Condition-based maintenance (CBM) is maintenance you perform based on the actual condition of an asset rather than a fixed calendar. You measure one or more indicators that say something about a machine's health, and you act only when those indicators show that something is changing. As long as the condition is good, you keep the machine running. The principle is straightforward: an asset that is demonstrably healthy does not need to be opened up or replaced as a precaution.
What sets CBM apart from other strategies
Maintenance strategies form a spectrum. At one end sits corrective maintenance: you repair only after something has failed. At the other end sits preventive maintenance on fixed intervals: every so many running hours or months you replace a component regardless of its real state. Both have a weakness. Corrective maintenance accepts the failure and its consequences; interval-based preventive maintenance replaces parts that are often still perfectly fine, while missing the failures that arise outside the schedule.
CBM sits in between. You replace based on measured behaviour, not on elapsed time. That places CBM close to predictive maintenance, but the two are not the same. With CBM you react to a threshold that is being crossed now: the vibration reading is too high, so schedule an intervention. With predictive maintenance you go a step further and predict, from trends and models, when the asset will fail, so you can plan the intervention ahead. CBM measures the current state; predictive maintenance extrapolates that state to an expected failure moment. To see the basis of this classification, read preventive vs corrective maintenance. A broader background on the concept is available on Wikipedia's condition-based maintenance page.
The P-F curve: the heart of CBM
To understand why condition monitoring works, you need to know how failure develops over time. Almost no component goes from "fully healthy" to "broken" in a single step. There is usually a period in which the condition deteriorates measurably before the function actually gives out. You describe that development with the P-F curve.
The curve has two points. P is the point of potential failure: the first moment at which a suitable measurement method can detect that something is starting to go wrong. Think of a bearing that begins to show a subtle vibration pattern, or oil in which the first wear particles appear. At that point the machine still works normally, but decay has begun and it is detectable. F is the point of functional failure: the asset no longer does what it is supposed to do.
The time between P and F is called the P-F interval. This interval is the core of every CBM decision, because it determines how often you have to measure. The rule is that your measurement frequency must fall well within the P-F interval, usually at least two measurements inside that window. An example: if an incipient bearing defect announces itself on average three months before functional failure, you might measure monthly. Measure once every six months and you are likely to miss P and be caught out at F. Measure daily and you measure far more often than needed, wasting capacity.
Two things matter here. First, the P-F interval depends on the measurement method. Vibration analysis often detects a bearing defect much earlier than a technician feeling or listening by hand; a more sensitive method effectively lengthens the usable interval. Second, the interval is an estimate based on failure behaviour and experience, not an exact number. You set your measurement frequency against the shortest realistic version of it, with a margin.
What you measure in CBM
CBM depends on choosing an indicator that reliably responds at P. Commonly used quantities include:
- Vibration: the standard for rotating equipment such as motors, pumps, fans and gearboxes. Vibration spectra reveal imbalance, misalignment and bearing wear at an early stage.
- Temperature: elevated temperature points to friction, overload or poor lubrication. Thermography makes anomalies visible in electrical panels and mechanical components.
- Oil and lubricant analysis: counting and analysing wear particles, viscosity and contamination gives insight into the internal condition of gearboxes, hydraulics and engines.
- Running hours and load: usage-based indicators say something about accumulated wear and the conditions under which an asset operates. See running hours and monitoring for how to record this and link it to thresholds.
- Sound and acoustics: abnormal sound patterns, including ultrasonic, indicate leaks in compressed-air and steam systems or incipient mechanical problems.
- Others: current draw, pressure, flow, and corrosion or wall-thickness measurements, depending on the asset.
The choice of indicator depends on the failure mechanism you want to catch. A single machine can justify several indicators at once, because different failure modes announce themselves in different quantities.
When CBM pays off, and when it does not
CBM is not a goal you apply to every asset. It delivers the most under a number of conditions. There must be a failure mechanism that develops gradually and measurably, with a P-F interval long enough to react to. The failure must also carry consequences that justify the measurement effort: unplanned downtime, safety or environmental risk, or high follow-on damage. And a practical, affordable measurement method must exist that reliably picks up P.
If one of those conditions is missing, another strategy is often better. For failure mechanisms without a warning period, where the asset collapses abruptly, condition monitoring adds little; preventive replacement on interval or a design change fits better there. For parts that are cheap, quick to replace and fail without notable consequences, deliberate corrective maintenance is often the most sensible choice: you run it to failure and replace it then. And where a fixed interval is simple, cheap and demonstrably safe, there is no need to add the complexity of continuous measurement.
In practice you use the strategies side by side. You decide per asset and per failure mode which approach fits, usually based on criticality and failure behaviour. CBM is then the tool you deploy where gradual, measurable decay coincides with consequences you would rather not wait for.
Frequently asked questions
What is the difference between condition-based and predictive maintenance?
CBM reacts to the current measured condition: if an indicator crosses a threshold, you act. Predictive maintenance uses trends and models to forecast the expected failure moment, so you can schedule the intervention ahead of time. CBM is often the measurement base that predictive maintenance builds on.
How do I decide how often to measure?
You derive it from the P-F interval: the time between potential failure (P) and functional failure (F). Your measurement frequency must fall well within that interval, usually at least two measurements inside the window, so that you catch P reliably before F occurs. A shorter warning period therefore requires more frequent measurement.
Do I need expensive sensors for CBM?
Not necessarily. CBM starts with choosing the right indicator, not the most expensive technology. Periodic manual readings, oil samples or route-based vibration measurements already deliver a lot. Fixed sensors and continuous monitoring make sense when the P-F interval is short or the failure consequences are high.