6 Benefits of Power Quality Monitoring for PLC, CNC, MCSA

Key Points

• The electrical supply interrupts control hardware, wears motors and drives, and distorts the current spectrum diagnostics read.
• Voltage unbalance produces current unbalance several times larger, concentrated in one winding, predictably shortening insulation life.
• A power quality study characterizes only the weeks it covers, and intermittent events usually fall outside it.
• Power quality events originate on both sides of the meter, so responsibility requires per-asset measurement to assign.

The Log Entry That Says Nothing

Some entries in a maintenance log say nothing. A line stops, the controller restarts cleanly, the tech finds nothing wrong with the machine, and they close the event as an unknown fault. With nothing to report, any amount of time can go by, then a nearly identical entry appears under a different asset number.

You could see these entries two different ways. Maybe the team is being careless. But likely, that’s not the case. These kinds of unknown fault entries are often evidence that a measurement is absent. That the sensing ability to capture the full range of data isn’t there, and something outside of current capabilities is occurring. Especially when it’s random non-mechanical failure.

As you know, electricity is not a constant, and it doesn’t act on a plant in only one way. It interrupts control hardware, it wears motors and drives, and it distorts the current spectrum that electrical diagnostics depend on, all at the same time. A plant that is not measuring the electrical supply can’t separate those three effects, and their consequences will be recorded under other names.

This article covers six things continuous power quality monitoring changes for programmable logic controllers, CNC equipment, and motor current signature analysis. Each one is a capability a plant either has or does not have, all known by what your team can or can’t prove after a stop occurs.

One Supply, Three Ways to Fail

Every plant treats electricity as an input. But many don’t view it through its real impact as a variable that acts on equipment.

For a programmable logic controller and the control hardware around it, the supply is something that interrupts. A disturbance far too brief to register as an outage can still drop a controller mid-sequence and leave nothing behind except a stopped line.

For motors, spindle drives, and the servo amplifiers inside a CNC machine, the supply is something that wears. That damage accumulates rather than announcing itself, and by the time it surfaces, it has usually stopped being a repair.

For motor current signature analysis, the supply sits inside the measurement. The technique reads the current spectrum to find what is happening inside a motor, and supply-side distortion lives in that same spectrum.

Those are three different mechanisms operating on the same plant at the same time. The supply is disrupting the controller, wearing the asset, and contaminating the measurement simultaneously, and nothing separates the three without measuring what the supply is actually doing.

Most plants do not measure it, and the reason is not carelessness. Voltage disturbances happen faster than anyone can observe them, and they are not included in the reliability statistics utilities report. A plant can experience them week after week for years and hold no record that they ever occurred.

When the Supply Interrupts the Controller

The hard problem with these events is that they happen without leaving evidence behind, which is why they keep happening.

1. Stops have a cause instead of being unexplained resets

A voltage sag is a short drop in voltage, lasting anywhere from half a cycle to a minute, that never falls all the way to zero.

It’s a familiar pattern for most plants. The line stops, and the controller comes back clean. Maintenance tests the machine and doesn’t find anything. The event closes in the log as an unknown fault. Six weeks later, a nearly identical entry appears under a different asset number. You keep getting random unknown faults. 

Nobody can predict which controller will drop, because sag tolerance is not something a specification sheet reports. Published testing across typical industrial controllers found that the knee point of a controller's voltage tolerance curve ranges from roughly 35 to 85 percent of nominal voltage, with knee durations from zero to five cycles. Two controllers of similar rating, mounted on the same panel, can behave completely differently in the same event.

What follows is disproportionate to what caused it. Business losses track production downtime rather than the length of the disturbance, so a partial voltage loss can cost as much as a full hour without power once equipment has to be rebooted and a process restarted.

On a machining center, the same event looks different and costs more. When a drive faults mid-cycle, the part sitting in the fixture becomes scrap, and the spindle has to be re-referenced before anything runs again.

None of it gets prioritized, because none of it gets attributed. Unplanned downtime that closes without a cause is downtime nobody can build a case around, and it stays in the log as a category rather than as a diagnosed equipment failure.

2. Causes are located, whether inside or outside the fence

Power quality events originate on both sides of the customer meter. That single distinction decides who owns the fix, and most plants settle it by inference.

Inside the fence, the usual sources are:

• Unevenly distributed single-phase loads
• Large motors starting on shared buses
• Loosened terminations
• Switching activity that wasn’t logged

Outside the fence, the events belong to the supply and arrive without warning.

From inside a control panel, the two are indistinguishable. Both produce the same stopped line and the same unknown fault entry.

