Three ways to decide when to service a machine
Every maintenance team has to answer the same question: when should this machine be worked on? There are three broad answers.
- Run to failure. Do nothing until it breaks, then repair it. This is cheap in the short term and sensible for inexpensive, easily replaced items that do not affect safety or production.
- Fixed-interval servicing. Work on the machine at set times or hours, such as changing oil every so many months or rebuilding after a set number of operating hours. This is the idea behind preventive maintenance.
- Condition-based or predictive. Measure the machine’s condition and service it when the evidence says it is needed, ideally before it fails.
Predictive maintenance belongs to the third group. Its premise is that many failures do not happen suddenly. Parts typically show signs of wear in measurable ways for some time beforehand: more vibration, higher temperature, changed oil, a shift in electrical current. If those signs are caught, work can be planned.
How it differs from fixed intervals
A fixed schedule is simple, but it is a guess about average behavior. Two problems follow. Some machines are serviced earlier than they need to be, which spends labor and parts and sometimes introduces errors during the work. Others fail before the next scheduled date because they were worked harder or in dirtier conditions than assumed.
A condition-based approach tries to replace the guess with evidence. Compared with a calendar schedule, it may reduce unnecessary tasks and catch early faults. It also has costs: sensors, instruments, data handling, trained people and time to learn what normal looks like. For some equipment those costs are not worth it, and plain scheduled servicing remains the better choice.
In practice most plants use a mix. Simple, low-cost, low-risk items are run on schedules or to failure, while the most critical or expensive machines get condition monitoring. Ranking machines by how much their failure would matter, in what is often called asset criticality ranking, helps decide where to put the effort.
Common condition-monitoring techniques
Predictive maintenance is not one technology. It is a set of techniques that each reveal certain failure modes. Condition monitoring techniques commonly include:
- Vibration measurement, useful for bearings, gears, imbalance and misalignment on rotating machines.
- Temperature measurement, including thermal imaging cameras, which can reveal hot connections, overloaded components or blocked cooling.
- Oil analysis, which looks at lubricant condition and at particles in the oil that may point to wear.
- Ultrasound, which can detect leaks and some bearing and electrical problems by their high-frequency sound.
- Electrical measurements, such as motor current and insulation checks.
- Performance trends, such as a pump delivering less flow for the same power, which can signal internal wear.
Some are collected by a technician on a route with a handheld instrument. Others come from permanently installed sensors that report continuously. Both count as predictive work. Handheld routes are often a lower-cost way to begin.
From reading to forecast
Collecting a number is only the first step. The useful part is interpreting it. The usual progression is:
- Baseline: record the machine’s normal readings when it is healthy.
- Trend: track readings over time, since direction and speed of change matter as much as the value.
- Thresholds: define levels at which attention is needed.
- Diagnosis: work out what the change likely means, based on experience and the machine’s known failure patterns.
- Action: plan the work for a convenient time, with parts and people ready.
The word “forecast” should be treated carefully. Estimating how long a machine can keep running is uncertain, and an estimate is not a promise. Many programs simply aim to give enough warning to plan a repair, rather than a precise date. Some use statistical models to help estimate remaining life, but these depend on good data and still carry uncertainty.
What a program needs
Predictive maintenance works best when a few foundations are in place:
- Knowledge of what fails. Monitoring is only useful for failure modes that actually show symptoms. A failure that comes with no warning cannot be predicted this way.
- Reliable records. Without maintenance history, it is hard to tell whether a reading is unusual.
- Basic maintenance discipline. Lubrication, cleaning and inspection must still be done. Prediction on top of poor basics tends to find problems that routine care would have prevented.
- People who can interpret the data. Training, or outside support, is typically required.
- A way to act. If the repair crew cannot plan work, get parts or stop the machine at a suitable time, an early warning loses much of its value. Spare parts management and critical spares are part of this.
A maintenance software system, often a CMMS, helps track alerts, work orders and history in one place.
Typical pitfalls
Monitoring everything. Collecting data from every machine overwhelms people. Start with a short list of critical equipment.
False alarms and missed alarms. Thresholds set too tight create noise, and thresholds set too loose miss problems. Expect to adjust them as you learn.
Treating the tool as the program. Buying sensors does not create a program. The routine of reviewing, deciding and acting is what delivers value.
Overpromising. Some failures are sudden, and no method catches them all. Present predictive maintenance as a way to reduce surprises, not to eliminate them.
Getting started in a small way
A reasonable first step is to choose a handful of machines where a breakdown is costly, identify how they usually fail, and add a simple measure for those failure modes, such as a monthly vibration and temperature route. Keep notes, review what you find after several months, and decide whether to extend the approach. If readings never reveal anything useful for a given machine, that is also a result: it may be better served by scheduled servicing.
Whenever measurements involve getting close to energized or moving equipment, follow lockout and tagout style procedures, use trained and authorized personnel, and follow manufacturer manuals and local regulations. This is general information and not engineering advice.