Estimated reading time: 9 minutes
Key takeaways
- Hidden variables in predictive maintenance. These are the keys to understanding differences in asset performance.
- Context often beats another sensor for explaining failures. Capture what changed around the asset, not just what the asset did.
- Start with one asset class, then scale what works. The goal is repeatability, not perfection.
- Build a simple signal map and stop guessing. Make it a shared artefact between maintenance and operations.
You know the story.
Two pumps. Same make, same duty, same PM plan, same spares.
Yet one runs for months and the other keeps eating bearings. If that sounds familiar, the problem is rarely a “bad batch” or a cursed line. More often, your predictive maintenance or condition-based monitoring data is missing the hidden variables that explain what the pump was experiencing, not just what the pump was doing.
Process context (recipe changes, starts per hour, suction conditions), lubrication detail, electrical stress from VSDs, and the local environment can quietly shift an “identical” asset into a completely different failure pattern.
See where Predict will make a difference on your site
Condition-Based Monitoring (CBM) is revolutionising the manufacturing industry, and AssetMinder Predict is your key to unlocking its full potential. With advanced sensors, real-time data analytics, and machine learning, AssetMinder Predict offers proactive maintenance, enhanced efficiency, cost savings, and improved safety. It continuously monitors your machinery, providing early warnings and actionable insights to prevent failures before they occur, optimise maintenance schedules, and reduce unnecessary interventions.
Ready to transform your maintenance strategy? Visit AssetMinder Predict to see how this innovative solution can revolutionise your operations and arrange a demo today!
Don’t wait for the next failure—predict and prevent it with AssetMinder Predict. Embrace the future of maintenance and discover unparalleled reliability and efficiency today.
See where Predict will make a difference on your site
Condition-Based Monitoring (CBM) is revolutionising the manufacturing industry, and AssetMinder Predict is your key to unlocking its full potential. With advanced sensors, real-time data analytics, and machine learning, AssetMinder Predict offers proactive maintenance, enhanced efficiency, cost savings, and improved safety. It continuously monitors your machinery, providing early warnings and actionable insights to prevent failures before they occur, optimise maintenance schedules, and reduce unnecessary interventions.
Ready to transform your maintenance strategy? Visit AssetMinder Predict to see how this innovative solution can revolutionise your operations and arrange a demo today!
Don’t wait for the next failure—predict and prevent it with AssetMinder Predict. Embrace the future of maintenance and discover unparalleled reliability and efficiency today.
The same pump story, same plan, different failure rate
Let’s make it concrete.
Imagine a standard centrifugal pump on Line A and Line B. Both are “identical” on paper. You have a standard lubrication route. A PM every four weeks. A decent technician who can strip and rebuild with their eyes shut. Yet Line B keeps failing.
When you finally slow down enough to compare the two properly, you find the differences were never in the pump. They were in the context:
- Line B runs a slightly different product mix. More solids. More temperature swings. More start-stop.
- The suction conditions are marginal on Line B during certain recipes. You get cavitation events that are short and brutal.
- The motor on Line B sits closer to a VSD cabinet and sees more electrical noise and heat.
- Line B’s operators “help” by cracking valves manually during changeovers to hit throughput targets, which shifts the pump into the wrong part of its curve.
All of this is real. None of it is in your CMMS. So every investigation starts from a false assumption: “Same pump, same conditions.” That assumption is what creates repeat fixes.
Hidden variables: the four buckets that quietly drive failure
If you want fewer repeat breakdowns, you do not need to track everything. You just need to get disciplined about capturing the small set of context signals that explain why identical assets diverge.
Here are four buckets that show up again and again in rotating equipment failures.
1) Environment
The stuff the machine “breathes”
Ambient temperature, humidity, washdown routines, dust, corrosive vapours, mounting stiffness. These affect bearing life, lubrication breakdown, electrical reliability and sensor readings. Even small changes matter when you run 24/7.
2) Lubrication
Not just “greased”, but “greased how”
Grease type, fill quantity, interval, contamination risk, who did it, and what they saw while doing it. Over-greasing and mixed greases are classics. So is a small seal defect that lets moisture in, which then turns into a bearing failure you blame on load.
3) Electrical
The invisible stress
VSD settings, harmonics, cable routing, grounding, motor current, and power quality issues. Condition-based monitoring is not only mechanical. Current sensors, temperature sensors, and vibration sensors all have a role, and the choice depends on what you are trying to catch. In IoT-based CBM systems, vibration sensors are widely used alongside temperature, pressure, current and oil analysis sensors, because different faults present in different ways.
4) Process context
What the asset was actually asked to do
Flow, pressure, product viscosity, solids content, duty cycle, starts per hour, recipe changes, valve positions, and upstream or downstream constraints. Most “mystery failures” stop being mysterious the moment you overlay operating context onto condition data.
