Why Throttling a Valve Changes More Than Flow

Why this matters

A tech closes a valve a third of the way because a branch is running hot, the complaint goes away, and everyone moves on. Six months later the pump seal is weeping, the valve seat is wire-drawn, and the far branch that was fine is now short. Nothing failed randomly. Closing that valve did not subtract flow from the system, it changed the whole balance of head, velocity, power and wear, and every one of those changes was predictable from the machine's curve. Understanding what a partly closed valve actually does is the difference between balancing a system and quietly damaging it.

Before you move any valve on a live system

Stored pressure is the hazard that gets people, and the action is to know the relief path before you turn the handle. On a positive-displacement pump (gear, vane, piston, progressing cavity, and most diaphragm metering pumps), throttling the discharge does not reduce flow, it raises pressure until the relief valve lifts or something ruptures: confirm the relief is present, correctly set and not isolated before you close anything on the discharge side, and if you cannot confirm it, stop the pump instead of throttling it.

On a running centrifugal pump, do not use the discharge valve as a stop. Fully closing it traps the water while the shaft keeps putting energy into it, and that energy leaves as heat in a small volume that can flash when the valve is reopened; shut the pump down first, then close the valve. Where the valve you are touching is part of an isolation for service rather than a flow adjustment, the mechanical isolation and stored-energy requirements of 29 CFR 1910.147 apply, which means lock and tag the energy isolating device, not just switch it off, and bleed the trapped pressure before breaking any joint. If the same job puts you in the starter or the disconnect, that is electrical work under 29 CFR 1910.333(b)(2), and the live-dead-live proving sequence on your meter comes from NFPA 70E-2021, 120.5.

Hot fluid scalds through a cracked packing nut. Loosening a packing gland or a bonnet on a hot loop sprays at body height: let the section cool or isolate and drain it first, and if you must repack hot, stand out of the plane of the stem, wear a face shield rather than glasses, and cover forearms.

A valve does not remove flow, it converts head into heat

A control valve is a variable resistance in series with everything else in the loop. It does not take fluid out of the circuit. What it does is dissipate part of the head the machine developed, turning it into a tiny temperature rise across the valve, and it does that whether the valve is a globe, a ball, a butterfly or a balancing valve with a memory stop.

Two consequences follow immediately, and they are the ones techs miss.

First, the pump or fan does not know that you wanted less flow. It only sees a steeper system, so it slides back along its own curve to a new intersection point at lower flow and higher developed head. The head did not go away; the pump is now making more of it and you are burning the extra in the valve.

Second, all of the energy you are dissipating shows up somewhere physical. In a valve that is nearly shut, the fluid accelerates hard through a small gap and decelerates on the other side. That acceleration is where cavitation starts in liquids, where noise starts in air and water alike, and where the seat and the trim erode. A valve used as a permanent flow restrictor is a wear part in a way a fully open valve never is.

Same rule, two opposite outcomes

The rule is: throttling moves the operating point along the machine's own curve. Whether that helps or hurts depends entirely on the shape of that curve.

Case one, a constant-speed centrifugal pump. Its curve falls as flow rises. Add resistance and the point walks up and to the left: less flow, more head, and shaft power that drops only a little because the head rise partly cancels the flow drop. Throttling works here in the sense that flow really does fall, and it is the reason balancing valves exist.

Case two, a positive-displacement pump. Its curve is very nearly a vertical line. Add resistance and flow barely moves while pressure climbs to whatever the relief or the weakest joint allows. Throttling does not work here at all, and treating it as a flow control is how casings crack. Reduce displacement or speed instead.

Fans split the same way, by wheel type rather than by pump family. A forward-curved centrifugal wheel draws more power as flow rises, so closing a damper reduces motor draw. Backward-inclined and airfoil wheels have a non-overloading power curve that peaks near mid-flow, so where damper closure lands relative to that peak decides whether draw falls, and the fan's own published power curve is the only place to settle it.

Worked example: trimming an over-flowing branch

A constant-speed centrifugal circulator serves a loop that was designed for 60 gpm. The balancing valve on the branch was never set, and a flow meter reads 78 gpm.

