- More valves are damaged during construction and commissioning than in years of normal service — industry failure surveys consistently attribute the majority of first-year seat leakage to weld slag, grit and hydrotest water, not to operating duty.
- ASME B31.3 and B31.1 both call for hydrostatic testing at 1.5× design pressure, but API 598 limits seat testing to 1.1× the valve rating — so never hydrotest a line against a closed soft-seated valve.
- Effective line flushing needs a velocity of at least 3 m/s (and ideally 1.5× the normal operating velocity), with control valves pulled out and replaced by spool pieces before the first drop of flush water enters the system.
Ask any valve repair shop where their busiest month comes from and they will tell you: right after a new plant starts up. Valve commissioning is the most dangerous period in the life of any valve, and the damage done in those few chaotic weeks — flushing, hydrotesting, chemical cleaning, first heat-up — routinely exceeds anything the valve will see in a decade of steady operation. A Class 300 A216 WCB gate valve is built to handle 51.7 bar at 38°C for thirty years; it is not built to swallow a fistful of weld slag at 4 m/s.
This guide walks through each commissioning activity in the order it usually happens on site, and explains exactly what it does to valves, what the standards actually require, and the field practices that separate a clean start-up from a punch list full of leaking seats. All of it gets skipped under schedule pressure, every single time.
Why Construction Kills More Valves Than Operation Ever Will
A valve leaves the factory having passed an API 598 or ISO 5208 shell and seat test. It arrives on site clean, greased, with end protectors fitted. From that moment on, everything that happens to it is hostile: it sits in an open laydown yard through a wet winter, gets welded into a line full of grinding dust, is slammed shut against 1.5× design pressure during hydro, then left full of stagnant chlorinated water for four months.
The physics is simple. Seating surfaces are precision components — a metal-seated gate valve relies on lapped faces flat to within a few light bands, and a soft-seated ball valve relies on an unscratched PTFE or nylon seat. A single grain of silica carbide grit dragged across a seat under full differential pressure does more damage than 10,000 clean cycles. In my experience, if a valve leaks in the first six months, the root cause is almost never the valve. It is what the construction phase did to it.
Line Flushing: Winning the War Against Debris
Every new piping system contains debris. Weld slag and spatter, grinding swarf, sand from hydrotest water, mill scale, dropped nuts, gloves, and — more often than anyone admits — the remains of gaskets and desiccant bags. Line flushing exists to move that material out of the system before it reaches something expensive. The mistake is treating flushing as a formality rather than an engineered activity.
Flushing Velocity: The Number That Matters
Debris does not move at operating velocity; it moves at flushing velocity. The accepted rule is a minimum of 3 m/s in every part of the flush path, and preferably 1.5× the normal operating velocity of the line. In large-bore headers that can mean enormous water volumes, which is why flush plans divide systems into loops sized to what the temporary pumps can actually achieve. A flush at 1 m/s is theatre — it rearranges the debris without removing it.
Flush until the discharge runs clean through a 100-mesh screen at the outfall, then flush some more. Some specifications require two consecutive clean screens 15 minutes apart before sign-off. That criterion, written into the test pack, is what stops a night-shift supervisor calling a line clean because the water ‘looks alright’.
Remove the Valves That Cannot Defend Themselves
Control valves must come out before flushing. Full stop. A globe-pattern control valve is a debris trap by design — the tortuous flow path, the machined plug and cage, the stellited seat ring — and a flush through an installed control valve will wire-draw the trim. Replace each one with a fabricated spool piece of matching face-to-face dimension, flush hard, then reinstall the valve with fresh gaskets.
The same logic applies to soft-seated ball valves, wafer check valves with elastomer components, and anything with a fine internal clearance: positive displacement meters, steam traps, orifice plates. Where removal is genuinely impractical, lock ball valves in the fully open position so debris passes through the bore rather than embedding in the seats, and pull the internals from check valves. A swing check flapper left in during flushing will be hammered by slugs of debris-laden water and may never seal properly again.
Strainer Strategy: Start-Up Strainers and Cone Strainers
Flushing never gets everything, which is why start-up strainers exist. The workhorse is the conical (witch’s hat) strainer installed between flanges immediately upstream of pumps, compressors and control valves. Fit them point-upstream so debris collects in the cone without collapsing it, and always specify a perforated backing plate — a bare mesh cone will implode under a heavy debris load and send its own fragments downstream.
Mesh selection is a balance. For centrifugal pump protection during start-up, 40-mesh (about 400 microns) backed by perforated plate is typical; for control valve and compressor protection, 100-mesh or finer. Monitor differential pressure across every temporary strainer — a rise of 0.3 to 0.5 bar is your signal to shut down and clean it. And put strainer removal on the commissioning punch list with a date. Every engineer over forty has found a forgotten cone strainer, torn and half-dissolved, during a turnaround five years later.
