Facility Piping & Pressure-System Mechanical Engineering

Everything earlier in Part 5 treated cooling as a thermodynamics problem: how much heat, at what temperature, rejected how. This chapter treats what is left after the thermodynamics is settled — the charged pipe that actually carries the coolant, under pressure, through a building full of energized racks, for twenty years of thermal cycles and the occasional pump trip. The moment a hall is plumbed for direct-to-chip liquid (Chapter 5.4), the facility has acquired a pressure system, and a pressure system is governed by mechanical codes, surge physics, pump hydraulics, corrosion electrochemistry, and structural restraint — not by ASHRAE setpoints. These are mature disciplines from the process-plant and building-services worlds, but AI density imports them into a context they were not written for: water within centimeters of a 1,400 A busbar, flow rates that push velocity into the erosion regime, and a goodput penalty measured in MW-hours for every hour a leak or a tripped pump takes a row offline.

Each fork in this chapter carries a downstream cost. Build to the wrong code section and your weld QA and hydrotest regime are either over-burdened or, worse, under-specified for the duty. Skip the surge model and a routine power blip becomes a burst manifold over a powered rack. Mis-match the pump to the system curve and you either cavitate at the low-pressure node or dead-head into a throttled, energy-wasting operating point. Couple aluminum cold-plate manifolds to copper headers through a steel CDU without isolation and the galvanic cell eats the least-noble metal on a schedule you did not budget for. None of these are exotic; all of them are how liquid-cooled halls actually fail in the field.

The first irreversible decision is jurisdictional: what code do the charged loops live under? In North America the governing family is ASME B31, the Code for Pressure Piping, and the live fork is which section. For closed-loop chilled-water and technology-cooling-system (TCS) loops carrying water or water-glycol at building-services pressures and temperatures, B31.9 (Building Services Piping) is the natural fit — it was written for HVAC and process-utility piping inside buildings and carries lighter design and examination requirements. But the instant the service becomes more demanding — higher pressure, a fluid classified as hazardous, or an owner/AHJ who wants process-grade rigor — the design migrates to B31.3 (Process Piping), with its fluid-service categories, stricter weld-examination percentages, and more onerous documentation. The choice is not cosmetic: B31.3 mandates more NDE, tighter qualification, and a heavier inspection paper trail than B31.9 for the same nominal pipe.

In the EU and much of the rest of the world the framework is different in kind, not just in number. The Pressure Equipment Directive (PED, 2014/68/EU) is law, not a voluntary consensus code: any pressure equipment above the directive's threshold (PS > 0.5 bar) must be CE-marked, and the assembly is sorted into hazard categories (I–IV) by pressure × volume and fluid group, which sets the conformity-assessment module and the degree of Notified Body involvement. EN 13480 (Metallic Industrial Piping) is the harmonized standard that demonstrates PED compliance for piping, the way EN 13445 does for vessels. The practical consequence for a global operator: a reference design validated to B31.9 in Virginia cannot simply be shipped to Frankfurt or Dublin — the EU loop must be re-evaluated for PED category, and a Category III/IV assembly drags in a Notified Body, a CE-marked assembly file, and a different acceptance regime. This is a recurring tax on multi-region standardization, and it is cheapest to pay at design time by engineering to the stricter of the two from the start.

Steady-state pressure is the easy part — every loop is designed to a static pressure rating. The hazard is the transient: the pressure spike when a velocity changes suddenly. The first-order magnitude is the Joukowsky surge, ΔP = ρ·a·ΔV, where ρ is fluid density, a is the pressure-wave speed in the fluid-filled pipe (roughly 900–1,500 m/s in water-filled metal pipe, lower in flexible hose), and ΔV is the velocity change. The brutal implication: a loop running at a benign 2 m/s that stops instantly can spike on the order of 20+ bar above its operating pressure — multiples of the static rating — independent of how thick the pipe is. In a data hall that spike lands on manifolds, quick-disconnects, and cold plates sitting directly above energized GPUs.

The two events that drive it are valve closure and pump trip, and they behave oppositely. A fast valve closure sends a positive (compression) wave upstream toward the source. A pump trip — far more common, because it rides on every utility blip, ATS transfer, and breaker event — sends a negative (rarefaction) wave downstream from the discharge as fluid momentum carries on without the driving head. The negative wave is the dangerous one: if local pressure falls to the vapor pressure of the coolant, the column separates (a vapor cavity forms), and when that cavity collapses the rejoining columns slam together and produce a surge that can exceed the Joukowsky estimate. This is precisely the column-separation case that the single-equation estimate does not capture, which is why a hand calc is a screen, not a design.

The fork is therefore: do you run a transient (method-of-characteristics) surge model, or do you trust a static rating with a fudge factor? For any non-trivial charged loop in a dense hall the answer is the model — software in the lineage of AFT Impulse and equivalents — exercised across the credible event set (single and simultaneous pump trips, the worst-case valve slam, an emergency-stop that drops all pumps at once). The model tells you the maximum and minimum pressure envelope at every node, including the negative-pressure nodes where column separation and check-valve slam live, which a steady-state rating never reveals.

The pump is the heart of every charged loop, and two failures recur. The first is cavitation: if the absolute pressure at the pump suction falls below the coolant's vapor pressure, vapor bubbles form and then implode on the impeller, pitting the metal, collapsing the head, and roaring like gravel in the casing. The guardrail is the NPSH inequality: NPSH-available (set by the loop — suction static pressure, fluid temperature/vapor pressure, suction-side losses) must exceed NPSH-required (a property of the pump at its flow) with margin, typically a meaningful fraction of a meter or more above NPSHr at every operating point. In a closed, pressurized data-center loop NPSHa is usually engineered with a generous expansion-tank/fill pressure — but the worst-case node is the part-load, hot-coolant, redundancy-degraded state, and that is the one to check.