From the pump's published curve at this impeller and speed, three points: 60 gpm at 42 ft of head, 70 gpm at 38 ft, 78 gpm at 33 ft. So the system as built, with the balance valve wide open, is sitting at 78 gpm and 33 ft.

Model the piping and terminal as a square-law resistance, which is a fair approximation for a fully turbulent water loop: head loss equals k times flow squared. From the measured point, k equals 33 divided by 78 squared, which is 33 divided by 6084, or 0.00542 ft per gpm squared.

At the 60 gpm we actually want, that same piping needs 0.00542 times 3600, which is 19.5 ft. The pump at 60 gpm makes 42 ft. The gap, 42 minus 19.5, is 22.5 ft, and that is what the balancing valve has to absorb.

Now read what that means. The valve is dissipating 22.5 of the 42 ft the pump is developing at the new point, about 54 percent of it. Flow fell by 18 gpm out of 78, which is 23 percent. And the hydraulic power product, flow times head, went from 78 times 33, or 2574, to 60 times 42, or 2520. That is a 2 percent reduction in hydraulic power for a 23 percent reduction in flow.

Compare that against what speed control would do. Slowed to land on the system's own curve at 60 gpm, the pump would make 19.5 ft, so the product is 60 times 19.5, or 1170, which is 45 percent of the original 2574. That is the honest comparison because both figures are hydraulic-power products at the same reference; neither is shaft power. Shaft power differs from both because pump efficiency changes as the operating point moves, generally falling as you move left of best efficiency, so the real saving from slowing down is somewhat less than 55 percent and the real draw in the throttled case is somewhat worse than the 2 percent suggests.

The failure mode if you skip this arithmetic: a tech sees flow drop and assumes the pump is now loafing. It is not. It is working nearly as hard as before, at a point further from its best efficiency, with radial load on the shaft that rises as you move away from that point, and with 54 percent of its output being destroyed in a valve seat two feet away.

Where the throttle shows up later

• At the seat. Sustained high-velocity flow through a partly open gap wire-draws the trim. A valve carrying most of the loop's pressure drop is consumable.
• At the pump. Off-BEP operation raises radial thrust on the shaft and shortens seal and bearing life. The further left you throttle, the worse it gets.
• In the noise. Air and water both start to hiss and then roar as the gap velocity climbs. In liquid, that noise is usually the announcement of local vaporization rather than turbulence, which is a different mechanism and a different fix.
• In the neighbouring branches. Every branch shares the same pressure difference across the loop. Adding resistance to one raises that difference and pushes flow into the others.

What would make me not throttle at all

Reach for a different tool when any of these is true. Speed control, on a machine and drive rated for it, is the answer where the flow reduction is large, permanent, and on a centrifugal machine, because it moves the machine's curve instead of steepening the system's. Impeller trim is the answer where the reduction is permanent and there is no drive. Reducing the number of parallel paths or resizing the terminal is the answer where the branch is oversized rather than over-pumped. And where the pump is positive displacement, the valve is never the answer.

How to verify you got this right

Put a differential gauge across the balancing valve alone, then across the terminal alone, on the same instrument and the same reference so the two numbers are comparable. The valve drop plus the terminal drop plus the piping loss should account for the pump's differential; if it does not, you have a path you have not found, usually a bypass or a check valve that is not seating.

Then confirm flow independently of the valve position, by an in-line meter or by a heat balance across the terminal, because valve handle position is not a flow reading. On water, the sensible heat relationship is roughly 500 times gpm times the temperature difference in degrees F, where the 500 assumes plain water near 60 F; on glycol or on a hot loop the constant shifts and you go to the fluid's own property data.

Last, come back at a different load and read it again. A balance set at one condition and verified at only that condition is a guess about every other hour of the year.

References

• 29 CFR 1910.147, control of hazardous energy, for mechanical isolation and stored pressure when a valve is being serviced rather than adjusted
• 29 CFR 1910.333(b)(2) for electrical work at the starter or disconnect, and NFPA 70E-2021, 120.5 for the live-dead-live proving sequence
• Hydraulic Institute standards for pump operating-point, best-efficiency and minimum-flow terminology
• Manufacturer pump and fan curves, including the published power curve, for the machine in front of you
• See related: Reading Pump Curves Reference; How Suction Conditions Fail in Practice; How a Branch Steals From the Branch Next to It