Hydrotesting Valves: Pressures, Positions and the Rules People Break
Hydrostatic testing proves the mechanical integrity of the assembled system, and hydrotesting valves correctly is where more commissioning damage is done than anywhere else. The piping codes are explicit: ASME B31.3 (process piping) requires a hydrostatic test at 1.5× design pressure, corrected for temperature, held for at least 10 minutes while every joint is examined. ASME B31.1 (power piping) similarly specifies 1.5× design pressure. So far, so familiar.
Body Limits vs Seat Limits: The API 598 Distinction
Here is the distinction that gets missed. Under API 598, a valve shell is tested at 1.5× its maximum allowable pressure at 38°C — that is a body test, with the valve partially open and both ends pressurised. The seat test is only 1.1× the rated pressure. The seat was never designed, and never factory-proven, to hold 1.5× design across a closed obturator.
The consequence is a hard rule: never use a closed valve as the test boundary for a hydrostatic test at full test pressure, unless the valve’s seat rating explicitly covers it and the test procedure documents it. Hydro against a closed soft-seated ball valve and you will extrude the PTFE seat past the ball; hydro against a closed gate valve and you can overpressure the bonnet cavity or bend the gate. Test boundaries belong at blind flanges, spectacle blinds or test spades — hardware designed for the job.
Valve Positions During Hydro
Within the test envelope, valve positions matter. Gate and ball valves go fully or partially open so both sides of the seat see equal pressure. Globe valves should be partially open — never pinned shut, because test pressure under the disc of a closed globe valve can push the disc off the stem or damage the seat. Check valve flappers should be tied open or removed if the test pressurises the downstream side, otherwise the flapper takes reverse pressure it may not be rated for.
Where a closed valve genuinely must isolate live plant from a test section, apply the double block and bleed rule: two closed valves in series with an open, monitored bleed between them, and the test pressure limited to what the seats are rated to hold — 1.1× per API 598, not 1.5×. If the bleed weeps during the test, stop — the first seat has let go.
One smaller hydro casualty deserves a mention: double-seated valves such as through-conduit gates and trunnion ball valves can trap and pressure-lock test water in the body cavity. Confirm cavity relief is functional before anyone applies heat to that line.
Trapped Test Water: The Corrosion Nobody Sees Until It Is Too Late
The hydrotest ends, the pressure comes off, and everyone moves on. The water does not. It sits in valve body cavities, dead legs, low points and instrument tappings, sometimes for months, and it is quietly the most destructive substance on the entire project. Three mechanisms do the damage.
First, chloride attack on austenitic stainless steel. Test water for 304/316 and 316L systems should be limited to 50 ppm chloride at most — many specifications for stainless and sensitive alloys demand under 30 ppm, and potable-quality or demineralised water is the safe default. Chlorides concentrate as trapped water evaporates, and the result is pitting and, on hot lines later in life, chloride stress corrosion cracking initiated at commissioning. Never, ever, use untested well water or seawater on a stainless system to save a tanker cost.
Second, microbiologically influenced corrosion (MIC). Stagnant, untreated hydrotest water is an incubator for sulphate-reducing bacteria, which pit carbon steel and stainless alike at rates that can perforate a valve body wall in under a year. If water will stay in the system more than a few days, it needs a biocide and an oxygen scavenger, and the drain-down needs to happen on schedule.
Third, plain oxygen corrosion and freezing. The fix for all three is the same discipline: drain from every low point immediately after test, open valves to the half-open position so cavities drain, then dry — hot air blowing or, for critical service, dry air or nitrogen to a dew point of −20°C or better, verified with a hygrometer rather than optimism. Drying requirements belong in the test pack as a signed step, because an unsigned step is a skipped step.
Pneumatic Testing: Treat It Like the Explosive It Is
Sometimes hydro is impossible — refractory-lined lines, systems that cannot tolerate water, sub-zero conditions. Pneumatic testing is the fallback, and it deserves genuine fear: compressed gas stores orders of magnitude more energy than water at the same pressure, and a brittle failure under pneumatic test is a fragmentation event. ASME B31.3 sets the pneumatic test at 1.1× design pressure with tightly controlled stepwise pressurisation, and PCC-2 guidance requires calculated exclusion zones based on stored energy.
For valves specifically, pneumatic tests find smaller leaks than hydro — gas will pass a seat that water will not — so expect and document minor seat weepage rather than panicking about it. And the same seat-pressure rule applies with even less forgiveness: never use a closed valve as a pneumatic test boundary. A soft seat letting go under gas pressure gives no warning weep, just an instantaneous release.
Chemical Cleaning and Passivation: Protect the Trim
Steam generation circuits, compressor suction lines and lube oil systems typically get chemically cleaned — alkaline degreasing, acid pickling (commonly inhibited citric or hydrochloric), then neutralisation and passivation. Those chemistries are chosen to dissolve mill scale, and they do not distinguish between scale and valve internals. Inhibited HCl will still attack 13% chrome trim, stellite overlays can suffer selective attack at the fusion line, and any acid left in a gasket crevice keeps working long after the flush.