Per-asset measurement timestamped from the first symptom rather than from the moment a threshold was crossed changes what conversation is possible afterward. Therefore, a plant holding that record can take it to a supplier. It can also discover the problem was its own, which is less satisfying and considerably more actionable.

Either way, root cause analysis on an electrical event is only as good as the timeline underneath it, and inference is not a timeline.

When the Supply Wears the Asset

Nearly every plant is paying this cause, and they’re doing so on a schedule. But it never appears as a line item, which is exactly why it keeps happening.

3. Stop invisibly spending motor and drive life

Voltage unbalance means the three line-to-line voltages feeding a motor are not equal. It is the most expensive small number in an industrial facility.

The reason is that the asymmetry doesn’t stay small. Department of Energy guidance for manufacturers reports that current unbalance can run 6 to 10 times the size of the voltage unbalance, and gives a measured case in which 2.5 percent voltage unbalance produced 27.7 percent current unbalance on a 100 horsepower motor.

That extra current is not shared evenly across the machine. It concentrates in the phase with the lowest voltage, so one winding runs hot while the other two look entirely normal to anyone taking a reading.

Heat then does its work on a schedule. The same guidance states that winding insulation life is cut in half for every additional 10 degrees Celsius of operating temperature, and that unbalance above 1 percent requires derating under NEMA MG-1.

Unfortunately, this also voids most manufacturers' warranties.

This reaches further than a motor list suggests. Assets across the plant absorb this. We’re talking about induction motors, VFD-driven equipment, and inside a precision machining center, where the spindle and servo drives absorb it first. There, the early signal is rarely a failure. It is thermal foldback, a drive quietly derating itself to survive conditions nobody has characterized, which reads on the floor as a machine that has become slower and less consistent for no obvious reason.

The invisibility is the crux of the problem, if not the problem itself depending on perspective. The motor eventually fails, the failure gets recorded as a motor failure, and the supply condition that aged it never enters the record at all.

Programs built on thermal monitoring see the elevated temperature and still miss the reason for it, because temperature is the symptom of this particular failure mode rather than its cause.

4. Electrical stress becomes attributable to a named asset

An aggregate utility bill and a plant-level power quality report share one weakness. Neither of them names any particular machine.

Unattributable numbers do not survive budget reviews, because the first question is always which asset and how much. A finding that applies to the facility as a whole without any differentiation can’t provide any argumentation, and it competes badly against requests that arrive with an equipment tag attached.

Per-asset measurement turns the same information into a cost with a name on it. That is the difference between raising a concern and making a request.

Three things carry weight upward.

• Unplanned stops that stop closing without a cause, which is where the cost of downtime actually accumulates
• Asset life recovered on critical assets currently being derated by a supply condition nobody has measured
• Engineering hours that stop being spent on ambiguity

However, the following is noteworthy because it runs against the way power quality is usually sold. Consumption savings is the weakest part of the entire argument here. Energy management is a legitimate benefit, but it is not the one a reliability case should rest on. This is because the bottom line lives with the asset, and in the downtime, rather than the meter.

When the Supply Contaminates the Measurement

The cost here isn’t money, like the first two, but certainty. And at the decision stage, it is the more expensive of the two.

5. Motor current signature analysis becomes defensible

Motor current signature analysis (MCSA) is well established, and most reliability teams already know what it can find. The constraint with it, though, has never been the technique.

Harmonic distortion means current and voltage components appearing at multiples of the fundamental frequency. Those components occupy the same spectrum MCSA reads to find rotor and stator conditions. That is a problem of arithmetic rather than of instrument quality, and no amount of care with the instrument resolves it.

Published reviews of the method state it directly. Spectral MCSA approaches degrade under voltage unbalance or harmonic distortion, producing false detections and misinterpretations. The same body of work sets out a precondition most programs never verify, which is that steady-state current methods assume the machine runs at a constant known speed with a constant supply frequency. On VFD-driven equipment, and on a machining center changing spindle speed through a cycle, neither assumption holds on its own.

The traffic runs both directions, and this is the part worth sitting with. Federal guidance on voltage unbalance advises that a finding of 120 hertz vibration should prompt an immediate check of voltage balance. A vibration program is already picking up electrical problems. It is simply reading them as mechanical ones and routing them accordingly.

What this costs is not accuracy. It is defensibility. A condition monitoring program produces findings that somebody eventually has to justify, and a reliability engineer who cannot say what the supply was doing at the moment of measurement cannot fully assign the finding to the machine.

Automated fault assessment shortens that loop considerably, but only when the electrical reference underneath it is sound. Hesitation is expensive, and hesitation is the specific thing predictive maintenance was supposed to remove.

