Put the Expansion Tank on Return, or Static Exceeds 600 kPa

Where is the expansion tank connected? If the answer is the supply header, the manifolds are about to see pressurisation plus pump head. Conclusion 25 of the hydraulic basis is one of the very few decisions in the document that can be frozen today, and it says so in the same breath as the consequence of getting it wrong. The pressurisation point — expansion tank and make-up pump — must sit on the return side, near the secondary pump suction, so that every rack manifold lands on the pump's discharge but in the lower-pressure region close to return. Put it on the supply side instead and the manifold carries pressurisation pressure plus pump head; against the loop pressure drop of 235 to 450 kPa built up two sections earlier, manifold static can easily exceed 600 kPa, approaching the sealing boundary of the cold plates and quick disconnects [MODELED].

Read 600 kPa Against 250 kPa Working, Not the 1,000 kPa Rating

Three pressures appear in the protection table and they are not interchangeable. Secondary loop design pressure is 1,000 kPa system rating, graded an assumption pending the supplier [TARGET]. Normal working pressure at the rack manifold is recommended at 250 kPa gauge [TARGET]. The relief valve is set at 400 kPa, discharging to a dedicated recovery drum, with an explicit prohibition on discharging to the room floor or an ordinary drain [TARGET].

So 600 kPa is not read against the 1,000 kPa rating. It is 2.4 times the recommended working pressure at the point where it would occur [SENSITIVITY], which is the comparison that matters, because the working pressure recommendation is about seals and warranty rather than about burst.

Either the 400 kPa Relief Lifts or the Manifold Has None

Sitting 600 kPa beside a relief set at 400 kPa produces an inconsistency the source does not address: either the relief lifts under the mis-placed topology, or the relief is not located where the 600 kPa occurs. Both readings have consequences — one puts coolant into a recovery drum during normal operation, the other means the manifold has no relief protection at its own worst point. This is an observation, not a source finding, and it is raised as a question for the reviewer rather than resolved here.

A vendor packet on a different block makes the same class of error in the other direction. Mean heat-exchanger differential pressure is not local plate differential pressure. Counterflow ends can sit near +75 kPa and −75 kPa while the average looks balanced. Sixty kilopascals of suction gauge [MODELED] is not a proven net positive suction head. Average balance cannot close a one-sided pump-loss, relief, isolation, dead-head or fouling case. Those are the statements a return-side tank is meant to make answerable, and they are still owed.

Topology can freeze; volume cannot

The conclusion is topological. It does not need the rack count, the temperature difference, the liquid fraction or the supplier's curves; it needs only the loop's connection order, so the hydraulic basis recommends writing it into the unit specification and the secondary-side diagram immediately. The drawing instruction is unusually specific for a study-stage document: the expansion tank and make-up connection are to be drawn at the return-side pump suction and annotated on the drawing itself as a point that may not be moved. The item appears in the freezeable-decision register as F-02. Freezeable is not frozen.

Sizing, by contrast, is not available. Fill volume is item H-09 and is estimated at 25 to 45 m³ with ±40% uncertainty from an assumed temperature difference, ring and branch lengths and unit internal volumes [MODELED]. The hydraulic basis allows the expansion tank and make-up system to be selected preliminarily on the upper bound of 45 m³, and refuses to let the same estimate support a coolant purchase quantity or a leak-rate criterion, because that band is also the denominator of the leak judgement.

Allowable Static Pressure Is Still Owed by Both Vendors

Item OEM-08 asks the rack supplier for the allowable static pressure range — working limits, proof pressure and test pressure — and records that until it arrives, both the 250 kPa working recommendation and the pressurisation topology cannot be verified. The recommendation is therefore a facility-side position waiting on an equipment boundary, not a checked design.

The distribution unit review asks the same question of the other vendor and lists what is owed: manufacturer confirmation of required differential pressure limits covering one-sided pump loss, relief, isolation, dead-head, transient and fouling conditions, together with available and required net positive suction head, expansion volume and protection settings. Available suction head is where the return-side location repays itself, and it is the one part of this decision that a pressurisation topology cannot be argued without.

The parametric engine records the working fluid's 1,012 kg/m³ and 3.921 kJ/(kg·K) as engineering assumptions held under item H-05 [MODELED]. It states in its own words that without manufacturer curves, pipe pressure drops, valve authority and pump curves it cannot determine a pump or a final operating point. A uniform heat-and-flow split is for comparison only. It is not a pressurisation calculation.

Limits and open items

Confirmed on the current record: pressurisation belongs on the return, near pump suction; placing it on supply adds pump head to manifold static, and manifold static can easily exceed 600 kPa. [FACT]

Sensitivity, not a source assertion: 600 kPa is 2.4 times the 250 kPa working recommendation. [SENSITIVITY]

Recommendation, not confirmed: working pressure at the rack manifold not more than 250 kPa gauge; relief at 400 kPa to a dedicated drum, not the floor. [TARGET]

Modelled and not frozen: 235 to 450 kPa loop drop, and 25 to 45 m³ fill with ±40% uncertainty. The calculation that would replace the estimate, fill volume with expansion vessel volume and pressurisation, is registered as item 07A-CALC-14 and waits on both the rack count and the layout. [MODELED]

Held open: OEM-08, H-05, H-09, and the vendor packet's unanswered NPSH and protection-settings request [HOLD]. One condition would make the topology argument moot rather than wrong. If the rack supplier returns an allowable static range whose upper limit sits above the supply-side case, the seal-boundary argument weakens and the choice reverts to pump suction conditions and vessel sizing alone. If it returns a limit at or below the recommended 250 kPa, then the return-side location stops being a recommendation and becomes the only arrangement that keeps the manifold inside its own equipment rating. Nothing above reflects Engineer of Record review, authority approval, supplier confirmation, commissioning acceptance or observed operation, and no vessel or valve here is a procurement release.

© 2026 K&K Data Service Inc. All rights reserved. Reproduction or republication is permitted only with clear attribution to K&K Data Service Inc. and a working hyperlink to the canonical URL of this article. Excerpts must preserve the technical context, maturity labels, assumptions, and limitations. No excerpt may imply project approval, field validation, certification, or endorsement that the original article does not state.

Source: K&K Data Service Inc., “Put the Expansion Tank on Return, or Static Exceeds 600 kPa,” https://www.kkdatasvc.com/lab/hydraulics-and-transients/put-the-expansion-tank-on-return-or-static-exceeds-600-kpa/.