Predictive maintenance has a simple premise: most failures of rotating machinery announce themselves before they happen. A bearing that is beginning to fail vibrates differently months before it seizes; an imbalance or a misalignment shows up as a change in the signature long before it becomes a repair. The programme that reads those changes converts an unplanned stoppage into a planned one.
The value is not the measurement; it is the notice. A failure detected weeks ahead can be scheduled into a stop, the part ordered, the labour planned, and the production sequenced around it. The same failure detected on the day it becomes audible stops the line, at the worst moment, with no part on the shelf.
That is why the case for condition monitoring is a case about downtime, not about the instrument. The technology only pays where an unplanned stop is expensive - on a critical machine, in a continuous process, or where the repair lead time is long.
A vibration reading on its own says very little. What makes it useful is a baseline: a reading taken when the machine was known to be in good condition, at a defined point, under a defined load. Every later reading is compared with that, and it is the change, not the absolute number, that carries the information.
This is why the baseline should be captured at commissioning, when the machine is new or freshly overhauled and the reading can be trusted. A machine that has run for years without a baseline can still be monitored, but it takes longer to establish what is normal for it, and in the meantime the programme is reading trends without a reference point.
Once a baseline exists, three things are watched: the level, the trend over time, and the rate at which the trend moves. A slow rise over a year may be normal wear; the same total rise over a fortnight is a fault developing quickly. The rate of change is often the earliest warning.
| Signal | What it often indicates | The right response |
|---|---|---|
| Steady overall level | Stable condition | Continue routine monitoring |
| Gradual rise | Normal wear, or a slow deterioration | Trend it; plan for the next stop |
| Sudden step change | An event - a hit, a loose component, a lubrication loss | Investigate now |
| Rising with speed only | Imbalance or a resonance | Check balance and support |
| Rising with load only | Misalignment or a gear/teeth problem | Check alignment and meshing |
Alarm bands set on absolute values alone will either miss a fast-developing fault or cry wolf. Bands set on both level and rate of change match the way failures actually develop.
Vibration tells you that something has changed; it does not always tell you what changed or when it will fail. A rising reading says a component is deteriorating, but the remaining life is a judgement, not a formula, and a machine can be monitored perfectly and still fail from a cause the instrument does not measure - a blocked lubrication line, a cracked bracket, an electrical fault.
The honest position is that condition monitoring narrows the uncertainty rather than removing it. It moves the odds in favour of planning, and it should sit alongside the ordinary disciplines - lubrication, alignment, inspection - rather than replace them.
Detecting a fault weeks early is only useful if a replacement can be on site in time and someone can do the work. The programme therefore depends on two things outside the instrument: the spares position and the maintenance skill.
That makes the spares decision part of the maintenance strategy. Holding a spare bearing for a critical machine is expensive; waiting for one to be ordered after the fault is detected may consume the entire warning the monitoring bought. Where a site would rather not hold and despatch those parts itself, the movement of critical spares can be handled by a specialist operation - a partner such as Dropioneer, which runs warehousing and fulfilment for industrial and e-commerce stock, can hold a defined critical-spares list and despatch against agreed service levels, so the lead time the monitoring needs is actually available.
A full condition-monitoring system across a whole plant is rarely the right first step. A better start is to pick the handful of machines whose failure would stop production, take a baseline on each, and monitor those on a simple schedule. The results usually show quickly which machines justify closer attention and which were never the risk.
The programme can then grow on evidence. Adding a machine because it stopped the line last quarter is a defensible decision; adding one because the technology is available is not.
Not to start. A route-based programme using a portable instrument on a few critical machines establishes the baseline and the trend, at a fraction of the cost of permanent monitoring. Permanent sensors become worthwhile when the machine is critical enough, or hard enough to reach, that route-based data cannot be relied on.
Often enough to catch the fastest credible failure for that machine. If a bearing can go from healthy to failed in a week, a monthly reading will miss it. Where the failure mode is slow, monthly or quarterly is adequate. The interval is set by the machine, not by the instrument.
No. It complements it. Scheduled tasks such as lubrication, cleaning, torque checks and filter changes address causes that a vibration reading cannot see. Condition monitoring tells you when something is changing; the scheduled tasks are what stop it changing in the first place.
Reading a machine before it fails is one of the most direct ways to protect production, and it starts with a baseline and one or two critical machines rather than with a system. ZH Industrial Machinery supplies and supports industrial machinery and handling equipment, and we are happy to help a site decide which machines are worth monitoring and what their baseline should be.
The method sits inside condition monitoring, and the vibration signature it reads is the subject of vibration analysis. The way failure modes are enumerated before a monitoring plan is built is failure mode and effects analysis, and the wider economics of maintenance and distribution equipment is covered by MIT Center for Transportation and Logistics research.