The second failure is a pump/system mismatch. A centrifugal pump delivers head that falls as flow rises (the pump curve); the loop demands head that rises roughly with the square of flow (the system curve, dominated by friction). The pump runs where the two curves cross — the operating point. Get this wrong and the consequences are specific: select a pump that lands far left of its best-efficiency point (BEP) and you waste energy, run hot, and invite recirculation and shaft load; land far right and you risk cavitation and motor overload. Worse, the operating point moves: a CDU loop has variable flow as racks throttle and isolate, and the system curve shifts every time a quick-disconnect is mated or a row is valved out. Constant-speed pumps chase this with throttling (wasteful); variable-speed (VFD) pumps with differential-pressure control ride the system-curve family efficiently and are the 2026 default for secondary loops — but the control loop interacts with thermal-control transients and can hunt, which is why this couples directly to Chapter 5.12 (controls stability) and is set during commissioning in Chapter 13.5.

A charged loop is an electrochemical cell waiting to be completed. Galvanic corrosion occurs when two dissimilar metals are electrically connected in a conductive fluid: the less-noble metal (the anode) corrodes preferentially to protect the more-noble one (the cathode), at a rate that grows with the potential gap in the galvanic series and shrinks with the area ratio. Data-center loops are full of dissimilar couples — aluminum cold-plate manifolds, copper cold plates and headers, stainless CDU internals, carbon-steel facility piping, brass fittings — and the worst sin is a small anode, large cathode geometry (e.g., an aluminum fitting feeding a large copper header), which concentrates the entire galvanic current onto a small area and perforates it fast. The defenses are: (1) keep the wetted-metal set as compatible as possible, (2) break the electrical path with dielectric unions and isolation kits at every unavoidable dissimilar junction, and (3) dose the coolant with corrosion inhibitors and hold water chemistry (conductivity, pH, biocide) inside the vendor envelope — a poorly maintained PG-25-class fluid is both more conductive and more aggressive.

Erosion-corrosion is the velocity-driven cousin: fast or turbulent flow strips the protective oxide film off the metal faster than it can re-form, exposing fresh metal in a self-accelerating loop, with the worst damage at elbows, tees, reducers, and the back of partially-throttled valves where flow impinges. Copper is notoriously velocity-sensitive — practitioners cap copper-tube velocity in the low single-digit m/s range, lower still for hot water — and this is exactly where the chip-driven high flow rates of dense racks collide with material limits. The fork: upsize the header to drop velocity, or move to a more erosion-tolerant material (stainless) and pay the cost. Either is fine; ignoring it is not, because erosion-corrosion is invisible until a pinhole leak appears over a rack.

Charged pipe is heavy and it moves. Thermal expansion is the routine driver: a long warm-water header heats and grows on every load swing, and if it is rigidly anchored at both ends without relief, the thermal stress goes into the pipe, the welds, and the equipment nozzles — which is how a CDU connection or a flange gasket fails after a few hundred cycles rather than a few thousand. The classical solution is a deliberate flexibility scheme: expansion loops, expansion joints (bellows), anchors, and guides laid out so the pipe is free to grow in a controlled direction while being restrained against everything else. This is a formal pipe-stress problem (the same caesar-II-class analysis the process world runs), and in a dense hall it is complicated by tight pipe racks, CDU galleries, and the need to keep growth away from busbars and trays.

The second driver is weight and dynamics. Water-filled pipe is far heavier than the empty pipe the structural model sometimes assumes — a large header full of coolant is a significant distributed load, and an entire plumbed-and-charged data hall adds a coolant mass the slab and the supports must carry. That is the explicit handoff to the structural basis: the floor-loading and weight envelope for plumbed, charged, energized racks plus their distribution piping is owned in Chapter 6.2, and the piping engineer must feed the wet weights and dynamic (surge, seismic) loads into it rather than treating the slab as a given. The third driver is seismic: in any meaningful seismic zone, charged pipe needs transverse and longitudinal bracing, flexible connectors at equipment, and slack at building seismic joints, because a rigidly-run, un-braced charged line is exactly the thing that ruptures in an event and floods a hall of live racks. Anchoring, guides, expansion provision, and seismic restraint are one coordinated system — designed to a stress basis, not eyeballed off a hanger schedule.

The code basis chosen at the top of this chapter cashes out here, at acceptance. Charged piping is only as good as its joints, and the joints are proven by three layers. Weld QA requires qualified procedures (WPS), qualified welders, and procedure-qualification records (PQR) per the governing code — B31.x or EN 13480 — so that the metallurgy of every weld is reproducible and traceable. Non-destructive examination (NDE) verifies the welds that were made: visual on everything, plus radiographic (RT) or ultrasonic (UT) on a percentage that the code ties to fluid-service category — light for B31.9 building-services water, heavier for B31.3 process service or a PED Category III/IV assembly. The deeper the service category you chose, the more film and the more paper.

The final gate is the pressure test. Hydrostatic testing — filling with water and pressurizing, typically to ~1.5x design — is the default because water is nearly incompressible: a failure leaks rather than explodes, releasing little stored energy. Pneumatic testing (with air or inert gas) is used where the system cannot tolerate water or cannot be fully drained, but it is far more hazardous — compressed gas stores enormous energy, so pneumatic tests demand exclusion zones, stepped pressurization, and explicit engineering justification. The decision between them is a safety decision, not a convenience one. And it does not stand alone: the pressure test is the end of fabrication but the start of commissioning. The cleanliness, flush, fill, and chemistry steps that take a hydrotested-but-dirty loop to a coolant-ready, leak-monitored, accepted system are the subject of Chapter 13.5 (cooling acceptance) and tie back to leak-detection and water-quality monitoring in Chapter 5.11.