Best practice mirrors the flushing rule: remove control valves and instrumentation, and replace soft-seated valves in the cleaning loop with spools or sacrificial commissioning valves. Where valves must stay in line, verify the inhibitor is compatible with every trim material present, keep temperature and contact time inside the cleaning contractor’s qualified envelope, and confirm final rinse conductivity and pH at every low-point drain — not just the loop outlet.
Valve Preservation Between Hydro and Start-Up
On a big project, months can pass between a successful hydrotest and first process fluid. This is the preservation gap, and valves corrode faster in it than at any other time because they are full of humid air and hydrotest residue. A written preservation plan, with a named owner and a re-inspection interval, is not gold-plating — it is cheaper than one replacement control valve.
The toolkit is well established. Vapour-phase corrosion inhibitors (VCI) — powder, tablets or emitters placed inside drained systems — protect internal surfaces for 12 to 24 months if the volume is sealed. Nitrogen blanketing at a slight positive pressure, typically 0.2 to 0.5 barg, is the gold standard for critical systems and stainless circuits because it excludes both oxygen and moisture. Desiccant bags with humidity indicator cards suit valve cavities and actuator housings, provided someone actually changes them when the card turns pink.
Do not forget the externals. Exposed stems should be greased and wrapped or fitted with stem protectors — a corroded rising stem will destroy its own packing on the first stroke. Cycle every preserved valve a quarter-turn on a monthly schedule so seats do not take a set and stems do not seize, and log it.
After First Heat-Up: Torque Checks and Gland Adjustment
The first thermal cycle is a commissioning event in its own right. Bolted joints relax as gaskets seat and studs experience their first thermal transient; spiral-wound and graphite gaskets can lose a meaningful fraction of their initial stress. On hot service — steam, hot oil, anything above roughly 200°C — plan a hot-torque check of bonnet and body joints after the first cycle, at the fastener manufacturer’s specified values and in the correct cross-bolting sequence.
Gland packing needs the same attention. Compression packing consolidates with heat and stem movement, so a light, even take-up of the gland nuts after the first thermal cycle is expected maintenance, not a warranty claim. Adjust hot, in small increments, just enough to stop the weep — an over-tightened gland multiplies stem friction, and on an actuated valve that stolen thrust comes straight out of your seating load. Live-loaded packing with Belleville springs largely removes this chore, which is why it is worth specifying on critical and fugitive-emission service.
Commissioning Documentation: Test Packs and Valve Punch Lists
Paperwork feels like the enemy during commissioning, but the test pack is the only mechanism that forces the right sequence. A proper pack for each test system includes the marked-up P&ID showing test boundaries, the flushing procedure with velocity calculation and acceptance criterion, the hydro procedure with valve position table, water quality certificate, drying and preservation steps, and sign-offs for each. If the valve position table does not exist, positions get improvised at 2 a.m., and improvised positions are how soft seats die.
Run a dedicated valve punch list alongside the system punch list: every control valve reinstated with fresh gaskets, every cone strainer logged in and logged out, every locked-open valve returned to its P&ID position, every gland checked after first heat-up, every check valve internal refitted. Reconciling that list before start-up takes an afternoon. Not reconciling it costs a shutdown.
Horror Stories, and How to Avoid Your Own
Every commissioning engineer collects these. The combined-cycle plant that hydrotested against a closed Class 150 butterfly valve and peeled the seat liner out in one piece. The stainless cooling circuit filled from a fire hydrant — 400 ppm chloride — and left wet for a summer, discovered when pinhole leaks appeared through pitted valve bodies eight months into operation. The forgotten witch’s hat strainer that collapsed into a 6-inch pump suction and took the impeller with it.
None of these were valve failures, and none were exotic. Each traces back to a missing line in a procedure: no seat-pressure limit stated, no water quality certificate demanded, no strainer register kept. Commissioning teams are temporary and their lessons leave with them, which is why the defences must live in the documents rather than in anyone’s memory.
The Short Version
Flush fast and prove it clean, with the vulnerable valves out of the line. Hydrotest to code — 1.5× design per ASME B31.3 or B31.1 — but respect API 598’s 1.1× seat limit and never test against a closed soft seat. Drain, dry and preserve the moment the test ends, because trapped hydrotest water is a slow-motion failure. Re-torque and adjust glands after the first thermal cycle, and make the paperwork carry the discipline that schedule pressure will otherwise strip away.
Do those things and your valves will reach start-up in the condition the factory shipped them: tested, tight and ready for thirty years of unremarkable service. Skip them and you will meet your valve supplier’s service team much sooner than either of you would like.