I'm Trevor White. For a little over 11 years, I've worked on the reliability side of flow control—which means I've lived with Flowserve control valves, actuators, and positioners, plus a few aging orphan valves that should have been retired before I ever set foot on the plant floor. I've personally made and documented 13 avoidable mistakes along the way. Add them up and they're roughly $84,000 in wasted budget. That's why I write things down now.
The question newer engineers ask me, usually in a training session, is "what is the theory of drift?" They want a neat equation. I understand; I wanted one too. But in flow control, drift is not something you solve on paper. It's the slow gap between what a valve says it's doing and what it's actually doing. Because it grows in small increments, it never generates an alarm—just a control loop that slowly starts behaving like it has a bad attitude.
So this article compares two ways I've seen teams manage drift:
- Schedule-first maintenance—pull the valve on a fixed interval, calibrate it, put it back, and hope the timing lines up with reality.
- Trend-first maintenance—leave the valve in service, monitor position and command data, and intervene when the numbers say the valve has actually drifted.
I'm not going to tell you one approach is right for every plant. I am going to compare them on three things that matter long after the excitement of a new dashboard wears off: who catches the problem first, what each method really costs, and what knowledge is still available after people leave.
What the Theory of Drift Actually Means
A control valve is a promise. You send a 12 mA signal, and the valve agrees to hold the stem at a certain position. Drift is when that promise slowly breaks. The same command that used to mean "48% open" starts meaning "52% open"—or 44%, depending on the direction. At the process level, the controller has to keep pushing the valve a little further just to hold the same flow. Nothing alarms. The valve doesn't squeal, doesn't vibrate, doesn't fail. It just quietly moves away from the truth.
Calling it a "theory" makes it sound academic, but the causes are completely mundane: seat erosion, trim wear, packing friction, actuator spring fatigue, dirty instrument air, and a positioner feedback linkage working itself loose over time. All of these change the relationship between the command signal and the actual stem position. The longer a valve runs, the further the real relationship drifts from the calibrated one.
It's also not the same as deadband or stiction. Deadband shows up when a signal reverses—the valve doesn't move until the command has changed enough to overcome friction. Drift is different. Drift can appear while the command signal is completely still. That distinction matters, because it means you can't catch drift by watching setpoint changes alone. You have to watch position over time.
Two Ways I've Seen Teams Manage Drift
Schedule-first is the oldest playbook in the industry. Every 12 months, or every turnaround, you round up the critical valves, lift them out, bench-set them, put them back, and check the box. It feels disciplined. The problem is that the discipline is built around a calendar, not around the valve's actual behavior.
Trend-first uses a digital positioner or position feedback to compare the commanded position with the actual stem position over weeks and months. When the gap starts growing, you get a data point instead of a surprise. You still calibrate valves. You just calibrate the ones that need it, when they need it.
Where the Two Approaches Actually Separate
1. Detection: you can't schedule a problem that ignores schedules
In April 2019, my team pulled 21 valves during a turnaround and bench-set them one by one. Eleven of them were still inside tolerance. We spent real money on the other ten—parts, lapping, diaphragm kits, testing—and it felt like a productive week.
Two months later, one of those valves started drifting again. A 4-inch letdown valve we had repaired was slowly losing its ability to hold position. It wasn't due to be checked again for another year, and nothing in the maintenance system was going to catch it. We only found out when a console operator noticed the controller output climbing a little more every day, for no reason she could explain. That was the day I stopped believing that a bench test was a crystal ball.
Trend-first works differently. A positioner with feedback logs what the valve was asked to do and what it actually did. You can see a drift pattern forming while it's still small. In the last 18 months, my team has caught 47 potential drift issues by reviewing that kind of data. Not one of them was on a schedule. The conclusion here isn't glamorous: if you want to find drift, you have to look at drift, not at the calendar.
2. Cost: the cheaper quote is often the more expensive lesson
Late in 2022, we were replacing positioners on three critical flow control valves. Procurement found a lower-cost option without position feedback or data logging. The savings: $2,150. I argued against it. I lost that argument.
Four months later, one of those valves began drifting. Because the positioner had no diagnostics, we couldn't tell whether the problem was in the positioner, the actuator, or the feedback linkage. A technician had to go out with a handheld communicator, climb the structure, and manually exercise the valve while the process endured three separate interruptions. By the time we replaced the positioner and recalibrated the loop, we had spent about $9,800 including overtime. The $2,150 savings became a $9,800 problem. I still kick myself for not pushing back harder with the total cost numbers in front of me.
To be fair, schedule-first has hidden costs too. Pulling a healthy valve out of service, isolating it, scaffolding it, replacing gaskets, and putting it back can easily cost more than the calibration was worth. I've watched teams spend $1,500 in labor to confirm that a valve didn't need any work. The most expensive maintenance isn't always the work you do—it's the work you do on equipment that was fine.
3. What's left after people move on
There's a quieter difference between these two approaches, and it shows up when someone retires. A few years back, we lost our senior instrument tech. He was the guy who just knew which valves lied to you, which ones needed a little extra stroke, and which ones were fine even though the calibration sheet said otherwise. When he left, most of that knowledge left with him.
Schedule-first paperwork usually says "checked and adjusted." It doesn't say how the valve behaved between checks, how fast the error grew, or what changed after the last repair. Trend-first leaves a trail. You can look back at six months of position data and see exactly when a valve started to slip. That doesn't replace experience, but it does something better: it turns experience into something the next engineer can read.
There is a catch, though. Trend data only helps if someone actually reviews it. A dashboard with 200 valves and no owner is just another report nobody reads. The counterintuitive conclusion is that trend-first is more demanding, not less. It requires someone to look at the data, ask why the gap is growing, and decide whether to intervene.
What I'd Choose Today
If I had to pick a default for any control valve where drift causes product loss, downtime, or safety risk, I would choose trend-first. The whole point is to measure the problem directly instead of guessing from the calendar.
But I wouldn't put a smart positioner on every valve in the plant. For a stable, non-critical utility loop with no digital infrastructure and nobody assigned to review trends, schedule-first is still the honest answer. A 2-inch cooling water bypass that has not moved in three years does not need to be part of a predictive analytics program. It needs a reasonable interval and a competent technician.
Here's the rule of thumb I now give our engineers: if you can measure drift directly, don't manage it with a date. If you can't measure it, keep the schedule conservative and admit that you're managing blind.
A Practical Note Before You Spec Anything
One thing I've learned the hard way: the best positioner in the world won't save a mechanically worn valve. Trend-first only works when the valve and actuator are healthy enough to respond. So we still rebuild valves. We still replace worn trims. The data just tells us which ones actually deserve the attention.
If you're researching an upgrade, the Flowserve control valves product page is a useful place to start. It lets you compare valve styles, actuator options, and positioner capabilities without relying on a sales rep's memory. While you're reading, check what data the positioner can actually expose, not just what the brochure says on the front page.
And for anyone who found this article while looking for company information: the Flowserve Corporation stock symbol on the NYSE is FLS. I'm an engineer, not an investment advisor, so that's the extent of my commentary on the subject.
About the author: Trevor White is a reliability engineer focused on control valve diagnostics and repair. He has been tracking control valve drift stats for more than 300 valves across three facilities. The views here are his own and come from maintenance failures he'd rather not repeat.
Leave a reply