Most failures don't happen all at once. There's a window — often weeks or months — between the moment a problem becomes detectable and the moment the equipment actually fails. Reliability engineers call it the P-F interval. The good news for an operator is that the early part of that window usually doesn't need any instruments at all. It needs a person who knows what to look, listen, and feel for — and a habit of doing it consistently.
Three checks catch the majority of developing problems on rotating and fixed equipment.
1. Sight — what's changed since last time?
The eyes catch more than any single sensor. On a walk-past, the high-value things to look for:
- Leaks and seeps — a fresh drip under a seal or flange that wasn't there last week.
- Corrosion, scorching, or discoloration — heat marks on terminals, rust streaks, a panel that smells "electrical."
- Loose or shifted mounts — fretting dust around bolts, a guard that's started to rattle, a coupling that looks off.
- Lubrication and levels — sight glasses, oil colour, grease purged past a bearing.
- Blocked cooling — fins packed with dust, a fan shroud full of debris.
2. Sound — listen for the change in pitch
Experienced operators "hear" a plant. A healthy machine has a signature, and most mechanical problems announce themselves as a change in that sound before anything else:
- Bearings — a new whine, growl, or rumble where it used to run smooth.
- Cavitation — a pump that sounds like it's pumping gravel is starving for suction.
- Knocking or rubbing — looseness, misalignment, or contact that shouldn't be there.
- Electrical — buzzing or arcing in a panel or contactor is never "normal."
3. Temperature — and its close cousin, vibration
Heat and vibration are where developing faults show up physically. You don't need a thermal camera to start — though a cheap IR thermometer makes the check repeatable:
- Hot bearings or motor casings rising over time point to lubrication, load, or alignment problems.
- Hot electrical connections are one of the most common — and most preventable — causes of failures and fires.
- Excessive vibration, even judged by hand, flags imbalance, looseness, or misalignment early.
Never put hands on rotating equipment or energized parts to "feel" for heat. Use a non-contact IR thermometer, observe from a safe distance, and follow your site's safe-work procedures.
The signal is the change — not the number
Here's the part that gets missed: an absolute reading rarely tells you much on its own. A bearing that has always run at 60 °C is fine. The same bearing reading 75 °C this week is news. A motor that always hummed a certain way is healthy; the day the hum changes, pay attention. Early detection is about trend, and trend requires history. That's why a check you don't record is worth a fraction of one you do.
Why these checks fail in real life
If this is so cheap and effective, why don't more sites do it well? Three reasons, every time:
- Too long or too vague. A 40-point checklist nobody finishes, or "inspect pump — OK?" that means nothing.
- Inconsistent between people. Two operators check different things in different ways, so the data can't be compared or trended.
- Never captured or actioned. The observation lives in someone's head or on a clipboard that gets filed and forgotten.
Making it stick
A check program survives only if it's fast enough to actually happen and structured enough to be useful. In practice that means a round that takes seconds per asset, a consistent verdict everyone uses (OK / Watch / Issue), and a path that routes anything abnormal to maintenance with context — which asset, where, what was seen, and how it compares to last time.
This is exactly the problem we built EdgeSight to solve. An operator scans an asset's QR tag, sees the right short checklist, records the three signals in under ten seconds, and anything flagged routes to maintenance with the asset's history attached. The checks above are good practice with a clipboard — they're far more powerful when they quietly build a record for every asset over time.
