The Valve That Said Open and Was Not

Why this matters

A cooling loop on a heat exchanger was running the process hot. The controller was calling for 100 percent, the positioner was reporting 100 percent, and the stem indicator on the valve body was sitting at the top of its travel. Three separate things agreed the valve was open. Nothing was going through it.

The shop had already burned three site visits and roughly 7 labor hours on the controller and the temperature sensor before anyone measured across the valve. The measurement that settled it took about 1.5 hours, which is close to five times less effort than had already gone into the wrong question. The reason it took three visits is the thing worth carrying out of this: the word "open" is asserted at four different points in that chain, each one is a different claim, and the first three can all be true while the fourth is false.

Stop the heat before you diagnose anything

A process temperature climbing on a loop that has lost its cooling is not a diagnostic situation yet. Cut the heat input first, at its own control and not at the cooling valve, and confirm the input is actually off at the fired or electrical device rather than at a setpoint on a screen. A vessel that keeps absorbing heat with no flow through its cooling side raises pressure as well as temperature, and the relief device is a last resort, not a control strategy.

Once the input is off, do not use a relief valve, a rupture disc line, or a vent path as a temporary bypass. Those exist to open at a set condition and nothing else, and running normal flow through one fouls the seat so it will not reseat when it matters.

What containment bought, and what it cost

The skid had a manual bypass around the control valve, which is what it was there for. Opening it restored cooling flow, and the exchanger outlet temperature started falling within a few minutes. That confirmed two things at once and it is worth being explicit about both, because a containment move that also carries information is worth choosing over one that does not.

First, the cooling supply itself was fine: the pump was running, the supply line was cold, and the loop had capacity. Second, the exchanger and the return path were not plugged, because flow through the bypass entered and left the same exchanger. Everything downstream of the branch was cleared by one action.

The cost of containment was that the process was now running on a manual valve with no modulation, so somebody had to sit with it. That is a real constraint on how long you get to think, and it is why the next step was a measurement rather than a parts swap.

Finding the restriction instead of arguing about the valve

With the bypass carrying the load, the question became narrow: is the control valve the restriction, or is it open and something else is in the way? That is a differential-pressure question, and the branch had pressure taps immediately upstream and downstream of the valve body.

With the controller commanding 100 percent:

• Upstream tap: 42 psi
• Downstream tap: 9 psi
• Differential across the valve body: 42 - 9 = 33 psi
• Total dynamic differential available across the whole branch, pump discharge to return: 36 psi

So the valve was absorbing 33 of the 36 psi available, about 92 percent of the loop's dynamic loss, while nominally wide open. A control valve is commonly sized to take on the order of a quarter to a half of the branch's dynamic loss at design flow, and that is a sizing convention rather than a law of physics, but nothing in that convention puts a full-open valve at 92 percent. Read on its own that number says the valve is the restriction. Read against the bypass result from the step before, which had already cleared the exchanger and the return, it says the restriction is inside that valve body and nowhere else.

Take that reading with the taps confirmed clear by cracking each one briefly to a drain and watching it run, with the drain pointed away from you and into a container, because a plugged tap reads the last pressure it saw and will hand you a fabricated differential with total confidence.

The segment that should have been cold

The branch had no flow meter, which is common and which is why the temperature check earned its place. The short run of pipe between the control valve and the exchanger inlet should sit close to supply temperature if anything at all is moving through it.

• Cooling supply header: 55 F
• Return header: 68 F
• Exchanger shell: 118 F
• The pipe segment between the valve and the exchanger inlet: 96 F

A segment carrying supply water would read within a few degrees of 55 F, not 41 degrees above it. Sitting at 96 F, between supply and shell temperature, that pipe was being heated by conduction from the exchanger and cooled by nothing, which is what a dead leg does.

One condition on that reading, because it is where this measurement usually goes wrong: those were contact-probe readings on bare pipe. An infrared thermometer aimed at bright bare copper or polished stainless reads low, sometimes badly low, because the surface emissivity is far from the value the instrument assumes. If infrared is what you have, read a patch of flat dark tape or paint applied to the pipe, and say in your notes which method you used.