Why this matters (and why it is not just a maintenance problem)
When the quiet drivers are missing, maintenance is forced into guesswork. The result is repeat work, wasted parts, and a creeping loss of trust between functions.
When the quiet drivers are captured, three good things happen quickly:
- Maintenance gets better root cause and fewer repeat fixes.
You stop replacing the same component and start addressing the conditions that are killing it. - Operations gets fewer recurring stoppages and less quality drift.
In some environments, the cost of downtime is brutally visible. One AssetMinder Predict deployment referenced £15,000 per hour of downtime, which is exactly why preventing repeat failures matters more than perfect reporting. - Sustainability sees waste that shows up as consumption.
A struggling machine often consumes more energy to deliver the same output. And sometimes the waste is not subtle. In one energy monitoring deployment, unnecessary weekend energy consumption was identified and cut, delivering around £70,000 in annual savings, with a wider programme projecting £2.16 million of potential energy savings.
So yes, this is a reliability topic. It is also an operations stability topic and, increasingly, an energy and carbon topic.
“Do we need more sensors?”
Not always
There is a trap teams fall into: if we cannot explain failures, we assume we need more instrumentation.
Sometimes you do. But often, context beats another sensor.
One reason predictive programmes stall is that plants collect health data but fail to capture what the machine was experiencing at the time. That is why you can have “normal” vibration trends right up until the day the pump dies, because the failure was driven by short-lived process events that never got logged.
This is also why a lot of companies still feel like “predictive maintenance” does not work for them. Plenty of firms experience unplanned downtime, with machine failure as a major contributor, while predictive adoption remains limited.
The way out is not to boil the ocean. It is to pick one asset class and build a simple, repeatable way to capture the few context signals that explain most of the variance.
How to capture hidden variables without boiling the ocean
Here is a practical approach that works even when you are short-staffed and juggling reactive work.
Step 1: Choose one asset class, not the whole factory
Start with one: pumps, gearboxes, fans, compressors, conveyors. Pick the class that creates the most pain, or the most repeated work orders.
Step 2: Agree the “signal map” as a joint maintenance and operations artefact
This matters. If maintenance owns it alone, it becomes another form to fill in. If operations helps define it, it becomes a shared language.
Step 3: Capture context only when it changes, not constantly
You do not need a new logging burden. Most context is “steady state” until something changes. So log changes:
- recipe change
- setpoint shift
- washdown event
- changeover
- abnormal run mode
- known upstream issue
- operator intervention
Step 4: Make it easy to record in the moment
If it takes three screens and a password reset, nobody will do it. The best context capture is quick and “close to the work”. A simple dropdown list in the CMMS, a QR code at the asset, or a short “event tag” that can be added to a work order.
Step 5: Review weekly, briefly, and only on exceptions
Do not build a new meeting culture. Review exceptions: repeat failures, rising trends, energy anomalies, or top downtime contributors.
A rotating asset signal map you can actually use
Below is an example signal map for pumps. It is intentionally small. You can adapt it to your plant.
| Signal type | What to capture first | Why it helps | Who can log it |
| Condition | Vibration trend, temperature, run hours | Early warning and confirmation | Maintenance |
| Electrical | Motor current trend, VSD fault codes | Detect electrical stress, load changes | Electrical or maintenance |
| Lubrication | Grease type, interval, contamination notes | Prevent “maintenance-induced” failure | Maintenance |
| Process context | Recipe, viscosity band, starts per hour, suction pressure anomalies | Explains why “identical” pumps diverge | Operations |
| Environment | Washdown events, ambient heat spikes, dust ingress | Explains accelerated wear | Ops or maintenance |
If you already have vibration sensors, great. Use them. Just remember that vibration is one view of the machine. IoT-based CBM typically uses a mix of sensors such as vibration, temperature, pressure and current, because different failure modes speak different “languages”.
What this looks like in practice: one week, one repeat issue
Let’s say Pump B has a rising bearing defect signature.
Without context, you schedule a bearing change and hope.
With context, you overlay the last two weeks of operating conditions and see:
- defect growth accelerates only on Product X
- Product X runs only on nights
- nights also coincide with higher starts per hour due to upstream batching
- suction pressure dips during that batch cycle
Now your “fix” is not just a bearing. It is a small process adjustment, or a control tweak, or a line-side habit change. Maintenance still does the work, but operations helps remove the condition that causes the work to repeat.
That is how you reduce failure rates across lines and sites, without doubling headcount.
Value to the people doing the work
If you are a maintenance engineer or technician, this approach gives you three things that matter day to day:
- More confident diagnosis with fewer false assumptions. You stop starting every job from “unknown unknowns”.
- Practical guidance on what context is worth logging first. No extra admin for the sake of it.
- A shared language between maintenance and operations when things go wrong. The conversation moves from blame to evidence.
And if you are an operations manager, it reduces the steady drip of recurring stoppages and quality drift that never quite makes it into the headline KPIs.
