply pressure | Check the utility side rating |
| Continued upstream supply | Feed line into pump A suction | Source pressure | Only if the source can exceed 10 bar |
The decision that governs the whole study is the position of the recycle takeoff relative to pump B's discharge check valve. Takeoff upstream of the check and the 90 bar inventory is isolated when the pump stops; the recycle valve then only ever sees pump B's own discharge. Takeoff downstream and the recycle valve is the single barrier between the high-pressure system and 10 bar piping. Confirm it on the isometric, not the flow schematic.
Check: stroke the recycle valve to fail position with air removed. Note whether it drives open or closed on loss of signal and on loss of instrument air separately — they are not always the same.
Which scenario sets the relieving rate?
Size each credible case and take the largest. Three of them behave differently.
- Centrifugal pump deadhead. Read the required rate off the certified pump curve at the head corresponding to the relieving pressure, not at rated flow. With a set pressure at the 10 bar cooler design and 10 % accumulation, the relieving pressure of 11 bar already exceeds pump A's 10 bar shutoff, so pump A delivers zero at that point and contributes no required capacity. That credit only holds if you verify shutoff head at maximum speed, maximum specific gravity, and with a new (untrimmed) impeller.
- Thermal expansion. A liquid-full loop with a running pump and no cooling has no vapour space to absorb expansion; pressure rises hydraulically at the bulk modulus of the fluid. The rate is small, but the case is real whenever the cooling medium can be isolated while pump B keeps turning.
Confirm the scenario set and the accumulation limits against API 521 and the sizing method against API 520 Part I before it goes into the relief study. The rate for the backflow case is a valve-capacity calculation; the rate for the pump case is a pump-curve lookup — do not blend the two.
Check: tabulate all three rates in the same units and mass basis. If the backflow case is not the largest by a wide margin, re-examine the recycle valve Kv.
Setting and installing the relief device
Set pressure at or below the cooler design pressure of 10 bar, corrected for the actual coincident temperature — a design pressure quoted at ambient is not the MAWP at operating temperature. Locate the PSV on the low-pressure side of the spec break, physically between the recycle tie-in and the cooler, so the exchanger is inside the protected envelope for both the backflow and the deadhead paths.
Three installation details decide whether the valve works:
- Inlet line loss. Non-recoverable pressure drop from the vessel to the PSV inlet must stay within 3 % of set pressure at rated capacity, or the valve chatters and destroys its seat. On a full-flow liquid relief this drives the inlet size, not the orifice.
- Liquid trim. Use an API-certified liquid-service valve rated at 10 % overpressure. A conventional gas trim will not lift cleanly on incompressible flow.
- Discharge routing. The relieved liquid has to go somewhere at 100 m³/h scale. Route to the source tank or a closed drain sized for the rate, and check built-up backpressure against the valve type — balanced bellows or pilot-operated if backpressure exceeds the conventional valve's limit.
Fit thermal relief separately. A small dedicated thermal PSV on the blocked-in section protects against slow hydraulic expansion without the large valve simmering on every minor upset.
Check: hydro-test the inlet spool and verify the as-built inlet loss by calculation at the certified rated capacity, not at the required capacity.
Proving it end to end
- Isolate the high-pressure side and confirm the recycle valve's fail position by removing signal, then removing air.
- Dead-head pump A into the closed loop and record the cooler inlet pressure. Compare against the certified shutoff head at the fluid's maximum specific gravity.
- Run pump B on full recycle at 100 m³/h and log cooler inlet and outlet pressures. The differential is the cooler ΔP that stacks on the loop floor — confirm the sum stays below set pressure.
- Isolate the cooling medium with pump B recycling and watch the loop pressure trend. A rising trend confirms the thermal case is live and the thermal PSV is the only thing bounding it.
- Bench-test the PSV to the stamped set pressure and record seat tightness, then reinstall with car-sealed open isolation on both inlet and outlet.
- With the process at normal conditions, close the final discharge block valve and let the recycle take full flow. Record the cooler inlet pressure at one-minute intervals for ten minutes. It must stabilise below set pressure with no upward drift — that stable trace, not the calculation, is the proof the exchanger is protected.
FAQ
Can 100 % internal recirculation overpressure the pump suction by itself?
No. In a liquid-full closed loop the pump adds differential head only; the recycle valve dissipates the same ΔP each lap, so the absolute pressure repeats. Overpressure always arrives from a source outside the loop — backflow from the discharge side, a booster pump's shutoff head, cooling-medium in-leakage, or thermal expansion.
Does a discharge check valve count as protection for the 10 bar spec break?
Not on its own. A single check valve is a flow-preferential device, not a pressure barrier, and relief studies do not normally credit one against a 90 bar to 10 bar spec break. If the recycle takeoff sits downstream of that check, the recycle valve becomes the only barrier and a PSV on the low-pressure side is required.
Can I take pump-curve credit and reduce the required relieving rate?
Yes for a centrifugal pump: the required rate is the flow the pump delivers at the relieving pressure read from the certified curve. With relief set at 10 bar and 10 % accumulation, an 11 bar relieving pressure already exceeds a 10 bar shutoff head, so that pump contributes zero — provided you verify shutoff at maximum speed, maximum specific gravity, and a new impeller.
Does the cooler pressure drop need to be added to the relief set-pressure margin?
Yes when the recycle returns upstream of the exchanger. The exchanger inlet sits at pump suction pressure plus the cooler ΔP at recycle flow, so measure that differential at full 100 m³/h recycle and confirm the sum with the 10 bar loop floor stays below set pressure.