Industry 4.0 Predictive Maintenance: Definition and Efficiency in Industry 4.0

Editor: M.A. Frauke Finus

Predictive maintenance describes a forward-looking approach in which machines and systems are maintained proactively and on the basis of permanently collected data.

Related Company

The term predictive maintenance describes a system initially developed in industry 4.0 that can be regarded as a core component of industry 4.0.
The term predictive maintenance describes a system initially developed in industry 4.0 that can be regarded as a core component of industry 4.0.
(Source: ©tashatuvango -

With predictive maintenance, downtimes can be effectively shortened and reduced. Comprehensive data acquisition in production is essential for this.

What is Predictive Maintenance?

The term predictive maintenance comes from Industry 4.0 and has become an integral part of today's smart production. Predictive maintenance is basically about using measurement data from machines and systems to determine maintenance intervals of the individual components and machines on the basis of this data.

The goal of predictive maintenance is always to maintain the machines and systems proactively and with foresight, so that malfunction times are minimised and the maintenance effort can also be reduced to a minimum. Ideally, predictive maintenance can be used to accurately predict malfunctions and problems so that a company can act before real failures occur and problems persist.

As a result of predictive maintenance, production times and the service life of the machines used can be extended, since maintenance is always carried out “on point”. In addition, excessive costs can be effectively prevented by unnecessary maintenance intervals. Predictive maintenance is an Industry 4.0 tool that strengthens productivity and effectiveness in production and provides companies with targeted information on the entire machine and plant park.

Predictive maintenance versus conventional maintenance

Traditional maintenance was usually understood in a reactive way. Although reactive maintenance is easy to implement, it is a significant risk to organisations. This is because only when errors and malfunctions occur is the system reacted to and the problem analysed and the necessary troubleshooting carried out. In contrast to predictive maintenance, reactive maintenance can therefore neither prevent nor predict malfunctions, often resulting in considerable downtime.

In the worst case, the urgently needed spare parts for the repair of systems and machines are not available and thus the downtime increases considerably. If such a case affects an entire production line or a machine that is important for the production process, this can in the worst case lead to economic problems for the company. Predictive maintenance is thus the safe and above all effective variant of maintenance, which should now be standard in Industry 4.0.

The advantages of predictive maintenance at a glance

In contrast to other methods of maintenance, predictive maintenance offers considerable advantages in Industry 4.0. Unplanned and therefore often very expensive machine downtimes can be avoided. Thanks to the comprehensive knowledge of the respective condition of the machines and their individual components, maintenance can be planned exactly as required, which noticeably and significantly increases the productivity of the machines. The field service assignments of employees can also be better planned and thus more efficiently arranged, since maintenance can be carried out in a targeted manner and at an intelligent interval. In addition, spare parts management can also be significantly simplified, since the required spare parts are already known in advance and can be held in stock accordingly. Furthermore, the continuous and constant analysis of the machines and the read-out data offers the possibility to successively increase the efficiency of the individual machines and thus optimise the utilisation. Therefore the machines can amortise their investment costs faster and thus contribute to the economic success of the company.

All in all, the advantages of predictive maintenance are enormous, even if the effort appears complex for many users at first. However, once the system has established itself and the data is not only collected but also specifically evaluated, predictive maintenance can achieve enormous leaps in performance for many machines and systems. Predictive maintenance is therefore once again a clear advantage for smart production, whose performance is to be adapted to current needs and current demand. After all, the scope of companies is increasing enormously.

Three important steps for the use of predictive maintenance

Those who want to rely on the principle of predictive maintenance in their own company must bear in mind that three important steps lead to success in the long term:

  • The collection, digitalisation and transmission of all relevant data
  • Storage, analysis and evaluation of the collected data sets
  • The calculation of probabilities for defined critical events

The first step is to create a database and implement the corresponding sensors on and in the machines. Since many manufacturers now operate in the area of Industry 4.0 and its standards, such sensors are already integrated in many machines and systems. In the course of this, however, all relevant data must also be collected by the experts. Those who only measure the bare data of the machine, but leave out values such as room temperature and air humidity, for example, can often not bring the data into a meaningful context. Predictive maintenance must therefore always collect as much data as possible.

In the second step, the data must be combined in a database and placed in relation to each other. Here, many providers on the market offer practical solutions that can be operated either in the provider's data center or directly in the company itself. This is where the quality of the data and its possible applications is decided. Because only through targeted and structured storage and fast access to the immense data sets can these be analysed by the intelligent algorithms.

The last step is essential for predictive maintenance. In this step, failure probabilities are calculated for all relevant components on the basis of the collected data. The better the data, the more accurately the probabilities can be calculated. Predictive maintenance is based on these probabilities and predictions. Before an event X occurs, component Y can be replaced in order to prevent event X.

