The drip shows up at the NPT end, and only after the shipping plug comes out. That single detail fixes the geometry of the problem: the socket-weld end is welded into the pressurized spool, the wedge is between the two ends, and the threaded end is a dead-ended cavity. Follow the water backwards from the port and there is exactly one path it can have taken.
Where does the water actually exit?
Before arguing about acceptance criteria, fix the exit point. A gate valve on a socket-weld spool has five separate leak paths, and only one of them is a closure (seat) leak. Wipe every external surface dry, re-pressurize, and look at each location under a light before touching anything.
| Exit point | Path | Typical cause | Next check |
|---|---|---|---|
| Open NPT port, wedge closed | Past the downstream seat | Debris on the seat face, wedge not driven home, seat ring distorted by weld heat | Cycle and re-close under pressure |
| Body-to-bonnet joint | Gasket / bolting | Bolt relaxation, gasket damaged during handling | Retorque cold, cross pattern |
| Around the stem at the gland | Packing | Gland follower never adjusted after commissioning cycles | Take up gland nuts evenly |
| Socket-weld fillet, toe or root | Weld defect | Socket bottomed, lack of fusion, root crack from shrinkage | PT the toe, RT the root |
| NPT plug threads | Thread seal | Shipping plug sealant only | Not a valve defect |
Water at the open port with the wedge closed is a closure leak. Water anywhere else is a shell or packing issue and gets a different repair and a different acceptance clause. Do not merge the two.
Is it stored volume or a live leak?
With the plug installed and the wedge shut, the volume between the wedge and the plug is a sealed dead pocket. Any seat leakage bleeds into that pocket, pressurizes it, and then stops when it equalizes with the test pressure. Nothing is visible from outside. Pull the plug and that stored volume decompresses in one gush, followed by the true steady-state leak rate. That is precisely why the leak "appeared" only after plug removal, and why it can be missed on the other seventeen valves if their plugs stay in.
A 3/4in body cavity holds only a few millilitres, so it empties within seconds. Anything still moving after that is real flow. Take the reading this way:
- Hold the line at full test pressure with the plug removed.
- Wipe the port bore and the face dry with a lint-free cloth.
A sustained one drop per minute is roughly 3 mL/hour of continuous flow across the seat, and it will not stop on its own.
Which acceptance criterion applies - API 598 or the field test?
Two different regimes are being conflated here. API 602 cl. 5.2 sends pressure testing of compact steel gate valves to API 598. But API 598 cl. 3.1 requires those tests to be performed at the manufacturer's plant. A hydrotest on an erected spool is a piping system leak test governed by the piping code, not an API 598 valve test, and the valve is normally in the open position for it. Establish which document the inspector is writing the punch item against.
| Test (API 598 Table 1-A) | Medium | Acceptance for NPS ≤ 2 |
|---|---|---|
| 3.1 Shell | Water | No visible leakage (cl. 3.9.1) |
| 3.2 Back seat | Water | No visible leakage (cl. 3.9.1) |
| 3.3 Low-pressure closure | Air | Metal seated: less than one bubble. Resilient seated: zero |
| High-pressure closure (Table 5) | Water | Metal seated: zero drops, no visible leakage. Resilient seated: zero |
For a metal-seated NPS 3/4 gate valve, API 598 Table 5 gives no drop allowance at all. One drop per minute is a rejection under every branch of that table. There is no "minor leakage" tolerance to fall back on. If the project specification instead invokes API 6D (1999), cl. 10.4.4 requires no visible leak for soft-seated valves and refers metal-seated valves to ISO 5208, Rate D, which is the only route by which a measured, non-zero rate becomes acceptable. Check the valve datasheet for which document was purchased against before conceding or rejecting.
Why did 3 of 20 fail and 17 pass?
A 3 in 20 scatter on identical valves from the same batch points at the installation, not the design. Two mechanisms produce it on small forged socket-weld bodies.
