Definition Predictive Maintenance

A faulty machine or a system failure is disastrous for a company’s daily workflow, but at the same time you don’t want to invest time and money in potentially unnecessary maintenance. Finding the perfect middle ground was possible for a long time only through experience and with a good dose of luck. However, due to the extensive data available to us as we continue to digitise, the ideal maintenance time can now be determined mathematically – predictive maintenance is born.

Predictive maintenance, to German: Predictive maintenance, is a concept that determines the maintenance time for equipment based on its actual condition. In contrast to maintenance intervals, which are coupled to fixed periods of time or service life, costs and downtime can be minimized.

The basis is a corresponding data situation, which allows such accurate forecasts. If this is not given, predictive maintenance would be reduced to a pure glass ball reading and extensive damage could be the result.

However, if the machines and systems used can be analyzed by sufficient sensors, the maintenance times can be determined with higher precision. If additional historical data are available, the benefits of predictive maintenance increase considerably once again.

Since the digital transformations have led to unprecedented networking of closed-off systems and the value of data acquisition and transmission has been generally recognized, a corresponding data base is available for the first time in almost all industries.

Definition of digitization

Digitalization is, quite soberly speaking, simply the transfer of formerly analogue processes to digital ones. Even if we are currently increasingly encountering these and similar terms, this is a very old and simple process, because almost every form of digitization is rewarded with efficiency increases, cost reductions and new, previously unknown possibilities. No wonder we humans have always been very interested in her.

Due to the accelerating technical progress and the mutual support (new technologies enable new technologies …) digitalization has gained so much speed in recent years that it has now penetrated into all areas of our lives and is indispensable from there. This digital transformation is a technological, socio-cultural, economic and intellectual process that brings with it gigantic upheavals.

For companies in particular, digitalization has created unprecedented opportunities – but it also lurks with considerable dangers, especially if it is ignored.


The aim of predictive maintenance is to find a time for necessary work that causes the lowest possible failures in terms of use and cost. However, since overloads of the equipment must be avoided in order not to risk damage, a corresponding approximate value must be determined, which should take into account all aspects.

On the one hand, periodic measured values can be used for this purpose, which can also be recorded by external devices. Older machines that do not have appropriate sensors from the factory can also be used in this way within a predictive maintenance protocol.

The significantly better alternative, however, is continuous measurement, which is possible through appropriately networked systems. This real-time data enables more accurate mathematical models to be created and the maintenance time to be calculated even more accurately – further savings in maintenance effort and costs are the result.

Based on the data obtained, whether by continuous or periodic measurement, the condition of the respective system is determined and a time corridor is formed within which maintenance must take place. Key data is the latest possible time without risking damage and the earliest possible time when maintenance would be worthwhile. This time window can be further clarified with better data by a higher number of sensors and historical data.

Once an ideal from-to-value has been calculated, aspects such as machine utilization, accessibility and cost of maintenance personnel, status of other machines and other available information are included in the calculation. The result is the time (or timing – maintenance work can be spread over a longer period of time) over a longer period of time, during which the work causes the least cost and other damage.

Measuring techniques

The effectiveness of predictive maintenance depends on the available data has already been mentioned. But how can this data be collected without disrupting operations?

A large arsenal of measuring instruments is available for this purpose: acoustic and infrared measurements enable the condition determination of a machine without intervening in the normal workflow. Vibration analyses are often more complex to implement, but are particularly worthwhile for high-speed work equipment. Sound detectors can easily detect changes in noise levels. Temperature, internal pressure, oil consumption… A long list of values could be compiled.

In the context of predictive maintenance, the output volume of the respective machine is also considered: by taking productivity into account and its temporal fluctuations, further statements can be made about the condition of the equipment, which may be important for the calculation of the maintenance period.


Industrial production of any kind is a worthwhile place of use for predictive maintenance, as it is also the most common 24-hour operation. This type of work makes finding the ideal maintenance time and the associated reduction of downtime particularly worthwhile.

But there are also numerous uses in the area of traffic and transport: the condition of a road surface can be measured by ultrasound and a section of roadway can be renewed months, weeks or even just days before the first damage occurs visibly. As road construction is subject to strong seasonal fluctuations, repair work can be relocated in times of low capacity utilization.

The same applies to rail transport and the determination of the condition of train facilities and track beds. Bridges of all kinds can also be replaced by a new building at the ideal time, provided that measurements are taken. This may delay the start of construction by several months or years and save corresponding costs. On the other hand, predictive maintenance protocols show structural damage earlier, so that life-threatening collapses could be better prevented. However, the effort for close monitoring by appropriate measurements is significantly greater.

Predictive maintenance systems are already increasingly being used in trucks and private cars. Thanks to the increasing networking of vehicle systems in the course of digitalization, as well as the now usual access to the Internet or at least the mobile network, corresponding maintenance recommendations can be calculated algorithmically. Connected cars thus provide an ideal basis for intelligent maintenance patterns.

Predictive maintenance is also used regularly in IT. No data center today would replace a component or even perform a major update without first calculating the ideal time based on many factors. These include, for example, the availability of one’s own personnel, the utilization of the device and the maximum time that one could wait before the maintenance has to take place.

Companies that centrally manage a larger number of computers also have similar considerations when making major software updates. For example, it is common for not all departments to receive software updates at the same time.