In brief
Predictive maintenance is a maintenance strategy that forecasts failures from condition and operating data and schedules intervention just beforehand.
Contents
What is Predictive maintenance?
In predictive maintenance, machines are monitored continuously or at short intervals, for example using vibration, temperature, current or oil condition data. Statistical models or machine learning methods detect patterns that precede damage and estimate the remaining useful life of a component. Maintenance then intervenes not at a fixed interval, but when the condition requires it and the production schedule allows it.
Predictive maintenance should be distinguished from condition-based maintenance: the latter reacts when limit values are exceeded, whereas the predictive approach forecasts when the failure will occur. In practice, the two approaches merge into one another.
Prerequisites are suitable sensors, a reliable database containing historical failure cases and clear responsibilities for the analysis. The benefit depends heavily on whether the predicted spare part can actually be procured in the time remaining.
In practice: maintenance and procurement
A forecast is only worth as much as the replenishment lead time of the part concerned. For monitored units, lead times, sources of supply and possible alternatives should therefore be clarified in advance, especially for discontinued components.
Synonyms and abbreviations: PdM · forecast-based maintenance · data-driven maintenance
Frequently asked questions
What data do you need for predictive maintenance?
The basis is condition data that indicates deterioration early, such as vibration, temperature, current draw, pressure or particles in the oil. Operating data such as load, speed and running time is just as important, along with a history of previous damage. Without documented failures, patterns can hardly be assigned reliably. Many projects therefore start with a few critical units and build up the database step by step.
For which equipment is predictive maintenance worthwhile?
The approach brings the most benefit for units whose failure is expensive and whose damage announces itself measurably, such as large drives, pumps, compressors or gearboxes in bottleneck equipment. For inexpensive components that can be replaced quickly, the monitoring effort often exceeds the benefit. Random failures without warning signs, such as some electronic defects, can also only be predicted to a limited extent using condition data.
What is the difference between preventive and predictive maintenance?
Preventive maintenance works with fixed intervals based on time, operating hours or quantities, regardless of actual condition. As a result, parts are sometimes replaced too early or too late. Predictive maintenance bases the intervention on the measured condition and a forecast of the time of failure. It makes better use of the wear reserve but requires measuring technology, data analysis and flexible provision of spare parts.