Key Points
- Some of the most expensive losses in plastics never trip an alarm. Long changeovers, micro-stops and a cooling loop running a few degrees off quietly fill the scrap bin.
- One hour of a line running off-spec can cost thousands in wasted resin alone, before missed orders and labor.
- When a vibration reading changes, the first question is always the same: did the setup change, or is the machine going? Vibration alone often can't answer it.
- Guessing wrong costs you either way. Teams chase false alarms, or they write off a real failure as "probably the changeover."
- The answer is context. When equipment condition and operating context sit in one view, an operational change and a developing failure look very different.
The Most Expensive Losses Don't Trip an Alarm
When most people picture unplanned downtime, they picture the big one. The extruder trips, the line goes quiet and the phone rings in the middle of the night.
But in plastics and packaging, a lot of lost output never makes that much noise. The line keeps running. It just runs a little worse:
- A line running off-spec. One packaging operation put a single hour of bad product at thousands of dollars in wasted resin alone. Missed orders and overtime come on top of that.
- A cooling loop that's a few degrees off. A failing chiller pump doesn't stop the line. It keeps making scrap, hour after hour. The loss is continuous instead of catastrophic, which makes it easy to miss and expensive to ignore.
- Changeovers that keep getting longer. No single one looks like a problem. Added up across a short-staffed week, they're hours of lost output.
In Tractian's conversations with more than 230 plastics and packaging operations this year, nearly half raised exactly this kind of loss. It was also the one they could put the clearest price on.
These losses hide well because they look like normal operation. And they often share a root cause with the big breakdowns: a machine that's starting to change, and a team that can't tell yet whether that change matters.
Why "Probably the Changeover" Is So Hard to Rule Out
Plastics lines change on purpose all the time. A new mold, a different resin, a speed change or a new product run can all shift how a machine vibrates, heats up and draws power. So can a bearing that's starting to wear.
Looking at vibration alone, those two stories can look a lot alike. And when condition data lives in one place and operational data lives in another, a setup change looks just like a fault.
That's where teams lose time, in one of two ways:
- They chase false alarms. Someone pulls a technician off planned work to troubleshoot a machine that's running exactly as it should for the new setup. Do that enough times and the team starts tuning alerts out.
- They write off a real problem. A developing failure shows up right around a changeover, and it gets filed as "probably the changeover." Weeks later, the machine fails, and the warning was in the data the whole time.
Both cost you. One burns hours of troubleshooting. The other ends in a breakdown that was visible weeks earlier. The fix is alerts your team can trust enough to act on, not more alerts.
Four Signs That Separate an Operational Change From a Failure
The answer is context. When you can see equipment condition next to what the machine was doing at the time, the picture gets clear fast. Here's what to look for:
No single row settles it. The confidence comes when they agree. If the vibration shift, the motor current and the cycle data all point the same way, your team can act on the alert instead of debating it.
What It Looks Like on an Extruder
Take the same extruder gearbox on two different nights.
Night one. Vibration on the gearbox starts to shift. Motor current rises. Cycle behavior drifts. Seen one at a time, any of those could be a mold change or a new resin. Seen together, with no setup change on record, they point to a developing equipment issue. Maintenance orders the part and plans the repair for the next scheduled stop, before the line goes down.
Night two. Vibration shifts right after a changeover. Current settles at a new, steady level, and the cycle matches the new run. That's a normal process change. No work order, no wasted troubleshooting and no one pulled off another job to chase it.
Same machine, same first signal, two very different answers. The difference is whether your team can see the whole story at once.
How to Put Operating Context to Work on Your Floor
You don't need a full overhaul to start reading alerts this way. Begin with one line and a few habits:
- Pick the line where hidden losses hurt most. Look for the line with the most scrap, the longest changeovers or the most "we're not sure why" slowdowns. Extrusion and molding lines are common starting points.
- Watch the machines that set the pace. Extruder drives and gearboxes, molding machine pumps and motors, chillers and air compressors are where condition changes matter most.
- Add electrical and operational data where you can. Motor current, power, pressure, and flow turn a vibration reading into a story with a timeline.
- Make changeovers visible. If the team can see when a new run or speed change started, they can line it up against any shift in the data.
- Agree on who acts, and how fast. Name an owner for alerts and decide what happens when one looks like a real failure versus a process change.
- Close the loop. After every alert, note whether it was a process change or a real catch. That record is what builds trust, and it shows you where to expand next.
Where Tractian Fits
Tractian brings three kinds of data into one view of asset health:
- Mechanical: vibration, ultrasound, magnetic field and temperature from wireless sensors on the machine
- Electrical: power, voltage, and motor and servo current, showing how hard the machine is working
- Operational: pressure, flow, level and discrete pulses, showing what the machine was asked to do
On their own, each signal leaves a question open. Together, they answer it. Mechanical monitoring tells you whether a machine is going to break. Adding electrical and operational context tells you why a line is giving up output, and what to do about it. That's the difference between keeping one machine alive and keeping the whole line moving. And that’s what we like to call multimodal monitoring.
The results are already adding up. In the last 12 months, packaging and materials plants running Tractian identified 1,127 failures before they stopped production, avoided 2,545 hours of downtime and saved $1.82 million across more than 60 plants. That's 2,545 hours that didn't turn into purging, startup scrap, overtime or an order moved to a sister plant.
Your machines are already telling you the difference between a changeover and a failure. The question is whether your team can see it in time to plan the fix.
Results source: Tractian customer results across plastics, packaging, pulp and paper, and glass plants, trailing 12 months.
Frequently Asked Questions
Can a changeover cause a change in vibration readings? Yes. A new mold, a different resin, a speed change or a new product run can all change how a machine vibrates and how much current it draws. That's why vibration alone can't always tell a process change from a developing failure.
How do you tell a process change from an equipment problem? Look at timing, trend, motor current and cycle behavior together. A process change usually lines up with a setup change and settles at a new, steady level. A developing failure appears without a setup change and keeps getting worse over days or weeks.
What are hidden losses in plastics manufacturing? They're losses that don't stop the line: running off-spec, micro-stops, slow cycles, long changeovers and cooling that drifts out of range. Because the line keeps running, they often go unmeasured, even though they can cost as much as a breakdown.
What is multimodal monitoring? Multimodal monitoring combines mechanical data (like vibration and temperature), electrical data (like motor current and power) and operational data (like pressure and flow) into one unified view of asset health. Together, they explain not just whether a machine will fail, but why a line is losing performance.
Will this create more alerts for my team? The goal is fewer, better alerts. When condition and operating context are read together, process changes stop showing up as faults, so the alerts that remain are the ones worth acting on.
Which equipment should we monitor first? Start with the machines that set line speed or stop the line when they fail: extruder drives and gearboxes, injection molding pumps and motors, chillers, air compressors and blowers.