First, weld heat. On a 3/4in Class 800 body the fillet weld sits within a few tens of millimetres of the seat ring. With the trim assembled and the wedge in the body, that heat reaches the seat pocket, and unrestrained thermal expansion of the ring against a cold wedge leaves a permanent distortion of the seat contact circle. The valve leaves the shop passing a zero-drop closure test and fails after welding. Which valves fail depends on welder technique, position, and how long the arc dwelled on that side - hence the scatter.
Second, debris. Hydrotest water carries mill scale, weld spatter, and cutting swarf. A single particle bridging a metal seat line holds a drop-per-minute flow indefinitely. This is the cheapest failure to clear and should be eliminated before any disassembly.
What welding practice stops the next batch failing?
Strip the valve before it goes in the line. Remove the bonnet, stem and wedge from the body, weld the empty body into place, and reassemble afterwards with a new gasket. The trim never sees the heat and the seat pocket is free to expand and contract without a wedge trapped in it.
Set the socket gap. A 1.5 mm (1/16in) gap between the pipe end and the socket shoulder is required before welding - see ANSI/ASME B31.3-2002, Fig. 328.5.2C. Bottoming the pipe puts the shrinking weld metal in direct tension against a rigid joint and cracks the root. Random RT verifies the gap survived fit-up and tacking.
Preheat evenly. Once the body is tacked, bring it up uniformly with the torch before completing the fillet, rather than heating one side under the arc. Uneven heating of a small forged body is what warps the seat pocket.
Repair, re-test, and final verification
- Depressurize and drain the spool. Confirm zero pressure at a vent before opening anything.
- Back the stem off the backseat, then close the wedge with the handwheel only - no impact tools, no cheater bar.
- Cycle the valve fully open to fully closed three to five times to sweep the seat faces, flushing the upstream side between cycles.
- Re-plug, re-pressurize to test pressure, hold until stable, then pull the plug, wipe dry and re-count. Most debris-related leaks clear here.
- Still dripping: remove the bonnet and inspect the wedge and both seat faces for scoring, embedded scale, and a broken contact line. Blue-check the seat contact ring for continuity around the full circle.
- Distorted or interrupted seat contact on a 3/4in Class 800 valve is a replacement, not a lapping job. Replace the valve; strip the trim from the new body before welding it in.
- Leakage at the fillet weld or body-bonnet joint instead: PT the weld toe, RT the root, and cut out and re-weld to the B31.3 fit-up gap rather than adding cover passes over a leaking root.
Where the specification also calls for the low-pressure closure test, follow with air at the API 598 closure pressure and confirm less than one bubble under water on each metal-seated valve.
FAQ
Why does API 598 allow zero drops on a small gate valve when larger valves get an allowance?
Table 5 of API 598 sets the maximum allowable closure leakage for NPS ≤ 2 metal-seated valves at zero drops - no visible leakage - and zero for all resilient-seated valves. The seat contact length on a 3/4in valve is short enough that any measurable flow indicates a defect, not a tolerance.
Why does the leak only appear after the shipping plug is removed?
With the plug installed and the wedge closed, the cavity between wedge and plug is sealed. Seat leakage pressurizes that dead volume until it equalizes with the test pressure and then stops. Removing the plug vents it, releasing the stored volume first and then the steady leak rate.
Why did only 3 of 20 identical valves leak?
Scatter within one batch points to installation variables: weld heat input reaching the seat pocket with the trim still in the body, and debris from hydrotest water bridging the metal seat line. Both vary valve to valve with welder technique and flushing.
Why does a socket-weld joint need a 1.5 mm gap before welding?
ASME B31.3-2002 Fig. 328.5.2C requires roughly 1.5 mm (1/16in) between the pipe end and the socket shoulder so weld shrinkage is not restrained by a bottomed joint, which otherwise cracks the root. Random RT verifies the gap survived tack-up.
Why can a site hydrotest result not be judged directly against API 598?
API 598 cl. 3.1 requires the shell, backseat and closure tests to be performed at the manufacturer's plant, on an unwelded valve in a fixture. A test on an erected spool is a piping system leak test; the applicable acceptance comes from the piping code and the project specification, with API 598 Table 5 used as the reference for what the valve was originally certified to.