Proving the stem actually moved

At this point the valve is the restriction. That still leaves two explanations: the actuator never travelled, or it travelled and the valve stayed shut anyway.

Travel is measurable without opening anything. Command 0 percent, mark the exposed stem against a fixed reference on the yoke, command 100 percent, and measure the change. The valve's rated travel was 1.5 inches and the exposed stem grew by 1.5 inches. Take that measurement with a rule against a fixed point rather than bracing a hand on the body, which was still at process temperature and will burn on contact.

So the actuator did its whole job. The positioner feedback and the body indicator were both telling the truth. Three assertions of "open" were now confirmed rather than merely reported, and the valve was still passing nothing.

The test that separated the last two explanations

The remaining candidates were a blocked port inside the body, something like debris or a dislodged gasket, and a plug that was no longer connected to the stem.

The distinguishing evidence is the force signature at crack-open. On an unbalanced plug sitting on its seat with differential pressure across it, the actuator has to overcome that unbalance force to lift the plug off the seat, and then the force falls away once the plug is clear. On a valve with a positioner you see this as a distinct rise in output pressure right at the break-open point, followed by a drop back as travel continues. It is one of the more reliable things a positioner gauge tells you.

Stroking slowly from 0 to 100 percent with the bypass carrying the load, the positioner output climbed smoothly through the whole stroke with no break-open step anywhere in it. The stem was lifting against packing friction and nothing else. A blocked port would still have produced the break-open signature, because the plug would still have been attached and still have been pinned to its seat by differential. A plug that is not attached to the stem produces exactly what was observed.

That is a hypothesis with a confirmed mechanism, not a guess, and it is what justified opening the valve rather than replacing it whole.

Opening it up

Before the bonnet came off: both isolation valves closed and their positions verified at the stems rather than at handle labels, the branch drained through a low-point drain and watched until it ran dry, the body cooled below hand-hot, and the residual pressure confirmed at a gauge that is on the isolated side rather than one that a closed valve has cut off from the body. Isolation and blocking of that stored pressure and of the actuator spring falls under 29 CFR 1910.147; lock and tag the isolation points, do not just close them, and bleed the instrument air to the actuator so the spring cannot drive the stem while hands are in the bonnet.

Inside, the plug was sitting on the seat, free of the stem. The retaining pin that keys the plug to the stem was absent. The service record showed a repack on that valve two visits earlier. It recorded the packing part and the torque on the gland. It recorded nothing about the plug-to-stem joint, because nobody had a field on the form for it, and the gap in the record is part of the finding: a repack requires pulling the stem, which means the plug-to-stem joint was disturbed and never verified on reassembly.

The four different claims that all say "open"

The word travels down the chain and changes meaning at every step, and the shop had been treating all four as one:

1. Commanded open. The controller's output. This is an intention. It survives a dead I/P converter, a severed signal wire, and a valve that is not there.
2. Actuator open. The positioner's feedback. This asserts that the stem is where it was told to be. It survives everything downstream of the stem.
3. Stem travelled. A measurement against the body, which is stronger than the feedback because it is independent of the feedback linkage.
4. Flowing. A differential, a flow reading, or a temperature that moved.

Only the fourth one is about the process, and the first three cannot substitute for it. Each one you confirm eliminates a stretch of the chain and nothing beyond it. In this case the shop confirmed the first two on the first visit, concluded the valve was open, and spent the next two visits looking for a reason the process would run hot with cooling flow it did not have.

The general form is worth keeping past valves. Any time a system reports that it did something, ask which of those four kinds of statement you are holding: a command, an internal feedback, an independent measurement of the mechanism, or an observation of the result. The gap between the third and the fourth is where the mechanical failures live, and it is the one nobody instruments.

References

• 29 CFR 1910.147 for control of hazardous energy, including stored pressure, thermal energy and actuator spring energy prior to opening a pressure boundary
• Instrument Society of America and general trade-standard practice for control valve sizing conventions and differential-pressure allocation
• Manufacturer documentation for plug-to-stem retention and post-repack reassembly checks
• See related: What a Positioner Adds; What an Actuator Does and How Far It Actually Went; How Valves Fail