Drift Costs More Than Calibration: What Is the Theory of Drift? A Sweco Service Engineer's Field Notes

If you are here because you searched 'what is the theory of drift', here's the short version: drift is the slow, quiet gap between what a machine was doing when it was set up and what it is doing now. The more useful answer is less poetic: drift is why corrective maintenance bills exist. I've handled service orders for Sweco's vibratory screens, separators, finishing machines, and oilfield wellhead equipment for seven years. I've personally made—and documented—23 significant mistakes, totaling roughly $60,000 in wasted budget. The cheapest fix to most of those mistakes was measurement, not new parts.

Among Sweco medarbetare (a Swedish term that we use for Sweco employees), the running joke is that old equipment has a personality. It doesn't. It has drift.

What is the theory of drift?

In engineering terms, drift is a gradual deviation of a process variable from its setpoint over time. It is not a sudden failure, and it is not a one-time glitch. It is a shift that develops slowly enough that the people closest to the machine stop noticing it.

On a vibratory separator, drift shows up as an orbit radius that changes by half a millimeter, screen tension that drops, or motor amperage that creeps upward. On wellhead equipment, it shows up as torque or stroke settings that move after a seal replacement. None of those things 'break' on the spot. They just move away from the original setup. The further they move, the more expensive the correction gets.

The counterintuitive part is this: most people assume that a machine that still runs is fine. In my experience, a machine that still runs is just a machine whose drift has not been measured yet. I learned that the hard way on a $3,200 order of replacement screens where every screen had mismatched tension. It looked fine on paper. It didn't run fine in practice.

Where drift hides in Sweco equipment

After enough service calls, I started looking for drift in the same places:

• Screen deck bolts. They loosen after run-in, and the tension changes. The screen still works, but the cut point changes.
• Vibration amplitude. The eccentric weight setting is right at the factory, then shifts during transport or after a motor swap.
• Bearing temperature. Everyone watches the alarm threshold. Nobody watches the trend. A bearing that runs 8 degrees hotter every month is drifting, even if it is still under the alarm limit.
• Finishing machine media load. Media wears down, process time stretches, and the operator adjusts the timer instead of the load. That adjustment masks drift until the part quality falls off spec.

These are not catastrophic failures. They are slow, measurable changes. The reason they matter is compounding: a small drift in machine settings produces a larger drift in product consistency, which finally becomes a customer complaint.

The peanut butter test

I use a food product to explain this to new hires: peanut butter. Take a jar and push it through a sieve. The first pass is thick, slow, and messy. Warm the peanut butter slightly and it moves differently. Add moisture and it moves differently. The peanut butter didn't change its identity. The conditions changed. That is what happens on many screening and separation calls when a customer says, 'We process the same product we always have.' The product might be identical, but the machine's orbit has drifted because nobody checked the eccentric weight setup, the motor speed, or the screen tension. The material is still peanut butter; the machine is what changed.

I once watched a customer blame a brand-new screen for poor throughput. The screen was fine. The separator had drifted into a lower amplitude setting after a motor service, and the operator had gotten used to the new sound. By the time they called us, the machine had been running at roughly 60% of its original separation efficiency for almost three weeks.

Miranda, the wellhead valve, and the missing torque check

One incident that changed my checklist involved a client engineer named Miranda. In September 2022, she called about a wellhead gate valve that would not seal during a hydrostatic test. It had passed the same test two weeks earlier. The initial instinct was to blame the packing or the valve body. But when we reviewed the records, the valve's torque spec had drifted after a technician replaced a worn seal and used a 'close enough' setting. Nobody intentionally changed the procedure. It just shifted.

That's the theory of drift in one sentence: every system drifts toward error unless a person, a control system, or a measurement catches it.

Miranda's team spent about $7,800 in labor and downtime to re-certify a small batch of valves. The root cause was a missing torque check after a minor repair. I now write 're-verify after ANY disassembly' at the top of every service checklist. That one line has caught more problems than any other rule we added.

The real cost of drift: an emergency in February 2024

Drift also affects how you buy service and spare parts. I used to think rush fees were just suppliers taking advantage of an emergency. My view changed in February 2024, when a client in Pennsylvania needed a replacement screen deck within 48 hours. Standard lead time was close to five business days. We quoted a rush premium that included priority shop time and expedited freight—roughly $400 more than the normal job. The client signed immediately.

The alternative was a production line sitting idle for three days. Their downtime cost was about $15,000 per day. The rush premium bought certainty, not just speed. That is exactly what the 'time certainty' argument is about: in an emergency, a 'probably okay' promise is the most expensive option there is.

I'm not saying rush premiums are always worth it. For a non-urgent spare part order, paying double for speed is waste. But when the cost of missing a deadline is bigger than the cost of arriving guaranteed, the choice is simple.

What Sweco Infra & Rail Oy taught me about measuring drift

This way of thinking is not unique to my equipment line. A former colleague moved to Sweco Infra & Rail Oy—the Finnish infrastructure and rail engineering business—and told me how their survey teams deal with rail alignment drift. They don't wait for a train to feel rough. They monitor millimeter-level changes and decide whether each measurement is noise or actual movement. They compare the measurement to the baseline, not just to the tolerance limit.

The same logic applies to separators, finishing machines, and wellhead equipment. If you only react when product quality falls, you've already paid for the drift. If you monitor the trend, you can correct it while it's cheap.

A simple checklist

I now maintain the team checklist that I wish I'd had in my first year. It has five lines:

1. Write down the baseline: amplitude, RPM, torque values, screen tension.
2. Measure at scheduled intervals. Do not rely on 'it sounds fine.'
3. Re-verify after every disassembly, repair, or screen change.
4. Track trends, not just pass/fail readings.
5. If a critical deadline depends on this machine, buy certainty in advance.

Since we put this in place 18 months ago, the checklist has caught 47 potential errors. Almost all were caught because someone measured a value that had drifted—not because something broke.

Where the theory stops

I'll add the boundary condition. Drift is a useful lens, but it's not a universal law. If a machine is catastrophically overloaded, drift is not your problem. If a component was wrong from the day of installation, that's not drift, that's procurement. And in a true breakdown emergency, you don't need a checklist first; you need a contingency plan and a supplier who can deliver quickly. That's why we keep a small stock of critical parts: to buy time when drift finally turns into a failure despite our best monitoring.

Bottom line: drift is going to happen. The question isn't whether you'll pay for it. It's whether you'll pay a little for measurement now or a lot for failure later.