Preventive versus predictive

Even if the terms preventive maintenance and predictive maintenance initially sound similar and are also used in a similar way, they still differ enormously. Similar to predictive maintenance, preventive maintenance also tries to avoid downtimes or keep them as short as possible. Preventive maintenance, however, does not collect any data, but determines the maintenance intervals according to a fixed pattern or experience. In the worst case, for example, wear parts that still function smoothly are replaced. In the long run, this causes considerable costs, as the material costs for the company increase without any concrete cause.

On the other hand, excessive wear cannot be detected either. If a component wears out particularly quickly and, above all, faster than the maintenance plan provides for, an unexpected failure occurs, which means further costs for the company. In the end, preventive maintenance is all about guessing as well as possible or estimating based on experience when replacement and maintenance would make sense. It is of crucial importance for the cost-benefit factor to arrange maintenance and repairs as early as necessary and as late as possible. These points become obsolete with predictive maintenance. This is because the condition of all relevant components can be checked on the basis of the data collected. This means that neither unworn components are replaced nor signs of wear are missed. Although the initial effort involved in this form of maintenance may seem higher, the data and data records collected can be used specifically to improve the performance of the systems.

This means that the data is not only available for predictive maintenance, but can also be used in a variety of ways within the company. In summary, this can be broken down into the following points:

Preventive Maintenance:

  • Defined maintenance patterns
  • Replacement of components independent of wear and tear
  • High costs due to high spare parts requirements
  • Can neither predict nor prevent failures

Predictive Maintenance:

  • Maintenance dependent on the condition of the machine or system
  • Targeted replacement of worn parts
  • Low costs due to fewer services and spare parts
  • Breakdowns are avoided and maintenance intervals are oriented to demand

Predictive Maintenance and Big Data

The great difficulty for predictive maintenance can be found in the processing and storage of the collected data records. For effective predictive maintenance, different data sets must not only be collected, but also stored, related to each other and processed by intelligent algorithms. Only through this combination can reliable predictions be made about the condition of machines and plants and thus enjoy the advantages of predictive maintenance.

This is compounded by the fact that the data collected can have completely different formats and variables. After all, not only the data of the machines and plants themselves play an important role, but also the associated environment variables. Temperature, humidity and air pressure can also play a role in many systems and their wear. In the context of predictive maintenance, enormously large data streams are thus collected, which must be updated and processed at regular intervals. Only in this way can trends and developments be recorded on the basis of the various core measurement data and made available for analysis.

However, this means that within the framework of smart industry and smart production, enormously large databases with huge capacities must be used, which can process the collected data at the appropriate speed. Basically, you should always bear in mind that the size of the database and the performance of the algorithms used have a lasting influence on the quality and reliability of the knowledge obtained. If we now look at the economic side, on the one hand there are the investment costs for predictive maintenance, which can very quickly assume very large proportions for many companies. On the other hand, however, there are the decreasing costs for maintenance, service staff and spare parts and at the same time the increase in productivity. If one compares these two cost centers, the non-recurring investment costs and the running costs for predictive maintenance appear to be significantly lower than initially thought. The larger the machine park and the better the capacity utilisation, the more the pendulum swings in the direction of predictive maintenance.

Examples from the world of work

A very good example of the use of predictive maintenance can be found in many vehicles. Thanks to the extensive data collection by many different sensors of a vehicle it is possible to reduce expensive breakdowns and repairs of a vehicle to a minimum. For this purpose, the sensors in the engine and chassis record a wide variety of data and compare them both with the optimum and with the previous data history. This means that any damage that may occur can be detected at an early stage and reported to the driver by the software. Vehicles with networked telemetry, which are able to report this data directly to the workshop or the vehicle manufacturer, are even more advanced. In such a case not only the vehicle owner can be directly informed about such a system, but also the responsible authorised workshop. This enables the authorised workshop to stock up on the required spare parts at an early stage and thus reduce the repair time to a minimum.

However, predictive maintenance is also becoming increasingly widespread in industry. Thanks to the sensors that are also installed here, vibrations, temperatures and machine noises, for example, can be permanently monitored. Even the smallest deviations are thus registered and can, for example, indicate the failure of a bearing at an early stage. In such a case, the bearing can be replaced in good time without further delays. Since predictive maintenance already knows which component is to be replaced in which area of the machine, maintenance time can also be minimised. The downtimes of the entire machine and also the working time of the service technicians can thus be reduced to a minimum.

Predictive maintenance in industry includes wind turbines and turbines with reducing turbine downtimes to a minimum. Thanks to intelligent mathematical algorithms, the vibration analysis of the various components can be optimally adjusted so that reliable predictions of the failure probabilities of individual components are possible. If these predictions are now combined with the wind conditions prevailing and the planned downtimes of such a plant, the replacement of the endangered components can be carried out at an early stage within the framework of predictive maintenance and thus without considerable effort. This saves time and money and also prevents a longer and unplanned breakdown of the entire plant.