6. The diagnostic reference is no longer limited to a two-week snapshot

Power quality is usually encountered as an investigation. Something recurs, an analyzer goes onto a panel for a week or two, a report comes back, and the plant returns to normal operation.

A measurement characterizes the window it covers. That is not a criticism of the instrument or of the engineer holding it. It is arithmetic, and it holds regardless of how good either one is.

Two consequences follow from this. The events that justified the investigation are intermittent by definition, which makes them the least likely to occur inside a fourteen-day window. And a reading taken at a panel cannot be attached to the individual machine that was absorbing the stress.

The second consequence compounds in a way the first does not. Every later diagnosis inherits its reference. If the electrical baseline is two weeks old and was captured at the wrong point in the distribution, the context for every subsequent finding is borrowed rather than measured, including the current-spectrum findings from the previous section.

There is a number worth finding out. How many hours of the last year does the plant actually hold power quality data for, and how does that figure compare against its count of unexplained stops over the same period?

Machine condition monitoring programs answered this question on the mechanical side years ago, which is why programs built around equipment that runs intermittently had to abandon scheduled sampling to work at all.

What This Changes About Decisions

Three questions are now answerable, and the answers are actionable.

• How many of last quarter's unplanned stops on controlled equipment closed with a recorded cause rather than a reset
• What the supply was doing during the last motor diagnosis that turned out to be wrong
• Whether anyone can name the plant's three worst assets by electrical stress without going to look

Each of those requires continuous, per-asset measurement. Nothing else produces an answer.

How Tractian Approaches Electrical Monitoring

The constraint that has kept continuous electrical measurement out of most plants is not cost, and it is not analytics. It is installation.

Tractian built its electrical monitoring layer around this constraint.

The electrical monitoring sensor clips onto existing conductors rather than being wired into the circuit. No rewiring, no shutdown negotiated with production, no electrical work order standing between a plant and its first reading. That's why per-asset coverage becomes practical across a fleet instead of one panel at a time.

What it captures runs continuously on every phase. Current and voltage per phase, power factor, phase angle, and harmonics, around the clock rather than during a scheduled check. From the first week, each asset holds a baseline of how it actually behaves electrically, and alarms fire on deviation from that asset's own established pattern rather than on a threshold drawn without reference to the machine.

Electrical Signature Analysis (ESA) reads the same waveforms for signatures inside the machine, mapping rotor, stator, winding, and insulation conditions to named failure modes that carry a severity and a progression state.

The part that answers this article most directly is what happens when the electrical and mechanical layers are one system rather than two. Vibration, ultrasound, temperature, and speed from the condition monitoring side land on the same asset records, the same timeline, and the same fault classification layer as the electrical data. Separating a supply event from a machine fault becomes a property of how the diagnosis was assembled rather than a reconstruction someone performs weeks later across two reports. The 120 hertz problem gets answered structurally instead of by whoever happens to remember the connection.

Learn more about Tractian's electrical monitoring for critical assets to see how high-quality, decision-grade IoT data transforms your program into AI-powered closed-loop workflows.

FAQs about Power Quality Monitoring

Is power quality monitoring worth adding if we already run vibration monitoring?

Yes, because the two cover different domains. Vibration shows the mechanical consequence. Electrical measurement often shows the cause, since supply conditions, drive behavior, and internal motor faults all sit upstream of the symptom vibration eventually detects.

Do we need a power quality study or continuous power quality monitoring?

A study characterizes the weeks it covers. Continuous monitoring captures intermittent events, which are usually both the expensive ones and the least likely to occur while a temporary analyzer happens to be connected. Most plants need the second.

Can power quality monitoring tell us whether a problem is coming from the utility or from inside our plant?

It can, provided the measurement is continuous and per-asset. Events timestamped from the first symptom show whether a disturbance arrived at the service entrance or originated on a specific internal bus.

Does power quality monitoring actually help with PLC and CNC nuisance trips?

It gives those trips a cause. Sag events are recorded with magnitude, duration, and timestamp, so a stop that previously closed as an unknown fault can be matched to a measured disturbance and prioritized against everything else.

How does power quality data improve motor current signature analysis?

It supplies the reference the technique assumes. Knowing the supply condition at the moment of measurement lets a finding in the current spectrum be assigned to the machine rather than to distortion arriving from the network.

Can power quality monitoring be installed without shutting down the line?

Non-invasive sensors clip onto existing conductors instead of being inserted into the circuit, so no shutdown, rewiring, or process interruption is required.