Predictive maintenance is quite simply THE “Industrie 4.0” theme for the German market. In Germany, production in many sectors has become highly automated, and therefore provides the perfect basis für this promising technology. To avoid this being written off as an unsubstantiated assertion, here are a few references:

  • Report of the German Federal Government “Skill and Qualification Requirements up to 2030“
  • Lufthansa statement: “Improved fleet maintenance with predictive maintenance “
  • Staufen AG study: “Industrie 4.0 Index 2017 “

We have been able to confirm the potential of predictive maintenance with the major vehicle manufacturing OEMs.

A precise analysis of the procedure and the transfer of the methods to strategy can be found in the following.



Breakdowns and faults often occur during operation, and need to be dealt with by the maintenance team. If a machine is idle and there is a break in production, this is a direct breakdown. On top of this comes the additional cost of other machines in the line which are also forced to stop production due to the fact that they are connected in series.

Maintenance is called out to settle the problem. Spare parts often have to be ordered. In acute cases, when express delivery is essential, the cost of these can escalate significantly.

In addition to this, there are also subsequent costs. If a line breaks down, it might be necessary to reschedule the production of other lines, as production is generally carried out in just-in-time und just-in-sequence mode.



Predictive maintenance uses the data created during the production process to detect irregularities in the data. This makes it possible for maintenance staff to take early action in production-free areas.

To this end, an international, multi-disciplinary team from EDAG PS has developed a method which, using sensor and control data, is capable of working out predictive maintenance solutions for any customer.

The sensor data is analysed by data scientists and technical/process experts in joint workshops. Working on the Pareto principle, according to which 20 % of the components and processes are responsible for 50 to 80 % of all breakdowns, an effective field of action for production stoppages is first of all identified. To begin with, the project will be restricted to this 20 %, and therefore promises great cost-benefit potential.

In later stages, a close look will be taken at every single process that is susceptible to faults, and the technical experts will then derive hypotheses from their findings, which will enable them to prevent faults.



Industrial robots have several servos. Typically, these exhibit irregularities in their power consumption data before a breakdown occurs.

Hypothesis: action can be taken before a breakdown occurs if the power consumption of a servo is under observation.

Fixtures usually have a large number of mobile clamps which are moved either pneumatically or by means of electric power. The control unit must receive confirmation that the clamps have reached their end positions, ensuring that a component is properly clamped, before work can be started on the component. These mobile clamps frequently become soiled or worn, which slows them down, until they open or close so slowly that it is no longer possible to adhere to the required production rate.

Hypothesis: if the opening and closing times are measured and a gradual deterioration is detected, maintenance staff can take action before production breaks down.



EDAG PS has an unbeatable advantage over IT companies and startups when it comes to the big data and predictive maintenance boom: thanks to our years of experience in the production environment, we are familiar with all production – and therefore maintenance – components and processes. The involvement of EDAG PS’s own in-house data scientists means that our comprehensive process knowledge is seamlessly linked with big data processes.



We work with you to develop your predictive maintenance strategy. This is based on the hypotheses we have put forward together and model-based condition monitoring.Signals from a wide variety of sensors are observed, analysed and compared with learned value limits, to identify any anomalies.

The observation of this condition monitoring is then taken over by machine learning algorithms.Together, we teach the system how to identify process problems itself. It does this by searching data for patterns, and can warn the employee of any irregularities it finds. In this way, the capabilities of man and machine are put to optimal use. The machine constantly observes data quantities that are so large that a person would soon lose track of things. Information is only provided when necessary. Maintenance staff can therefore spend more time concentrating on an area where man has an unparalleled advantage over machines: using technical knowledge to solve complex problems.

We will be happy to demonstrate how to get from the method to the strategy.



To make your first steps in predictive maintenance easier; EDAG PS offer hackathons, at which, in just a few days, you can work out your own predictive maintenance potential with us: quickly, cost-effectively and at little risk